This Trading Bot Is Too Good To Share On The Internet (but i do anyway)

Summary

Moon Dev reveals a powerful trading bot strategy that he believes is too good to be shared, yet he does so anyway, reiterating the importance of sharing knowledge and helping others. Throughout the video, Moon Dev stresses the significance of testing and continuously improving trading systems, inspired by the legendary algorithmic trader, Jim Simons. The strategy, which boasts overwhelming returns, undergoes rigorous testing to ensure its robustness against overfitting and market changes.

Highlights

• The trading bot shows an astronomical return, raising skepticism due to its 'too good to be true' nature. 🚀
• Moon Dev does extensive testing, including walk forward and permutation tests, to ensure strategy robustness. 🧐
• Jim Simons serves as an inspiration, highlighting the importance of continuous improvement in trading systems. 📚
• The results show significant variations in returns and risks across different market conditions and assets. 📊
• Ultimately, the strategy's robustness is validated, though it's clear that trading requires constant adjustment. 🔄

Key Takeaways

• Moon Dev believes in sharing trading knowledge to empower others, aligning with Jim Simons' philosophy. 🤝
• Automating trading can mitigate emotional pitfalls, providing a structured approach to investing. 🤖
• Rigorous testing is essential to verify the authenticity of seemingly too-good-to-be-true strategies. 🧪
• Trend filters and robust testing can highlight the real effectiveness of a trading strategy. 📈
• A good strategy thrives through continuous optimization and testing across various market conditions. 🔧

Overview

Moon Dev kicks off with an explosive declaration about the power of his trading bot. While he acknowledges its almost unbelievable success, he's committed to revealing its secrets on the internet, inspired by the ethos of sharing knowledge and lifting others. This stance is deeply aligned with Jim Simons' philosophy that the key to success is continuous improvement and sharing insights.

Throughout the video, Moon Dev dives into rigorous testing of his strategy. From out-of-sample testing to walk forward analysis, he leaves no stone unturned. Each test is a method to either validate or challenge the bot's exceptional returns, making sure there's substance behind the hype. The need for empirical testing is a core theme, as it weeds out overfitting and ensures that the strategy isn't just a flash in the pan.

In a field where secrecy is often the norm, Moon Dev's approach is refreshingly transparent. He emphasizes that helping others understand trading strategies can lead to mutual benefits. While the results are promising, he underscores that trading is never infallible and that vigilance and adaptation are vital to navigating the ever-changing market landscape.

Chapters

• 00:00 - 01:30: Introduction and Trading Philosophy The chapter introduces a trading bot that the speaker is enthusiastic about sharing online despite its potential effectiveness. The speaker believes in transparency and the power of code as an equalizing force, vowing to share daily findings regardless of their success. They express admiration for Jim Simons, renowned as the best algorithmic trader, emphasizing his achievements.
• 01:30 - 04:00: Overview of Trading Strategy The chapter titled 'Overview of Trading Strategy' provides an insight into the speaker's philosophy towards life and trading. They emphasize the importance of helping others achieve their goals as a pathway to achieving one's own desires, encapsulated by the phrase 'if you help enough people get what they want, you can have whatever you want in this life.' This mindset extends into their approach to trading, focusing on automation to simplify and optimize the trading process. The speaker's attitude reflects a casual and positive outlook towards life and trading, encapsulated in the repeated phrase '777. That's the vibe.' The chapter introduces the idea of achieving desired outcomes in trading by leveraging automated strategies, emphasizing the importance of mutual success and community support.
• 04:00 - 08:00: Testing and Optimization Methodology In this chapter titled 'Testing and Optimization Methodology,' the speaker discusses the testing and optimization processes used to evaluate a trading strategy. They mention the importance of not relying on results that seem 'too good to be true' and highlight the significance of further testing to ensure realistic outcomes. The speaker notes that while the Sharpe ratio might not be exceptional, the Sortino ratio shows improvement. There is a specific focus on Bitcoin, emphasizing that the strategy is heavily optimized. The exposure time of the strategy is 75%, indicating that the investment is active most of the time. The chapter stresses on looking at non-optimized returns for an accurate assessment of the strategy's performance.
• 08:00 - 13:00: Parameter Sensitivity Analysis The chapter explores a strategy that allegedly offers a 192,726% return, which vastly outperforms the common buy-and-hold strategy's 29,000% return. To evaluate this seemingly too-good-to-be-true strategy, various tests such as permutation tests, out-of-sample tests, and walk-forward testing are proposed. The speaker emphasizes the power of code as a great equalizer, encouraging transparency and sharing of strategies to collectively understand and evaluate their validity.
• 13:00 - 18:00: Robustness Checks and Recommendations The chapter titled 'Robustness Checks and Recommendations' seems to focus on the importance of positive thinking and mindset as foundational for success. It suggests that you can achieve anything in the world if you start with love and maintain a positive outlook. The speaker emphasizes the power of thoughts and suggests that maintaining positive thoughts is crucial, even though it's not a requirement for everyone. The speaker shares data related to trading, including a 64% win rate and mentions the issues of overfitting when using optimization libraries such as backtesting.py. There is a discussion on optimizing strategies and the associated returns, though the transcript cuts off before elaborating fully on the results.
• 18:00 - 23:00: Walk Forward Analysis and Results The chapter titled 'Walk Forward Analysis and Results' discusses the concept of overfitting in the context of financial returns, indicating that some results might seem too good to be true with percentages like 59.87%. The speaker plans to use a specific method labeled '03' to perform various tests, which are not specified in detail, to validate the effectiveness of their strategy. This involves a walkthrough of different tests that can be conducted to determine the viability of the strategy.
• 23:00 - 28:00: Comparison of Different Strategies The chapter discusses testing different strategies to determine their validity, with a focus on out of sample testing, walk forward testing, and permutation tests. The speaker expresses the need for analysis to decide which test to run first and mentions the availability of statistical data for nonoptimized returns.
• 28:00 - 33:00: Real World Application and Final Thoughts The chapter discusses real-world applications of automated trading and the importance of rigorous testing to validate past performance. It cautions against overfitting in backtesting and emphasizes the necessity of a daily routine to optimize trading strategies. The returns in trading can appear exceedingly large, prompting a need for thorough verification.
• 33:00 - 37:30: Algorithm Automation and AI Tools This chapter explores the process of automating algorithms and utilizing AI tools, particularly in the context of trading strategies. It begins with the author having a simple trading idea and testing its viability. The initial results appeared promising, prompting further optimization. However, the optimized results seemed unrealistically successful, prompting skepticism as advised by the author's mother. The chapter cautions against unrealistic promises of instant wealth through automated bots, emphasizing a critical approach to assessing such claims.
• 37:30 - 43:00: Additional Testing with Different Assets The chapter focuses on the importance of building your own trading strategies rather than relying on external systems. It introduces the concept of the RBI system, which stands for Research, Back test, and Implement. The author emphasizes the need to research trading strategies, back test them to evaluate their past performance, and then implement the successful strategies. The chapter stresses that using self-developed systems can provide a unique trading edge.
• 43:00 - 48:00: Conclusion and Future Directions The conclusion chapter discusses testing strategies and future plans for a developed bot after its positive back-testing results. The speaker emphasizes conducting further testing with small scale and exploring optimization methods to ensure robustness. They plan to perform systematic stress tests, guided by a detailed menu, to verify the bot's reliability, aiming to optimize and strengthen its future performance.

This Trading Bot Is Too Good To Share On The Internet (but i do anyway) Transcription

• 00:00 - 00:30 This trading bot is actually too good to be sharing here on the internet, YouTube, Twitter, whatever. Dude, I don't care though. I believe code is the great equalizer. And everything I find, I find it here live with you. So, I'm going to show you every single day, no matter what. I don't care how good it is, how bad it is. I don't care. I don't care. I don't care. I don't care. I'm going to show you everything. Jim Simons is the person that I'm following. Jim Simons is the best algorithmic trader to ever do this. 31 31.4 billion. Do you think I need 31.4 4 billion. No way. I
• 00:30 - 01:00 got the crib already. I'm good. So, I'm just going to chill and vibe out every single day and show you everything that I find. Good, bad, or ugly. I don't care. I believe co great equalizer. And I believe this, too, dude. This is what I really believe. If you help enough people get what they want, you can have whatever you want in this world, dude. 777. That's the vibe. I believe you can have whatever you want in this life if you help enough people get what they want. So, let's get what you want. You want to automate your trading? I did too, dude. Well, why? because dude, looking at these charts going up and
• 01:00 - 01:30 down, how do you trade this? I don't know. I don't know. So, I'm just going to show you everything here. Um, this is a too good to be true. So, we're going to do some more testing in here in order to make it um less too good to be true. You can see the sharp's not crazy, but the sortino is better. Of course, it's Bitcoin. The heavily optimized, if we look at the regular return here, let's just look at the the nonoptimized return here. Exposure time 75%. So most of the time it's in there, but you can see it's a 20
• 01:30 - 02:00 29,000% return for buy and hold, but 192,726% for for this strategy here, dude. It's too good to be true. So what I'm going to do today is I'm going to do everything I can in order to test it. Maybe some permutation tests, maybe some out of sample tests, maybe some walk forward testing. I don't know. We're going to we're going to try it all, dude. But um I believe code is great equalizer. That's why I share all this stuff. And yeah, dude, just keep giving.
