I Tested The Strategy That Turned 5000 Into 15 Million

TITLE: I Tested the Strategy That Turned $5,000 Into $15 Million CHANNEL: Unbiased Trading DATE: ---TRANSCRIPT--- This is Ed Seykota. He turned a $5,000 client account into over $15 million and built one of the first computerized trading systems to ever exist. Now, that is a 300,000% return. And if you’re anything like me, I’m wondering how did he do it? Is it even possible? And does it still work? Well, I went looking for answers and I got to say, what I found really surprised in this video. I’m going to code his strategy from scratch using Cloud Code and backtest it across stocks,

[music] gold, oil, and in crypto and show you exactly what happened. Now, the rules are actually simpler than you think and if you in case you’ve never heard of Seykota before, he was primarily a trend follower. Now, unfortunately, Seykota didn’t write a single book. So, we’re having to reverse engineer his approach from his actual systems. Now, his systems actually boil down to five particular rules. They became the building blocks that every legendary trend follower such as Richard Donchian, the Turtles, that build on use since. Now, his first rule was a trend filter. This is primarily an EMA 100. Now, above that EMA, we’re going long or we’re having a long bias and below that EMA, we have a short bias. Now, the actual entry signal was a new 20-day high, meaning we had a rolling window of 20 days and if a new high was made, then we’re going long. And if there was a new low that was made, then we’re going short. Now, one of the most important factors was pyramiding. So, this is where he would add up to three units as price kept moving in his favor. And then for for his risk management, he used an ATR-based trailing stop that let winners run as long as possible and cut losers as fast as possible. Emphasizing, he used apparently a fixed percentage per trade and scaled by ATR, which meant the more volatility, the smaller the size for that particular trade. But, these rules only really matter if you can understand one thing about how trend following actually makes money. So, let me show you. Now, I’m going to give you a quick one-on-one on trend following. But first, most retail traders judge a strategy by its win rate and in trend following, it’s kind of the complete opposite. In fact, the best trend following systems in history have a win rate below 40%. While many YouTube videos like to claim, you know, 90% win rate, 70% win rate, trend following primarily has a low win rate, which means around 30 to 40% of trades are winners. Most trade entries don’t actually work, and that’s completely normal. But, this strategy can still make money because it has big winners. Winners are normally three to 10x the size of losers. One huge trade pays for a year of small losses most of the time in trend following. And this leads me to the next point that’s very important, which is cutting losers as fast as possible. When you’ve heard that phrase, they’re normally referring to trend following. Applying that to trading in general for every single strategy type doesn’t really make sense, but for trend following, it definitely is one of the key characteristics. In this case, we we want to be using a trailing stop to kill bad trades as fast as possible and never stay married to one particular position for a long time. And then lastly, the thing everyone forgets about trend following is that diversifying markets is one of the key unlocks for trend following. If you just have a trend following strategy on one particular market, you are highly under using trend following and ultimately destroying the strategy. And this is because you can’t predict which market’s going to trend. So, you have to be available to all those markets, so looking over them and looking for signals because then wherever the trend does exist, you want to be capturing it as much as possible. Now, how does this look in practice? Well, primarily, you could have six losses in a row, but then one huge win pays for all of those and more, which means you get a net profit on that particular kind of trade window. So, now that you know what the characteristics actually look like, let’s actually see what real results could happen from them. And to do this, I need to walk you through how I actually built this backtest using Claude code. Now, I actually recently made a video similar to this one around Larry Connor’s strategy. Feel free to check that out after this one, but in that previous video, I also went through that I gave Claude my existing backtesting code as style references. Now, for context, I’ve been coding since I was around 12, but I use Claude code and AI tools just like anyone else, mainly because they’re getting super good at writing code. But, the one key thing is they need context and they need references. So, ideally passing in some of your older code, but if you’re not technical, you could pass in template code or code you find online that you think is, you know, good and you can see the results from it, for example, or from a course that you like potentially. Once you’ve pasted in that reference, you can then just give it a very simplistic prompt and it can do a lot of the heavy lifting. For example, my prompt was very simple. It was just going over the trend following with the five rules as we listed above. Now, the core of what Claude actually wrote me was primarily around the indicators, which is quite simple lines. Primarily, it was adding the trend filter, the long breakout, the short breakout, and then we had the ATR calculation as well. Now, all these calculations are very simple mathematic calculations, so nothing really complex going on here. And then for the actual long entry, this is actually such clean code you can basically just read it. Uh primarily, we’re looking at the close value, so the close of a bar, if it’s above the EMA 100, so that’s our first check. Then, we’re also looking if the close is above a 20-day high, and this would be a long entry instead of a short entry. Now, the trickiest part, especially even for AI nowadays, was around pyramiding. So, adding to a winner. Most beginners get this wrong, and most of time AIs get this wrong as well. In the sense of beginner traders, most of the time you guys think about of averaging down on a loser. And don’t get me wrong, I did that in my first year of trading as well. But, actually what you want to be doing in trend following is once you identify a trend and it’s a strong trend, you want to be adding as much as possible to it because you want to extract as much value and returns from that one winner. And all the other losers you want to just cut as fast as possible. Now, to give you a basic idea of my kind of code logic, primarily, I looked if I was in a current position, meaning one would be in a current position, zero would be not a current position, and if I already had free units or free extra additions to that position already. If I did, then I don’t want to be adding any more because we only want to add up to free extra kind of pyramids onto a particular position. But, if I don’t, then I want to