• 02:00 - 02:30 You can have anything in this world if you leave with love. I promise you. I promise you, dude. You are your thoughts. They have to be positive, good thoughts. And you don't they don't have to be, but if you want a good life, 64% win rate. Okay, best trade 30 worst trade. The draw down is negative 63%. Now, we used optimize and you know overfitting is obviously a thing when you optimize with libraries like backtesting. py um but you can see here the return on the optimized one. I can't
• 02:30 - 03:00 even tell you what the return is. It's like 2723. It's it's overfit, right? U 59.87%. Okay, so this is all too good to be true. So, what I want to do here is I'm going to just use uh 03 today and I'm going to do a couple different tests. First off, I'm going to say this. Here's the code. Please go ahead and walk me through all of the different types of tests that we can run in order to see if this strategy actually works
• 03:00 - 03:30 because it's too good to be true right now. So I want to do some out of sample testing of course and I want to do some walk forward testing maybe some permutation tests but give me your analysis of what the best test is to run first and let's start from there. 03. Um, I want to give it some of the stats. Did I give it the stats yet? I don't know. Yeah. Okay. You can see all the returns and stuff of the nonoptimized and the
• 03:30 - 04:00 optimized here. So, you can see that the returns are just disgustingly big. That the returns are just disgustingly big. And it's too good to be true. So, we need to do all the testing. we can in order to um see if this actually works in the past. Okay. And the reason this is important is because when you optimize back tests, you can overfit them. And I follow this process each day. The process of automating your trading
• 04:00 - 04:30 starts with research of trading strategy. So yesterday I was just I just had an idea. It was a simple idea. Um and I just tested it and it looked really good. So I was like, "Okay, let me optimize Okay, so now it's optimized and now the stats are too good. Okay, so too good be true. My mom always told me if too good be true, it probably is. So just like anybody that's saying, "Hey, you can just go build a bot and make a bajillion dollars or you can just have my bot and make a bajillion dollars." That's too good to be true, dog. If it's
• 04:30 - 05:00 too good to be true, it probably is. Never run anybody else's bot. You have to build your own edge. This is a system I built. I use build my own edge all day. The RBI system, that's all it is, dude. process of automating your trading strate with research. That's the R. Research of trading strategies. R. Then back test those strategies to see if they actually work in the past. That's the B, the RBI. Back test. Back testing is the process that we just did here to find these crazy stats on this return on this strategy. Okay. Well,
• 05:00 - 05:30 after you back test and you see it actually works in the past, well, it might work in the future. So, since it's just a might, I built a bot with tiny size. Okay. So that's where we're at now, but I want to do some more testing in between and that's why we asked Chad TBT how we will go ahead and optimize it. All right, let's stress test this too good to be true monster in a systematic way. Below is a menu of robustness checks, why each one matters, what it tells you, and the order I'd run. Run with the pack. If
• 05:30 - 06:00 everyone is trying to solve the same problem or a whole group of people, that's the latest and greatest thing to do. Don't do that. Do something original. Don't follow the pack. Do something original. D hold out a sample test. Do this first. Make sure the edge wasn't learned exclusively from the data you optimized on. Split the history into training where you did your parameter search and a completely unseen test segments.
• 06:00 - 06:30 Typically splits 70% trained, 30% test or old regime versus recent regime. Pass criteria strategy on the test set should keep at least 50% of the train set sharp draw down profile should remain similar and trade distribution hit rate payoff shouldn't fall apart. Okay, so let's do that first. Let's go ahead and split this data here. Uh, this is a little ghetto, but whatever. You know, I'm a ghetto. I'm a ghetto dude,
• 06:30 - 07:00 so it just is what it is. Okay, let's go ahead and say um split data. Split data py. Okay. And here, please build me a tool that will split this data into whatever percentages I want. So, we can have a test set and a sample set. So uh if I said like 30% OS and 70% is then it would take the
• 07:00 - 07:30 first 70% as in sample and the last 30% as out of sample and then it would save it right into the same data folder starting the name with OS or is save it right back. Save it right back. Save it right back. Just keep running it back every single day. I don't care. Copy path. I don't care, dude. I don't care. I'm here forever. I'm not going
• 07:30 - 08:00 anywhere. I'm sorry. I'm locked. I'm always going to be locked. You guys are going to leave. I've seen you guys leave. There's less people here than there used to be. Why? Because people are [ __ ] they're kill. They're quitting crypto. Why would you quit? Why would you quit? Everybody's down bad. Let's step on the gas. Let's step on their head. Are you really a killer, bro? Are you really a killer? They down pad right now. Let's step on the guys. Let's step on the guys, dude. Come on. This is when we eat. This is when we eat, bro. Come
• 08:00 - 08:30 on. Come on, dude. Come on. Don't stop. Don't stop. I know you I know you had those thoughts. You want to stop now. Don't. Please. Thank you. Okay, let's run this. Let's split the data up. We got them split. Okay, let's say in sample first. Copy path. I know this is a little bit of cheat, but I'm going to test this some other ways, too. So, let's go over to the strategy here and let's put in the new um where is that? Where's the data at? CSV.
• 08:30 - 09:00 Okay. Okay. So, this is test. Okay. First off, hold up. Hold on. Let's do this. This is the in sample. And then I want to go ahead and save this data. This is the OG data. Okay, I know this is kind of cheating, but whatever. I just want to I want to get a bunch of test going. Okay. OG. Okay. CSV.
• 09:00 - 09:30 Okay. Let's say the uh data path here. Let's call this the OG. The OG. And then this will be data path equals that. Okay. Okay. Okay, now let's see how it does on this data here. Am I still a data dog or what? You still a data dog or what? Okay, so I'm just going to do a bunch of tests up here. So, first off, I'm going to go ahead and run this
• 09:30 - 10:00 one. So, this is the is Okay. And now let's run it. Damn, my fingers is quick today. My fingers is quicker. They quicker and quicker every day, dude. Okay, that's what the equity curve looks like. So, you can see it. It's optimizing
• 10:00 - 10:30 now. And I'm just going to be quick with it. And I'm not going to overoptimize it. I'm not even going to look at the optimization. Okay. So, you can see it's a 73,000% return versus buy and hold of 7,000 draw down. Okay. So, this is unoptimized because if you optimize it, then it's just going to be less real. I mean, you know, that's I that's how I interpret it at least. I'm only on year four, though. So, if you got ideas, let me know, please. I'm here I'm here just learning. Just learning,
• 10:30 - 11:00 showing everything all the time. Okay. 188 winning trades. Okay. So, that's the in sample and then we'll go ahead out sample. See how I do. Um I haven't optimized it. This is literally the first idea I had um with 10%. Okay, stop loss and take profit equally. It looks like it's profitable, but the draw down is heavy. The draw down is heavy, but here in the optimized one, the draw down is not as heavy.
• 11:00 - 11:30 Let's go look at it. Max draw down - 26. So, you know, maybe you can get it lower. Just be careful with overoptimizing. That's why I run all my bots with tiny tiny size because just cuz it worked in the past does not mean it's going to work in the future. And you know that dude, maybe you don't. So now you do. Wow. Wow. So so much knowledge coming at you every single day, my Wow. Ow. The sample copy
• 11:30 - 12:00 path. Wow. CSV. Okay. Okay. In sample, out of sample, data pi. That's right, son. And then let's copy this. Copy. And then I'm gonna have the AI talk to me about
• 12:00 - 12:30 this. All right. Let's see if this still works on the other data. Okay, it does. You can see that still works. Um, final peak max draw at 77. Draw down days is 124. Okay, that's pretty that's pretty heavy V. That's heavy B. That's heavy V. That's heavy
• 12:30 - 13:00 B. Okay, I'm not going to let it optimize. Let's see how this did over here. Proper factor 1.2 expectancy one okay sharp 0000 damn exposure time return is ne4%. So it didn't work. There we go the
• 13:00 - 13:30 flop. Okay. So here we go. I did an in sample and out of sample and that's why you got to do those dude. You got to test them on different datas. Okay. Um Okay. You can see that my in sample did well but my out of sample didn't. Please um recommend what the next step I should have uh what what's the next step I should do in order to keep uh robustness testing
• 13:30 - 14:00 this. I did a 70% train and a 30% test. Okay, let's go ahead and see what she has to say up here while she's doing that as well, my dear. Yes, please. Thank you. Permutation test walk forward analysis rolling reoptimization here. Goal: Simulate how you'd really run it. Predict periodically retune parameters then trade the next chunk. Pick a window eg 3 years and a step of
• 14:00 - 14:30 six months loop. Optimize on window t. Trade on out of sample in window t +1. Rule four. Compile the stitch together equity curve for all OS slices. Parameter sensitivity and heat map check. Make sure the edge is a plateau, not a single unsustainable spike. That's actually a good idea. That's a good idea. Run a grid MA fast
• 14:30 - 15:00 slow TPSL a wide neighborhood of parameters within say 10% best sharp razor thin equals over fit. Um exposure time in sample the buy and hold is 71% versus -14.9 in the last three years. four years really 12 thou 1300 days
• 15:00 - 15:30 exposure time 77%. I want to test it on some other data as well. I think that would be interesting. Let's try um let's try this 1 hour 5,000 weeks. Is that what it says? 500 weeks. 5,000 weeks. Jesus data path. I've done it on hourly or daily daily data thus far. So now I'm gonna try it with
• 15:30 - 16:00 hourly. There it goes. Oh, bad, bad, bad, bad, bad. You can see it falls off a cliff early. Let's try it on some six-hour data here. No, it's it's what's called machine learning. So you find things that are predictive. You might guess, oh, such and such should be predictive, might be predictive, and you test it out in the computer, and maybe
• 16:00 - 16:30 it isn't, maybe it isn't. Test it out on long term. All right. So, I'm grabbing six hour BTC data real quick. And uh price data and other things, and then you add to the system this if it if it works, and if it doesn't, you you throw it out. So, throw it out. There aren't elaborate equations, at least not for nothing elaborate around here, dude. What time is it? 9:09. We got a bit of time until the stock market opens here. It looks like Tesla is up a little bit. QQQ is down a little bit. Nvidia is
• 16:30 - 17:00 uh down a tiny bit. Ethereum is steady. BTC is up a little bit. Okay, so we got about 20 minutes. So that happens. Uh you can see I'm getting six-hour data here for a thousand weeks. So that's a lot of weeks, but I'm going to do the uh this I'm going to use it on the split. I'm going to use it on the split again.
• 17:00 - 17:30 Let's go ahead and read what she has to say. Permutation test, randomization test, aka scramble the labels. Quantify data mining bias. Compute a p value. Break the link between price mo movement and entry logic while preserving distributional properties. Two common flavors. Return shuffle. Randomly permute daily returns. Keeping order of signals. Signal shuffle. Keep return path intact. Randomly permute your long,
• 17:30 - 18:00 short, flat signals. Run the strategy. N equals 1,000 times on shuffled data. Build distribution of equity finals. Sharp. See where the sharp See where the sharp sits. Live sharp should be in the extreme right tail. 05 Monte Carlo resampling of trades. Understand luck verse skill by bootstrapping trade level results. Sample with replacement from your realized trade list. Rebuild equity
• 18:00 - 18:30 curve 10,000 times. Evaluate distribution and kagger max draw down. The median bootstrap performance should stay positive and the fifth percentile should not wipe you out. You can see I got the data here. So, let's go ahead and data dog it. Let's go ahead and split it first. No, no, no. Let's not split it because we haven't even seen it works yet. So, well, let's split it. Let's split it. Let's be careful. 6 hours. Copy path
• 18:30 - 19:00 here. And let's go to that uh data split here. Um split data here. and we'll go to that split data part here and see what happens. Here we go. Let's split it first. Okay, it's split. Now, let's go get the data dog. 6 hour in sample out of sample. Copy
• 19:00 - 19:30 path. Coinbase data. Close that. 2x. Let's try this one here. The 1 hour is trash. Most of it's going to be trash. Remember that. Okay. So, this one looks like it. Um, we'll see. This one's profitable. 2.61 expectancy. Okay, that's in sample 6 hours. Uh, 72 days. Okay. 296 trades.
• 19:30 - 20:00 The six-hour data sharp ratio trash sortino though 3.13 remember sharp penalizes for the upside and downside exposure time 86 return 37,000 verse 8,600 8,600 Okay, so that looks pretty good on the 6 hour, but let's go ahead and try the 6 are out of
• 20:00 - 20:30 sample. That's actually right, dude. Good job. Good job, dude. Okay, so here we go. Worst trade, best trade. You can see this is profitable in sample and out of sample. There we go. Sharp is still low. The return is 161 verse 216.
• 20:30 - 21:00 So, not better here in the last few years. Interesting. Um, I should have written those down. I can run them again. I'm really curious about actually the optimization of this one. But you don't want to optimize on the out of sample. Okay. So, 6 hour looks decent. Decent. It's a little too high exposure time in my my liking,
• 21:00 - 21:30 but 6 hour OS 6h hour is. So, I flap that
• 21:30 - 22:00 back. Okay, you can see this one was profitable. 37,000 verse whatever it was. Okay, so sweet, sweet, sweet, sweet. I kind of want to get an optimization going for this six-hour one. So, I'm going to do that copy path. I'm just curious, you know? I'm just a curious cat. Because I'm using really random and this is good that I use random, but um constants. I'm not
• 22:00 - 22:30 even I haven't I don't want to overoptimize, but I also I'm I just put like random constants in there like take profits and stop losses. So, let's go ahead and say Python. run that. Oh, I'm going to have to activate TFlow, bro. There we go. So, let let's let it optimize here. I'll put it down here on the screen. Um, and we'll come back to that.
• 22:30 - 23:00 So, currently optimizing the 6 hour. Wait a minute, though. Wait a minute. Wait a minute. Wait a minute. 10 to 30. 10 to 40. Take profit. Nah, this ain't it. This ain't it. This is two cuz we noticed that the
• 23:00 - 23:30 um lower take profits do better. So, lower take profits and stop losses. Here we got 9:16 a.m. 9:16 a.m. Let's flip this to a uh more fun chart. Maybe the 5 minute so we can watch this open here in I don't know 14 minutes. Yeah. QQQ, Tesla, QQQ down a bit. Nvidia, BTC, Ethereum. Okay, you seen
• 23:30 - 24:00 everything. Let's go ahead here and say let's see what she said here to this response. If number one looks good, quick out of sample split code sketch. Okay, we already did that. Your OS flop equals early warning that the edge you saw was probably SMA curve fitting alpha. Don't throw the idea away yet, but switch from searching for the perfect numbers to proving the concept can survive rough handling.
• 24:00 - 24:30 Below is the logical next move and how to run it. Stress the parameters before anything else. Why? When you optimize four knobs at once, a single lucky combination can beat out a sample even if the underlying logic is okay. Here's what to do next. Step 1 a. Build a dense grid around the is champion plus 10 bars. Produce a heat map of sharp on the IS.
• 24:30 - 25:00 Run every grid point on the same OOS slice you already used. Run every grid point on the same OOS slice you already used. Okay. So, um, send me the code for the full code for 1 A and then 1B as well so I can
• 25:00 - 25:30 test both those.
• 25:30 - 26:00 here's a single drop in script that runs both robustness steps. Okay, step 1 A, dense parameter sweep on your in sample data and a sharp ratio heat map. Okay, step 1B. The identical grid of your out of sample data with another heat map so
• 26:00 - 26:30 you can eyeball whether the edge survives. Just copy paste into robust SMA plateau. Adjust the two file paths if needed and run. Robust SMA plateau parameter robustess test. make a choice, right? You just decide what it's going to be, who you're going to be, how you're going to do it. Just decide. And then from that point, the universe is going to get
• 26:30 - 27:00 out your way. It's like it's water. It wants to it wants to move and go around stuff, you know. Produces two PNG heat maps. Okay. So, we got the path for the end sample and the OS. Okay. All right, let's go ahead and try to run this. Um, I'm going to call this 2x MA reversal robust
• 27:00 - 27:30 SMA. All
• 27:30 - 28:00 right, let's run this and see what the error is cuz you knew there was going to be an error. Copy path. bug please fix. thinking.
• 28:00 - 28:30 You ran into the version mismatch back testing py that should uh that shipped
• 28:30 - 29:00 with anacondo tflow env. The keyword return history is not recognized by back test.run newer releases added it. Okay, simply drop that keyword. Okay. Here we go. So, let's wait for this to be done. In the meantime, let's see if we can read a little bit more.
• 29:00 - 29:30 What's up with the chat today, dude? I've been using some GPT like Lauren and Python code to code some things and I don't know if this is good AI. Can you help me out? Guide me. Sure, dude. Use cursor. Use cursor. Jeremy
• 29:30 - 30:00 says, "My bot is stuck trying to find radium pools. Everything is running, but it can't make actual trades. Nice, bro. Yo, Art, what's up? 777KW2. Give thanks to the 777 for he is good. Be kind and compassionate to one another. Yes. 777. Reek out for the for I know the plans I have for you, declares the
• 30:00 - 30:30 777. Plans to prosper you and not to harm you. Plans to give you hope and a future. All you need is hope, dude. All you need is hope, dude. All you need is love. What does 777 means, dude? I got AI. That's beautiful. Do you have some prompts? I got mad prompts in my brain, bro. My brain is full of prompts. What type of prompts you need, cousin?
• 30:30 - 31:00 Mano de said ei me no way I don't speak Spanish but sometimes I pretend like I
• 31:00 - 31:30 do program mayor Okay, maybe you don't speak Spanish actually cuz I thought it was program. But hey, much love to you regardless. Glad you're here. Um, let's see how this does. Okay. Does it look like it's done? Is champion. I can't read this trash, bro. Can you? No. No. No. I think there's there's a bug or something
• 31:30 - 32:00 here. Oh, there we go. There we go. Oh, this is uh this is the uh this one, the optimized one. It looks like it's trash. No, it actually looks really good. But look how many trades there is. There are SQN profit factors. Unbelievable. 1,600. Yeah, dude. This is too good to be true, so I'm just going to probably
• 32:00 - 32:30 pass on this one. But um you can see the return is just trading perfectly. Can see all the red and green dots here. All right. So what I want to do though is take this and I want to do that in sample out of sample on the best params.
• 32:30 - 33:00 Okay, heat is saved. Perfect. Heat.I. is saved. So, let's go look at it.
• 33:00 - 33:30 and sample heat on the left. Add a sample heat on the right based off
• 33:30 - 34:00 sharp. I mean honestly I think I want to try this on um I'll just tell her please help me analyze these heat maps that sent back. Also make sure that it's saved to the data folder. We don't want to save to where you saved it. Um, also, uh, this is crypto data, so I
• 34:00 - 34:30 think it would be beneficial to look at the Sortino heat map as it doesn't penalize the downside or upside like, uh, Sharp does. Um, one more other thing that could be cool to look at is the calm ratio heat map. Um, I don't know though, you tell me. 03. What you got, fam?
• 34:30 - 35:00 So, it thought for 14 seconds in sample.
• 35:00 - 35:30 Okay. Can you see these? Probably not. Okay. Let me pull this screen up a little bit so you can see it too, dog. Cuz you're a data dog. You're a data dog. We're both dated off. Let's make this a little bit smaller so we can get all on screen here. You can see that fully. Yes. Yes. Dude, you got to leave. Peace, bro.