check if the last entry is above 0.5 times the ATR, which just means half the ATR value away from the last entry time. And if so, then we’re adding another unit because it’s a strong trend and it’s continuing in our direction. Now, an important disclaimer, this content is educational purposes only and it isn’t investment advice and I’m not advising you to run this strategy. Everything has to be based on your own research. Additionally, Public has compensated me to create this demo using their data, but all my opinions are expressed as my own. I personally only agreed to this because I could express if I liked it or if I didn’t. So, let’s get into how we used their data to actually backtest. And this is where it gets interesting. Trend following, as I’ve mentioned, needs diversification across asset classes as you don’t know which market will actually trend. So, normally this means on the development side of things, you need to juggle between three or four different APIs and sometimes even paying for multiple different API services to get all that data across all those markets. But Public’s API gave me stocks, ETFs, commodities, and crypto data all from one endpoint. And if you want to check it out for yourself, I’ll leave a link in the description, which is docs as well. Now, their data is primarily free as long as you have an account with Public. And to actually get access to the data, all you have to do is generate a secret key. A secret key is just access to your particular account using their API. You need to make sure to never share this with anyone because if you share it, they’ll be able to access your account using that API. And then two, that secret key can generate an access token. An access token is just kind of like a temporary key that expires after some amount of time. So, for example, it could be 30 minutes, a day, and this just means in the case of leakage of that key getting leaked somewhere or you you forget that you put it on a GitHub for example, it should expire within a certain amount of time so there’s not too much damage that could really happen. And then lastly, using that access token, we can pull bar data for any symbol as I mentioned above. And we primarily do this through getting the historical endpoint, but more on that in just a minute. Now, the six symbols I actually tested using Public’s data was spy, QQQ, gold, USO, BTC, and ETH. This normally would be a huge headache. I would have to have different crypto data, I’d have to have different commodities data, and equities data as well, but instead I can do this all just through Public’s API. And specifically, I used their new bar V2 with aggregation endpoint. This returns historical open, high, low, close, and volume bars, and lets me pick any timeframe from 1-minute to 1-month bars to use. I personally used around 5 years of data, and I used the daily chart for doing this backtest. So, now that we have the code done, and we’ve got Public’s data to actually backtest it over, it’s time to see if the strategy actually works historically. First, let’s just make sure our strategy is actually executing correctly. Let’s have a look at an example trade. This was from BTC in October 2020. As we can see from the chart, the first trade entered as the 20-day high was kind of broken or there’s a new 20-day high. Then we added as the range expanded, and we got more kind of like movement in our direction, and then we confirmed that by doing that again. Overall, we’d added then three units, and it hasn’t added any more, and then we’ve exited once that trend is no longer kind of present, and it’s starting to fade. Now, obviously in this case, it returned, so it’s a bit of an unlucky exit, but that’s going to happen with any trading system. Overall, this trade made $13,000, and it was held for about 45 days. But, that’s just one trade, so let’s actually see how it performed over many different years. As we can see, it actually performed really well. This is exactly what I would want to see from a trend-following equity curve. We see big spikes where obviously those big winners are really carrying us. Then we have these kind of drawdown periods where we have a lot of losers, but then we’re still getting big spikes once we’re getting those winners, and overall, we’re getting an uptrend to the right. Now, these were using all the same parameters that he previously had mentioned to use all those years back. So, there was no optimization, no parameter tuning, or anything like that. I simply just picked the assets, did the exact same rules as he stated, and ran the strategy. Overall, it did a 198% return. It had around a 16.7% CAGR, which is actually really solid. Only a 12% drawdown. This is including commissions and slippage. Now, to be fair to Public, and in case you’re running on Public, most of the time you won’t have commissions with any of these, but do look at the fee schedule documents, and you can see which things they charge commissions on and which they don’t. But, Public seems very fair on most things. You’re really not getting charged any commissions. And in some cases, you’re even getting rebates if you’re using like options, for example. And then in total, we had 334 trades. We had a 33% win rate, so a bit lower 40, but we’re still in that kind of 30 or 40 range. And we had a profit factor of 2.0. Now, most traders would probably call this a losing strategy in some cases, just because it’s got such a low win rate. And they’d often maybe quit after a few months or a couple of, you know, days or weeks of drawdown. But ultimately, this is the characteristics of trend following. And it’s why it’s worked for such a long period of time, because it’s so grueling to actually hold when it’s actually live. But it has historically made money for decades. Now, let’s actually see per simple. So, Spy was actually one of our kind of lowest returns. We only got 33% on Spy, and we had one of the lowest win rates as well on Spy. However, QQQ did amazing. Gold did really good. Obviously, that was more in the recent years. USO did okay. It was 48%, which is better than I guess Spy, but very low win rate as well. And then BTC and ETH, as expected, because they have more volatility, were able to add way more returns and obviously had very strong trends. So, let’s get to the verdict. Based on Sekoda’s rules, written before personal computers really even existed, it does seem to still work in 2026. Now, not on every market and not in every year. But across a diversified portfolio with strict risk management rules, they definitely have compounded from 100,000 to almost 300,000 over the past 6 years. Now, if you want to build a trading bot yourself, such as building backtests like this and then turning them into live algos, is exactly what we focus on in the Crypto Momentum Group, where we help you deploy and develop your first live algorithmic strategy. Link in the description if you want to check that out. We provide strategy templates, AI coding guides plus prompts, and live algo walkthroughs. Be sure to check out Public at public.com palm/ghost hawk to check out the trading API and all the services they offer.