• 35:30 - 36:00 Mila, peace. Peace, programmer. Okay, let's read through this analysis here of of this good good good good feller in sample is yes a big yellow green plateau from fast MA 7 to 20 verse slow MA 30 to 48 that's what you want to
• 36:00 - 36:30 see out of sample edge clearly survives but shifts downward sweet spot is se fast 7 to13 verse slow 30 to 45. So how do you interpret this? Okay. So you can see the slow MA here the fast MA. Okay. We're looking for higher numbers here. So the yellow blocks. Okay. And over here we're looking for
• 36:30 - 37:00 higher numbers as well. The sharp is 25 35. Okay. Lots of parameter combos work. Yeah. Everything you wanted. everything. Imagine you were everything you ever wanted to be and capture that feeling. That's the easiest way to manifest what you want. That's the easiest way to manifest what you want.
• 37:00 - 37:30 Still a respectable plateau sharp 0.9 to 1.3 takeaway the idea is alive but the fast MA must be shorter than 15 for live trading that gives you a clear rule for the walk forward test restricts search
• 37:30 - 38:00 space why bother with sortino and kmar sortino ignores upside volatility great for moonshot crypto where big spikes are normal. Calmar is the return divided by draw down punishes long deep draw downs. Perfect if you care about capital efficiency. Both metrics complement sharp. So plotting them makes sense. If the sortino and calm plateaus line up with sharp, the edges robust. If they
• 38:00 - 38:30 diverge, you'll know exactly which risk dimension needs work. Drop in upgrade. Save to your data folder and plot sharp, sortino, and calmore. Below is a patched version of the robustness script. Saves everything to robust test. Collect sharpino calm in one loop. Nice. spits the heat
• 38:30 - 39:00 maps. Beautiful, beautiful, beautiful, beautiful. All right, so let's run it back. Beautiful, beautiful, beautiful. Life is beautiful. I'm going delete all
• 39:00 - 39:30 these. Books are freaking flying, son. They're They're flying. They're flying. Books are flying. Longs and shorts, dude. Longs and shorts. Longs get licked. Shorts get licked. You know the game. Long get licked, shorts get
• 39:30 - 40:00 licked. All right. 21%. Oh, the market just opened. Okay, let's go. Peep game. Excuse me. Excuse me. Sorry for burping in your ear. I hope you didn't hear that. QQQ at -21 21. Tesla
• 40:00 - 40:30 here 50% done. It says 50% 50%.
• 40:30 - 41:00 right
• 41:00 - 41:30 here. Yeah.
• 41:30 - 42:00 Awesome. Awesome. Awesome. Awesome.
• 42:00 - 42:30 Awesome.
• 42:30 - 43:00 I was just grabbing some cheese,
• 43:00 - 43:30 bro. I was just grabbing a little
• 43:30 - 44:00 cheese. Give me a break. I had to get some cheese, bro. I had to get some cheese, dude. Little mouse. You know what I'm saying? Snack on some cheese and breakfast. Breakfast cheese. Why not? Why not have a little cheese for
• 44:00 - 44:30 breakfast? Why not? All right, I'm going to send all this to AI. So, where is she? She right here.
• 44:30 - 45:00 Please walk me through all of the data that we got back here and then uh teach me your conclusions. All right, let's look at them, too. That was sick, bro. All right, so
• 45:00 - 45:30 uh yellow's good. I mean, all these are fire. They're like over 10 camo. Not all of them. A lot of them are fire. Here we go. Oh, this is a big data file. She ain't going to be doing nothing with
• 45:30 - 46:00 this. Okay, so what the three metrics mean? Quick refresher. Nice. That's a good one. Sharp is the total return divided by the total volatility. Good for traditional assets. Crypto specific nuance. It penalizes savage upside spikes just as harshly as down draw downs may undervalue
• 46:00 - 46:30 moonshots. Sortino is a return divided by downside volatility. So sometimes better for crypto ignores those huge positive candles better for BTC. Kagger kmar is the return divided by max draw down. Kagger compound compounded annualized growth return or something tells you how efficiently equity compounds given the single worst hole you have to crawl out of. Rule rule
• 46:30 - 47:00 of thumb sharp above one toino above two calar above one in out of sample land. Interesting. Keep testing. Reading the heat map you produce 2.1 sharp. So the big yellow green block from fast. We already saw the sharp. So 2.2 is sortino. We must keep moving. Let's go get those. We must keep going. If you can't fly,
• 47:00 - 47:30 run. If you can't run, walk. If you can't walk, crawl. But by all means keep moving. So you can see here this is the in sample sortino. There are some spaces with 70 plus sortino which is nuts. This is the out of sample sortino. And wow it's a big difference. Okay let's read what she says. Same band 7 to 20 30 to 48 but a few neon yellow
• 47:30 - 48:00 outliers. Sortino equals 70 plus. Wow. Those are almost certainly statistical artifacts. Tiny max draw down periods with goofy upside. With goofy upside, dude, that's goofy. You goofy, bro. That's funny. Add a sample. Sortino. Healthy green ridge. Sortino equals 2.5 to 4.5. Sitting on the same fast 7 to 13. Slow equals 30 to 45.
• 48:00 - 48:30 Strip. So the positive skew edge survives once we ignore the upside noise calm ratio here is calm severe several come on let's look at it first so we're going to look at the calmar ratio in sample looks like this you can see that um there's a couple yellowish green spots those are like the outliers then the in sample Is
• 48:30 - 49:00 this uh sorry out of sample? Let's see what she
• 49:00 - 49:30 says. The in sample kmar several ridiculous 40 plus values at fast 20 to 30 23 with slow equals 33. That's one monstrous run with almost no draw down. Ignore it vanishes. Oh, out of sample oscar. The only cells with three plus sit at 79 slow 40 to 45. Everything beyond fast equals 15 collapses to calm. What the trio tells you edge is
• 49:30 - 50:00 real but parameter sensitivity. Keep fast MA under 13. Sweet spot is 7 to 12 and slow between 30 and 45. Overoptimize pocket around fast equals 20. blew up. OS those high-end sample sortino calm spikes were miragees. Good. We caught
• 50:00 - 50:30 them early. Nice. Risk adjusted reward is still attractive out of sample. Sharp equals 1. Sortino equals three. Kalmar equals 1 to three for a wide band. Passes the plateau test. Concrete next step stage. Okay. Walk forward reoptimization. Y simulates live retuning and trading practical settings based on the heat map. Rolling three-year in sample trade next 6 months out of sample grid fast 7 to 13 slow 30 to 45
• 50:30 - 51:00 TPSL keep plus one minus one band. Okay. Permutation test estimate P value verse data mining. Okay. Keep the signal squee sequence fixed. Shuffle returns 1,00 times. Look for P under 005. Trade level bootstrap. Luck verse skill and tail risk. Resampled trades 10,000x. Fifth percentile kagger should
• 51:00 - 51:30 stay above zero. Transaction cost shock. Crypto fees spread. Double commission plus ad 0.25 25 slippage ad should beat the hold files and folders. All your outputs are now dropped into this. Okay. Feel free to rename or move back testing code. We'll pick them up anywhere, but this keeps the data self-contained. Verdict, good news. The idea holds water once you keep the fast
• 51:30 - 52:00 MA short. Caution. Avoid the tempting fast equals 20 cluster. Pure curve fit action. Enroll the walk forward test with the restricted grid if it delivers sharp about equal to one or or calm. Please send me the code for the walk forward test that you want me to
• 52:00 - 52:30 run and then we will go from there.
• 52:30 - 53:00 Here's a ready to run walk forward script that rolls a three-year optimization window, trades the next six month slice with the windows test parameters. Stitches every out of sample slice into a single live like equity curve. Prints slice level and aggregate stat stats. Drops all artifact artifacts into the data walk
• 53:00 - 53:30 forward. Save it as walk forward ma reversal or overwrite an old file. Then hit run. Done. Done. Done. I'm going to call this um copy. Paste. Walk
• 53:30 - 54:00 forward. Paste it in. Damn. I hit copy again, bro. I hit copy. I hit copy. I meant paste. All right. So, you can see we're using one day. Okay. Train. It's going to split it. Fast MAS, slow MAS. Okay. 02 commission. So, double double the commission because, you know, better than no
• 54:00 - 54:30 commission. I'd rather use double commission than no commission. Bug, please fix. Send me full code. Thank you. You're using no commission. I'm using double commission. Bug, please fix. Send me back the full code. Thank you.
• 54:30 - 55:00 Looks like whiff making a move, huh? How big a move was that for the last week? What? What's up, bro? How you
• 55:00 - 55:30 doing? That's cool. I didn't know you could do that. I'm chilling. How you doing? That's dope. I didn't know you could just start typing on this. No, you can't just start typing. But below is a drop in replacement. Thank you.
• 55:30 - 56:00 All right, here we go. Looks Gucci now. Sharp 1.49.
• 56:00 - 56:30 49 sharp 2.04 04.
• 56:30 - 57:00 All right, here we go.
• 57:00 - 57:30 So don't waste it living someone else's life. Don't be trapped by dogma, which is living with the results of other people's thinking. Don't let the noise of others opinions drown out your own inner voice. And most important, have the courage to follow your heart and intuition. They somehow already know
• 57:30 - 58:00 what you truly want to become. Everything else is secondary. Stay hungry. Stay foolish. Stay hungry. Stay foolish. Stay hungry. Stay foolish. All
• 58:00 - 58:30 right, let's see how this did. I mean, we can see the stats here, but I want to give it to the AI 03 here because she's smarter than I am. Please go ahead and walk me through the data that we got back. Explain to me how it looks and how
• 58:30 - 59:00 I can interpret it and how we can move forward from here. So let's go to that file here that we had the data. It's actually not in robust. It's in walk forward equity curve. We can go ahead and reveal and finder here. And I'm just going to toss this all into this lady here. See what she has to
• 59:00 - 59:30 say. Then also, let's go ahead and open them up so we can use our tiny brains as well. Tiny brain, big heart. Tiny brain, big heart. If anybody knows where what that's from, you are permitted to let me know in the chat, please. Tiny brain, big heart. It's a little different than what was actually said, but tiny brain, big heart. If you know, please put it in the chat because I'll know you're a real
• 59:30 - 60:00 one. But if you don't know, still, okay, you're still dope. So you can see the walk forward stitched equity here. Okay. Curious as to why you're using Python for HTF when it's subpar and quite slow. Um high time frame. Did you mean HFT? Did you mean high frequency
• 60:00 - 60:30 trading? I I don't know how to respond until I know what you actually meant. So much love to you. Thanks for the question. Actually, it's not even a question. This a statement. But, you know, I know developers are like like that. Let me just make a statement, not ask a question. It's all good, bro. You mean HFT? Is in what way you connect AI generated code to running Python without
• 60:30 - 61:00 copying and pasting the code every time? I just copy and paste the code or I use cursor. Cursor is here. I've been using 03 today and you know 03 is like 30 cents per run inside of cursor. This is the AI IDE. Um so I just you I'm just using my credits here and I'm copy pasting it. Okay. What you just ran? Well, let's look at it with our eyeballs first. We saw you can see the equity here. I think this is equity. Yeah,
• 61:00 - 61:30 equity here. So it goes up. That's good. Um, the Calar here. Let's go ahead and look at this. So, this is train start, train end. You probably can't see this, huh? You're welcome, bro. Guess my other message didn't go through. Yes, Python for HFT. No, I don't do HFT. That's the thing. So, um, HFT maybe in a few years. Uh, the way I look at it is the lower time frames are easier to for us newbies to get off on.
• 61:30 - 62:00 So, I've only been at this game for four years. Um, I'll be here for a long time and I feel like you just go smaller and smaller time frames over time. It's more competitive HFT. And I I I may agree with you when I get there that Python's too slow. I do know Rust though, and Rust is super fast. So, if I ever run into a problem with latency, a better word for latency is speed. If I ever run into a speed problem, I will probably use Rust as I know that language. Um,
• 62:00 - 62:30 but yeah, I'm not doing HFT. Some people want to trade 2,000 times a day, and that's not me. Not me. Not right now, at least. Maybe later. Maybe later. Okay. So, you can see this set here from 2016 to 2019. This set 2015 to 2018. This set 2016 2019 max draw downs here -27
• 62:30 - 63:00 -12744 -39 cals 268 3.6 Sortino 124 sharp 1.4 four three. Okay. So, I'm curious to see what she has to say. And also, if you like to give an opinion on 03 compared to 04 compared to 01 dot dot dot. I don't know. I'm just using the newest ones. I got no opinion.
• 63:00 - 63:30 How to read the results. Equity curve. Huge spikes. Flat lines spikes. That saw tooth is expected because every sing every six months every six months slice reinitializes capital to 1 million for independence. 2020 to 21 bull run slice
• 63:30 - 64:00 rocketed to $16 million. Then the slice starts again at 1 million. The plot plunges which the max draw down routine counts as 96% draw down. Later slices trend up more smoothly. 2023 to 24 takeaway. Visually useful for seeing per slice
• 64:00 - 64:30 behavior. Slice level metrics a bridged here. So trade window you can see the return from 2019 January 2019 to July 2019 483% 1.49 49 sharp 269 calm 74 trades early BTC bull
• 64:30 - 65:00 pop then 20 to 21589 1.3 sharp 2986 calm 82 that's a 2020 parabolic run 21 to 21 so 1 to 7 again negative8% return zero sharp Sharp zero calm 46 trade flat chop after the blowoff
• 65:00 - 65:30 top 2022 07 to 2023 01 negative9 return zero sharp zero calmar 35 trades FTX driven bare shock 20237 to 202401 one 109% return 2.43 sharp 44 calm 57
• 65:30 - 66:00 trades post ETF rally then the recent 6 months here is 41% return 1.21 21 sharp calm 4.1 number of trades 48. All right. So average sharp is 1.04 across all
• 66:00 - 66:30 slices. Edge survives regime changes four losing windows returns from8% to -19% cap the mean kagger at 4.9 in the naive concentration concatenation kmar blows up for the 2020 parabolic slice because max draw down inside that six-month block was tiny once we reset the 1M the metric loses
• 66:30 - 67:00 Meaning parameter stability.
• 67:00 - 67:30 You have to in a business like
• 67:30 - 68:00 this just keep making things better, keep improving the system because other parts of it are going to wear out after a while. People will catch on to this or they'll catch on to that. So you you just have to like in any business in any business you just
• 68:00 - 68:30 have to make things better and better and better because that's what everyone else is trying to do. Every slice picked MA fast of 13, MA slow of 33 or 42, stop loss of 1%, stop loss of 3%.
• 68:30 - 69:00 What the did Siri just pop up on my mother freaking computer, bro? About Riverside, California. What? Dude, what? I don't want to know anything about that place. I literally don't want to know
• 69:00 - 69:30 anything about that place. Why would you ever tell me that, Siri? And why are you even here? I didn't even know you were on my computer. The [ __ ] That was crazy, bro. What these numbers really say. Concept survives real walk forward. The sharp is over one with a tight parameter neighborhood promising risk presentation is
• 69:30 - 70:00 distorted. Minus 96. Draw down comes from equity teleporting down when a new slice starts. If we compound slice equity next slice starts with previous ending capital, the true max draw down and kagger will sit between the individual slice stats and the naive contamination. Edge is regime dependent but not fragile. Positive slices align with trending BTC periods. Choppy bear
• 70:00 - 70:30 slices lose singledigit percent controllable with a simple trend filter or volatility sizing trend filter or volatility sizing. Let's go ahead up here and uh let's go ahead and put these results up here.
• 70:30 - 71:00 results. Yep. Aggregated walk forward. Well, let's put them all in there. Why not recommend recommended next steps rerun walk forward with
• 71:00 - 71:30 compounding start capital equals equity last slice. Why you want this and what you'll learn is a true Kagger max draw down and calm add a trend filter should st sideep the 19% bare market
• 71:30 - 72:00 slices. Interesting. I actually want to do that right now because that's a solid idea. So, let's go up here before we get there. Let's say copy and paste. Copy and paste. Trend filter.
• 72:00 - 72:30 Please update this code to only trade if it is over the 200 SMA. This is adding a trend filter. Okay, first off, let's get back to our regular data here.
• 72:30 - 73:00 return is 192,000%. So, okay. And now let's go ahead and run it.
• 73:00 - 73:30 Yo, what's up,
• 73:30 - 74:00 Germaine? What's up, Justin 777 to you both? Much love. He say he's setting up his new PC. Old one burned up last week. Damn. Damn, dude. Almost back. Banana Dev, what's up, dude? 777, bro.
• 74:00 - 74:30 Bro. Okay, so let's go ahead and read through this. added a new 200 SMA indicator calculation only takes long positions and uptrend. This changes everything. We'll see how it does
• 74:30 - 75:00 though. Let's get rid of this. Well, let's keep this one. Hey, hey, hey. Except um it has the 200 now. Perfect.
• 75:00 - 75:30 Please do not have these debug prints outs because it makes it super hard to see the original stats.
• 75:30 - 76:00 Okay, so here we go. Profit factors two 58301. How did it do compared to the
• 76:00 - 76:30 OG? I mean, it got worse, but that that stuff was too good to be true anyway.
• 76:30 - 77:00 Shazam. Like the music. Shazam. I love that. I love
• 77:00 - 77:30 Shazam. Let's wait for this to be done. Optimize. I'm curious about that.
• 77:30 - 78:00 Don't even think about it, bro. Just
• 78:00 - 78:30 keep going. Let them find out who it was. They going to be on you.
• 78:30 - 79:00 Let them find out who you was. That's a good idea, dude.
• 79:00 - 79:30 Just waiting for this to optimize here.
• 79:30 - 80:00 Here it
• 80:00 - 80:30 goes. Yo, I appreciate you sharing
• 80:30 - 81:00 that. Okay, so you can see here. This is
• 81:00 - 81:30 crazy. These are crazy equity curves. Too good to be true again. Worst trade negative 22% 776 trades. Max draw down sharp 56.
• 81:30 - 82:00 The return percentage is immaculately crazy. So, I don't trust it. I don't trust it for a second.
• 82:00 - 82:30 Volatility is infinity.
• 82:30 - 83:00 Let's put the optimized results in here
• 83:00 - 83:30 for for funsies.
• 83:30 - 84:00 Nah, I'm not even going to put it in there. Let's go ahead and delete this file. It's too good. It's too good. I don't even want It's too good. I don't even want to, dude. Okay, so let's go ahead and do this. Let's do P1. Let's do priority one here and send me back the full code. Thank you.
• 84:00 - 84:30 Below is a self-contained walkboard
• 84:30 - 85:00 script with compounding. Each six-month trade slice now starts with whatever equity the previous finished one. No more artificial reset. So let's call this uh walk forward compounding copy paste
• 85:00 - 85:30 rename compound. I always do that, dude. I always put the copy instead of the paste. I hit the C again instead of that V. Should only be hitting the V, dude. Okay, here we go. This will take a
• 85:30 - 86:00 second.
• 86:00 - 86:30 All right, here we go. So, you can see
• 86:30 - 87:00 the max draw down is 52% here. Let's go over it with AI because, you know, she's fly and I'm just a I'm just me, dude. I'm just me. I'm just mobbing. above all the stats, please go ahead and
• 87:00 - 87:30 walk me through them and your analysis, please. So, let's go ahead here and say trading by walk forward here. Yo, Dan, what's up? 777, dude. Good to have you here, bro. compounding reveal and finder,
• 87:30 - 88:00 please. And let's put both of these in there. And then let's go ahead and open them up. Okay. So, this is not the one. This is the compounded work for Okay.
• 88:00 - 88:30 So, looks like heat, bro. It looks like heat. And then equity curve CSV here. Imagine you had everything you wanted. Everything. Imagine you were everything you ever wanted to be and capture that feeling. That's the easiest way to manifest what you want. I'm doing pretty good, man. Thank you. Thank you. Thank you. Good to see you.
• 88:30 - 89:00 All right. So, let's see what AI has to say about this. Number one, headline numbers. Compounded walk forward 100x 9.5 years. Total return is 101,000%. Simple ratio compounded slice to slice. The kagger 189% average geometric growth per
• 89:00 - 89:30 year. Max draw down is 52% worst peak to trough after compounding. The mean slice sharp here is the average of the six-month trade windows. The mean slice calmar here is the 259 slice level inflated when draw down is very small. Trade windows 13 nine winning four losing. What jumps out? The equity
• 89:30 - 90:00 curve now makes sense. No more artificial cliffs. Growth is lumpy but monotonic. 2019 mini run flat. 2020 Q1 2021 is vertical. Let's go look at that vertical thing, dude. This is what we're talking about here. This is the analysis. Walk forward equity
• 90:00 - 90:30 compounded. Please go ahead and explain how we did or why how and why we did the compounded verse the non-compounded walk forward and what the differences are. I kind of know him already. But um I just want her to explain it because you know I swear I'm not a know at all. I swear, dude. I don't know anything. I just want to learn it all. Outsized kagger is the real math is
• 90:30 - 91:00 real math but not real world scalable. when 1 million explodes to 16 million in a single slice and the new capital is redeployed one for one later trades assume tens of millions of BTC liquidity fine on Coinbase daily bars for analysis but in practice you'd cap size or switch the percentage of equity risk risk is high but proportional
• 91:00 - 91:30 negative 52% max draw down is ugly in traditional f finance yet onethird of The 190% kagger is 3.6 on the compounded equity compounded curve comparable to many quant equity funds. Yo DJ, it's been good man. How you been? Thank you.
• 91:30 - 92:00 regime dependence confirmed. The four losing slices match well known bearish or low BTC phases. Here's our trade slices. 2018 2019 negative 10%. Post 2017 bear still bleeding. 2020 to 2020 01 to 07 is a COVID crash
• 92:00 - 92:30 done sideways 01 2022 to 07 2022 is the Luna 3ac to Celsius meltdown. Wow. 2022 07 to 2023 01 is the FTX collapsy. Is that really true? Is that that then? Was that Ben? Yeah, I guess strategy is trend following in disguise and hates deep bare
• 92:30 - 93:00 chop. Parameter stability still rock solid. Every slice kept is the MA fast= 13 plus MA slowals 33 or 42 TP1 SL3. Okay. No sign of parameter drift equals low optimization bias. What this tells us edge survives multi-deade simulation of sharp over one draw down is tolerable relative to growth but a 50% hole in
• 93:00 - 93:30 psychological toe psychological to what's this word I've never heard it before tough tough is this how you t psychologically tough Dude, is that how you spell tough? Tough. I thought tough was spelled like this.
• 93:30 - 94:00 Tough and tough are homophones, meaning they sound the same but have different meanings. Dude, how do you Google something to see what the definition is and then it uses a word that I don't know either? What's a homophone? Well, it answered it in the sentence, but that's crazy stuff, dude. The acceleration of learning here. It's crazy. It's crazy. Are you coming back tomorrow? Because we'll be back tomorrow. I know it's Saturday, but please come back tomorrow because tomorrow is going to be
• 94:00 - 94:30 heat. Tough describes something as strong, resilient, or difficult. Tough, on the other hand, refers to a type of rock formed from compressed. What? I thought it meant hard, bro. Wait, wait, wait. Difficult. Yeah,
• 94:30 - 95:00 tough. Tough. Okay, I'm just I'm just slow. That's it. I just been spelling tough wrong this whole time. Scalability and liquidity becomes an issue once equity is over 30 to 50 million on a daily basis. You'd need position size caps or fractional allocation. Concrete next moves. Step a trend gate. Skip trades if close is under 200
• 95:00 - 95:30 SMA should erase the four losing slices. oneliner filter inside next or wrapper that sets position size to zero. Volatility risk targeting fixed per trade risk say 0.5% of equity caps runaway equity growth and limits draw downs position size liquidity cap permutation test
• 95:30 - 96:00 trade bootstrap fee slippage shock. Send me back the code for the 200 SMA filter on this exact same code. I want to see I want to compare them. Yeah, I want to compare the two and then we can do a comparison.
• 96:00 - 96:30 Copy this code over here and say
• 96:30 - 97:00 compound SMA 200.
• 97:00 - 97:30 Oh snap. But first I want to go ahead
• 97:30 - 98:00 and ask this or answer this question
• 98:00 - 98:30 here cuz I did ask something in between. Please go ahead and explain why we did why we did or why how talent you have naturally. Skill is only developed
• 98:30 - 99:00 by hours and hours and hours of beating on your craft. Oh god, dude. My prompts are so gross. Please go ahead and explain how we did or why how and why we did the compounded versus the non-compounded walk forward and what the differences are. Variance variant flat star walk forward non-compounded what we did began
• 99:00 - 99:30 every six-month trade with a fresh 1 million bankroll. Optimize the previous C three-year window. Trade the next six months, then reset capital to 1mm repeat. What questions the answers? If is the strategy profitable and stable independently of past gains, i.e. each slice is a self-contained live test.
• 99:30 - 100:00 Hidden caveats grows and draws down in percentage space. So overall return divided by draw down becomes meaningless. Equity chart has vertical cliffs. Looks like a negative 96% draw down. Compound walk forward. Same rolling windows and parameter search, but the ending equity of slice n becomes the starting equity of slice n plus one.
• 100:00 - 100:30 True compounding. If I actually reinvest my gains, how fast and how painfully does money grow. Real world P&L trajectory liquidity permitting. Hitting caveats. Later slices trade much larger nominal positions. So results assume you can execute tens of millions without slippage. Risk metrics now reflect real
• 100:30 - 101:00 draw downs. No slice independence. What each view told us. Flat start view. Slice level sharp equals about one across eight years. Core Edge survives regime changes. Four losing windows pop out instantly. Late 2018, 2021 Q1, 2022 bare and post FTX. Good for
• 101:00 - 101:30 diagnosing when the logic fails. Outliers don't distort risk numbers because every slice is capped at 1 M. Compounded view turns the same 1M to a B kagra of 190%. True max draw down is negative 52 not negative 96 calm equals 3.6 Six. Profitable relative to pain.
• 101:30 - 102:00 when to use
• 102:00 - 102:30 which use
• 102:30 - 103:00 case debugging parameter stability
• 103:00 - 103:30 detecting optimizer over fit flash
• 103:30 - 104:00 start stress testing testing strategy
• 104:00 - 104:30 Logic under different market regimes. Flat start sizing expectation compounded wealth path. Investor pitch deck compounded. Calculating realistic leverage. Use compounded. Key takeaways from running
• 104:30 - 105:00 both. The edge is real but trend dependent. Flat start slices isolate problem regimes. Compounded curve shows they merely pause long-term growth. Risk looks scarier in the real world. A
• 105:00 - 105:30 52% hole is survivable but requires iron stomach or a trend filter and position capsize. Liquidity matters beyond back test. Compounded equity assumes you can move 10 to 100 million notional on BTC.
• 105:30 - 106:00 Next step does discuss add a trend gate. Okay. Introduce riskbased position sizing. Okay. Permutation and bootstrap test. I asked for the full code back of the compounded with the the filter the simple moving average filter.
• 106:00 - 106:30 Transgate.
• 106:30 - 107:00 So, let's grab all of that code here and let's run this. So, this will take a minute and I'll answer any questions here in the chat if you have them. Um, how's your week been? Good, man. Thank you. I got access to the Bloomberg terminal at university. Is there anything you're curious about? What should I try first? I'll be able to
• 107:00 - 107:30 start using it at the end of May. But until then, I'd like to prepare. Sheesh, I don't know, bro. I never use that. I've never used a Bloomberg terminal. I've never Yeah, I've never seen it, honestly. Never seen it. Never used it. Never been curious about it. Yeah. What would you What do you think? What do you think? What are you going to try? Um, and by the way, how did the Google arbitrage work out? I didn't end up pursuing it actually. Where did you get this from?
• 107:30 - 108:00 Quote unquote Google Scholar. I don't know, bro. I just thought of it. What do you think about Monte Carlo market simulation? I believe if a strategy could work on simulated market, it would work on real. Um, I think Monte Carlo is a good test to run. Maybe we'll do that here. It's just about building
• 108:00 - 108:30 confidence. So, building confidence in your strat. Uh, okay. So, let's go ahead and give this data here. And then also there should be some more information. We got the mean sharp, the mean calm 97. Okay, that's solid. Kager 114 total return for 14 14,000% here data here equity curve
• 108:30 - 109:00 okay reveal and finder please your time is limited so don't waste it living someone else's life all right so I've attached all the data here. Please go ahead and walk me through it. Compare it to the original without the 200 SMA. Tell me which one's better. Um, and how you came to that
• 109:00 - 109:30 conclusion. Yo, I don't think Twitter chat getting forward here. G, just thought I'd let you know. Thank you, bro. Thank you, dude. I don't know why. Maybe it's just me though. Maybe maybe
• 109:30 - 110:00 baby. It's just using reream. So quick side by side of the two compounded returns. Okay, you can see here quick side by side. The final equity is a billion dollars here on the first side and 143 on the second side. So minus this 86. We saw that in our test too. We ran it. We ran this. But I'm glad that she's saying the same thing. I wonder what she has to
• 110:00 - 110:30 say about it though. Total return is 101,000% verse 14,000%.
• 110:30 - 111:00 Dana. So you can see the
• 111:00 - 111:30 Kagger 189% verse
• 111:30 - 112:00 114% so down max draw down 52% verse 71% so down it's actually not better to have this mean slice
• 112:00 - 112:30 sharp down losing trade window up verdict the simple 20 reduce return increase risk what actually changed and why observation missed early 2021 vertical ical run. Equity top now 75 or 760 ounce. BTC closed its 200 SMA until late 2020. The filter sideline the strategy during liftoff that produced the monster gains. More negative slices despite bare
• 112:30 - 113:00 filter. The filter kept you out of the rebound legs that usually re recoup draw downs. You sat in cash during recoveries, but were still hit by the intraday whips saws that triggered stop losses once once price poked above the 200 SMA.
• 113:00 - 113:30 Big lick here. Big lick coming here on BTC. Oh my god. A $ 122.6 million liquidation. We have to look at that. What is going on over here? A $ 122.6 million, dude.
• 113:30 - 114:00 Imagine that on a Friday. Imagine that, dude. Holy smokes, dude. He got smoked there. He got smoked. That boy got smoked. Let's put it there. 12 point something. 12.6 m. What was
• 114:00 - 114:30 it? I think it was 12.6M, but I can look it up. Yeah, 12.6. That's a big lick, dude. See how the price reacts to that. Does it fly through or does it bounce? I don't know. Not here to say all that. Just here to watch. He says test. Your
• 114:30 - 115:00 test went well. Anonymous. You and encrypted. You and encrypted both got weird ass names. Anonymous and encrypted. We got some dorky dorks in the chat. That's good to see. I love my dorky dorks. I'm a dorky dork, too. I'm a data dog. That's my cool way of saying I'm a dorky dork is I'm a data dog. Okay. I'm a data dog. You're a dorky dork. Anonymous and encrypted. And Mundev. Anonymous encrypted and Mundev. Okay. Let's get it. Dorky dorks, I love you. So glad
• 115:00 - 115:30 that you're here. Okay, so more negative slices despite the bare filter. Max draw down is worse. So equity curve now spikes later mid 2024 and then retraces with a smaller capital base. The percentage pullback looks deeper. Ah, he say been talking for a meme minute. Where you been talking
• 115:30 - 116:00 at? Um, okay. I see. I see what's going on. You got the same name. Same nameish kind of. I'm super curious to see how this reacts here to this 12.6 million liquidation. That is huge. That is absolutely massive. And we
• 116:00 - 116:30 didn't see one on anything else yet, which you know we good on Twitter now. Of course encrypted and anonymous are the same person, dude. Of course.
• 116:30 - 117:00 The filter did trim some downside noise. Which version is better? Pure performance and riskadjusted return
• 117:00 - 117:30 original higher sharp lower draw down smoother equity at any cost. It deepened the draw down. It's interesting to see compliance with position size liquidity limits both need a notional cap SMA filter doesn't solve that. So the raw MA reversal without the trend gate is superior on every headline metric. Why the SMA 20 idea failed here? Daily BTC already trends hard. When it
• 117:30 - 118:00 flips 200 SMA, the move is often halfway done. Strategy is mean reversion and breakout hybrid. It actually profits from the early transition zone around the long MA. The filter removes that sweet spot. Six-month trade windows are short. If the filter keeps you out for 3 to 4 months, you miss most of the slices opportunity. What to
• 118:00 - 118:30 test? Let's do that first. Actually, let's see if that makes it better. Let's make a shorter trend gate. I'm sure I can figure this out. Right.
• 118:30 - 119:00 change the trend filter from 200 to 100. Make a choice, right? You just decide what it's going to be, who you're going to be, how you're going to do it. Just decide. And then from that point, the universe is going to get out your way. It's like it's water. It wants to it wants to move and go around stuff, you know.
• 119:00 - 119:30 remember the imbalance from the start of the week from the Hyperlid 3D map? Um, yes, but I don't remember what the imbalance was. Was it showing more upside? I mean, probably. Is that what you're saying? I don't quite remember when that time was that you said that, but I do
• 119:30 - 120:00 remember you saying something about it. Okay, let's see how the 100 does.
• 120:00 - 120:30 All
• 120:30 - 121:00 right, here we go.
• 121:00 - 121:30 I forgot to show this. Dude, none of
• 121:30 - 122:00 this is financial advice. I'm not a
• 122:00 - 122:30 financial adviser. I need to show this every single time.
• 122:30 - 123:00 Let's go ahead and give this data here
• 123:00 - 123:30 um to the AI and
• 123:30 - 124:00 say I went ahead and tested it with a 100 instead of a 200. Here are all the details back um showing you how it did. Please go ahead and do an analysis now that we changed it to 100 instead of 200.
• 124:00 - 124:30 Quick scorecard. The three compounded
• 124:30 - 125:00 walk forward variants. No filter. Okay. No filter. The max draw. Okay. So the SMA filter makes the max draw down much
• 125:00 - 125:30 better. The KAG the KMAR is better as well. Well, that makes sense because the draw down's better. The equity is not higher. The return's not higher. risk profile max draw down collapsed
• 125:30 - 126:00 from 52% to 72 negative 72%. To negative 20%. A huge change. Return po uh fewer deep valley returns. return posture. Kager dropped versus no filter, but still outran SMA 200 and buy and hold BTC. Total return shrank 5x with no
• 126:00 - 126:30 filter because you sat out the craziest part of 2021 melt up, but it's still 214x on capital efficiency. Sharp jumped from 1.03 to 1.41. 4-1 because volatility fell faster than return. Only three losing slices
• 126:30 - 127:00 here. Why the SMA 100 beats the SMA 200. BTC often clears the 200 SMA late into a bull run. 100 day let you re-enter a few months
• 127:00 - 127:30 earlier and ride midtrend legs up. In bare markets, the 100 day still sits above price. So you avoid the worst bleed while jumping back in sooner for rebounds. Which version is best? Goal maximize dollar return can stomach 50% draw down then is no filter. I just need smoother smaller trades 200
• 127:30 - 128:00 SMA. Best risk adjusted growth here is the SMA 100 sharp 1.4 calm 6.4 for DD only 20%. That's nice, dude. That's real nice. Next tuning levers, position size, cap, ATR, risk
• 128:00 - 128:30 targeting, dynamic volatility filtering. Keep 1% equity at risk per trade or 10% or 10 million notional max. Okay. Should rain the 200 M to 1 billion notional assumptions of the raw back test dynamic volatility filter require a 20-day ATR over price permutation test. Okay, dude. Sick. I'll be right back. I got to take
• 128:30 - 129:00 a whizzle to remove all emotions and actually automate your trading. I can show you from a point of not knowing how to code at all. If you don't know how to code at all, I can show you exactly how to code. And then I can show you exactly how to automate your trading. You can see there's hundreds of reviews. I'll show you at the bottom. And there's a 100% money back guarantee. And the price is super inexpensive, 69 bucks. and I will literally hold your
• 129:00 - 129:30 hand step by step in the process of automating your trading. You can watch this video if you want to learn a little bit more about it. But we can quickly scroll through here and pretty much everything that I wish I could find on the internet when I was trying to automate my trading is in this boot camp. It's 15 days of videos that you can watch on demand in any language. Just in case you don't speak English natively, you can watch all videos at any time period of the day in any
• 129:30 - 130:00 language. And I show you step by step how to automate your trading. And not only that, I give you free AI. So I pay for your open AI claude. We have channels in our Discord that give you 100% free AI. So you can cancel all your subscriptions. You have a team of us that are all working towards the same goal. We're all chasing this guy. This is Jim Simons. He ran up a net worth of
• 130:00 - 130:30 $31.4 billion. And he always says, "You just have to make your systems better and better and better because that's what everybody else is trying to do." So, I can show you step by step how to get your ideas out of your brain into an automated trading system and you'll be able to code all of it. I spent 10 years in technology scared to code because I thought it was just for super smart people who went to Stanford or knew a
• 130:30 - 131:00 bunch of math and I was neither of those. Then I locked in because I had to automate my trading. I learned Python and I learned how to actually automate my trading. It's the most most secretive industry in the entire world. Nobody else is sharing this information. So, you can see a closer look of what's in the boot camp here. There's 15 full days. And on the first day, in case you don't know how to code, I show you how to code. So, don't
• 131:00 - 131:30 worry. I give you an intro to algo trading. I get all of your your orders set up, all your risk management set up. I get you to a place where you can back test any strategy that you have for the rest of your life. So, while I don't just say, "Hey, here's a bot that you can run make a million dollars, I think that's an a scam. If you ever see that on the internet, which I know you will, just run away because nobody is going to give you a trading bot that just makes
• 131:30 - 132:00 you a bunch of money. You have to come up with your own strategy and your own edge. But for a crazy low price, I can show you step by step how to get to that point where you can just sit here like I do. You see me on live every single day. I show everything on YouTube. And I just test test the RBI method. We must keep moving. We must keep going. If you can't fly,
• 132:00 - 132:30 run. If you can't run, walk. If you can't walk, crawl. But by all means, keep moving. And dude, I understand. I understand how emotional this game is. And sometimes you feel defeated because you're just tired of looking at these charts go up and down. You have FOMO, and then you have fear, and you're scared. you're in a position, you're in a long position, then all of a sudden you're in a short position, and all of a sudden you're using 40x leverage. Dude, I totally understand that. But that was Martin Luther King
• 132:30 - 133:00 Jr. And he says, "If you can't fly, then run. If you can't run, then walk. If you can't walk, then crawl. But by all means, keep going." And I know trading can rip your heart out. The sad thing about trading, which is like not good for entrepreneurs or people who like to work really really hard and just like just bang down doors and get anything done like you and me, dude. You know, we're ambitious. We're
• 133:00 - 133:30 data dogs. We're curious. We're fearless. You can't just work harder at trading and get better. You can't just put more hours in. Like everything else in the world. Most things you can just put more hours in, work harder, and be better at trading. With trading, you have to have patience, you cannot be greedy, and you can't be fearful. Those are three things that humans just can't do. So, in this boot camp, I show you step by step how to remove all of your
• 133:30 - 134:00 emotions, automate your trading. I give you free AI. I give you a bunch of examples of trading algorithms that will enable you to then build any strategy that you want. And then I connect you with hundreds of other algorithmic traders so you can make connections for life and work on this with other people opposed to just on your own. Remember, this is the prize. This is what I'm chasing. But I don't need $31 billion. I would rather
• 134:00 - 134:30 have a few hundred of us get $31 billion, you know? So, I'm not saying that this is this is guaranteed by any means, but for me, it's something that I can invest my time into and just like he says, just keep making my systems better and better and better. Now, I've been at this for 4 years. So, if you want to fastass and get everything I know about algorithmic trading and how to automate
• 134:30 - 135:00 your trading, you can see the prices right here. $69 per month. It's nothing crazy. And you can go to our Trust Pilot. Just Google Moon Dev reviews or Algo Trade Camp reviews. Here's some Q&A down here. I want to make sure you know that we have a no questions asked 90-day money back guarantee. So, for whatever reason you want your money back, just get your money back. You can keep all the code and just all the videos just you get your money back 100%. I follow the Costco guarantee here. This is the
• 135:00 - 135:30 90-day money back guarantee. Go ahead and check out the link in the description and just scroll down and see what people like you have to say. I know you're smart because I didn't run ads and I didn't beg you to come here. You found me somehow. You're a good researcher already. And the process of automating your trading starts with research. So obviously you're looking for a better way to automate your trading. So you're intelligent enough to know that everything can be done in this world as long as you're able to research. This is my RBI system. I'll
• 135:30 - 136:00 talk about in the algo trade camp, but it comes down to researching, back testing, and implementing. So, go do your research on the internet. Go look at all of these testimonials we have on our page. And if it seems like it's going to be a good fit, if you want to automate your trading, remove all emotions, connect with hundreds of other ALGO traders, connect with me, get free AI, have the 90-day money back guarantee, then click the link below and join right now. remove all emotions and
• 136:00 - 136:30 actually automate your trading. I can show you from a point of not knowing how to code at all. If you don't know how to code at all, I can show you exactly how to code and then I can show you exactly how to automate your trading. You can see there's hundreds of reviews. I'll show you at the bottom. And there's a 100% money back guarantee. And the price is super inexpensive, 69 bucks. and I will literally hold your hand step by step in the process of automating your trading.
• 136:30 - 137:00 You can watch this video if you want to learn a little bit more about it. So I think the next test I want to do is an uh size test. Send me back some code that test sizing. So we'll never we'll always use 25% size and make this a constant so we can change it.
• 137:00 - 137:30 So Below is a plug-andplay of the SMA 100
• 137:30 - 138:00 filter on it. Okay. Compound 25
• 138:00 - 138:30 perk. Copy the code over and let's run it. I get this bug. Please fix
• 138:30 - 139:00 it. Let's use whole units. But um please send me back the full code.
• 139:00 - 139:30 make a choice, right? You just decide what it's going to be, who you're going to be, how you're going to do it. Just decide. And then from that point, the universe is going to get out your way. It's like it's water. It wants to it wants to move and go around stuff, you
• 139:30 - 140:00 know.
• 140:00 - 140:30 Um, All right, let's check this out. I also want to test this on um a different asset. So, I want to grab Let's paste this and then let's run this. Damn. Focus, bro.
• 140:30 - 141:00 Focus. Okay, so you can see it's running now. Now, let me unfocus here and grab some new data here. I want to try it on um Ethereum because that's the next biggest symbol here. ETH here one day. I'll grab that after this.
• 141:00 - 141:30 Okay, I want to do that uh test
• 141:30 - 142:00 here. These are the results of the one that has 25% size. Please walk me through the comparison verse the others. Thanks.
• 142:00 - 142:30 The three SMA variants so
• 142:30 - 143:00 far risk 25% 214 million ending 100% is a billion max draw down 19% Ker okay big Solana liquidation
• 143:00 - 143:30 there. Let's go look at it. Damn, longs and shorts are getting liquidated as usual, dude. TR Trump is going crazy actually. Risk 25% rounded down to the integer. Always risk 25% fractional units
• 143:30 - 144:00 allowed. What the whole unit cap really did full 25% early on 610k when you deploy 1 to 4 BTC later buys only 500 fractional of the independent intended exposure. Oh snap I got to let this run. Sorry for the ADD, but sorry for the ADD, but that's how the show goes on, dude. That's how I'm
• 144:00 - 144:30 here all the time because my brain be going all over the place. Okay. Draw down cratered. Okay. Sharp per active slice shot up. Which sizing rule is better? Highest risk adjusted return while still BTC hold fractional 25%. Kyra is 128. Draw downs 20. realistic position sizes on spot or perpetuals. We got the data. I'm a data dog. But heck, you already knew
• 144:30 - 145:00 that. If you need ultra tiny low draw downs for a tiny discretionary sleeve. Maximum dollar return
• 145:00 - 145:30 here. Recommendations going forward. Use fractional sizing plus notional cap. or risk trade per trade ATR stop model. Run a fee slippage test. All
• 145:30 - 146:00 right, let's test this um Ethereum data here because I feel like we've done a lot of testing to see if this is robust or not. It looks pretty good. In the chat GBT, you use only one chat or jumping in multiple. I use one chat. I move when I need to when it gets too big or something.
• 146:00 - 146:30 the Coinbase data. We got the Ethereum one day now. Copy path and let's split it first. Let's split it. Let's split it to have in
• 146:30 - 147:00 sample out of sample. Split data here. Okay, we got it.
• 147:00 - 147:30 Let's grab that data here. Coinbase data here in sample. Copy path. Let's go to the original back test here.
• 147:30 - 148:00 28 reversal data path. Let's go ahead and
• 148:00 - 148:30 put in the end sample and let's run it, dude. You can see it's optimizing now. But I got the piano curve here.
• 148:30 - 149:00 So final is 8
• 149:00 - 149:30 million%. Expectancy 5.98 sharp ratios kind of low.
• 149:30 - 150:00 The buy and hold return is 13,000%. Anything we could think is real. And if it's not, we could create it. See, that's the gift we were given to evolve, advance, create. I create my own world.
• 150:00 - 150:30 I create my own world.
• 150:30 - 151:00 Oh, this is nuts. Nutsy, dude.
• 151:00 - 151:30 Nutsy bloodsy. Here's the optimized version, which is even nutser probably. Profit factor 4.5 returns
• 151:30 - 152:00 unbelievable. It's too good to be true. It probably is. It's too good to be true. It probably is. Okay, let's check out the outsamp or the Oh, we didn't do it on this first. Damn, I did on the full data first. My bad. Let's check it with just the
• 152:00 - 152:30 is copy path here. Really good as well. Expectancy 5.9. Buy and hold is that but this is the return for that.
• 152:30 - 153:00 So, let's try this but out of sample.
• 153:00 - 153:30 So, it looks good on both or the that one so far. And then let's check out OS. Still looks good on OS I
• 153:30 - 154:00 believe. Worst trade is negative 32%. Max draw down's pretty high. Return is 103% versus buy and hold of 7.3. So this looks pretty good, dude. Honestly looks better than
• 154:00 - 154:30 BTZ. So that makes me think what's it look like on Solana dog? I mean it's kind of biased. That's kind of selection bias, right? Coinbase data here. Let's say so. So, I'm grabbing the data for Solana for as much as I can get
• 154:30 - 155:00 really. There it goes. And let's do that same split here. Let's put it in the split data
• 155:00 - 155:30 here. Let's go ahead and run it. So now I have the split Solana data here in sample out of sample copy path.
• 155:30 - 156:00 That's what it looks like here. Max draw down 64%. That's pretty heavy.
• 156:00 - 156:30 That's pretty heavy B. See what the stats look like. Win rate 62%. Buy and hold return is 357 verse a
• 156:30 - 157:00 return percentage of 14437. Sharp is 877. Kalmar is 08 or 8.19. Best win rate, lose rate or worst trade. Best trade. Win rate is
• 157:00 - 157:30 62%. This is fascinating, dude. This is fascinating that I just show all this here on YouTube. Oh god. He doesn't care. He just doesn't care. He truly believes if I help more people, I'm going to be more successful. So, I'm just going to keep sharing. That's my belief and you can't take it away from me because you can't take you can't change somebody's belief. No way. This one looks like it's not profitable on the outer sample
• 157:30 - 158:00 though. See you there. And then let's go ahead and see here. So 75. So it did no it did not make money the buy and hold here. So you know this is been a very
• 158:00 - 158:30 interesting experiment. You've seen a lot of this is profitable. Um a couple of them were not profitable. We did all a lot of robustness testing. We did a compounded one, a compounded trend gate one. I didn't like this one. I'm going to delete it. I'm going to push all this to GitHub right
• 158:30 - 159:00 now. So, let's go ahead and say new terminal. Get add dot get commit. Um trend trend reversal back test looks good with robustness testing.
• 159:00 - 159:30 get push. Okay, so this is all on GitHub now. Um I'm going to show it show this actually in the inner circle here today. So that's in a few hours. No, sorry, in a in about 30 minutes. And um from there, I guess I'll
• 159:30 - 160:00 just see you when I see you. Come back tomorrow because you know I'll be locked in tomorrow. So make sure you're back here tomorrow. Same time, same place. Much love to you.