How the BEST Trading Bots Work in 2026 (And Why You Should Use Them Now)

0:00 By the end of this video, whether you're a manual trader doing prop firms, trading crypto, Forex, or stocks, you're gonna realize why doing firm back tests is so important. I'm gonna show you exactly how to do them. And last night, for example, I ran my Claude Opus, I handed it my trading engine, and it built me these strategies. It built over 85 strategies overnight and back tested them, and it ranked them by the best. Now, how did it determine what was the best one? Well, it ran it through bootstrapping. So it ran through Monte Carlo simulations. This is the original candlesticks up here. You can see all the random variations below, and it ranked them to see if it passed or failed how many of these other simulations. Why is that important? Well, because we get really good results. Look at this backtest. 1.13

0:46 sharp ratio in out of sample data. This is, here we go. Let's look at the out of sample and 1.12. So the in sample and out of sample are both extremely good. 1.13 and 1.12. And then we have the, this is the full training data as well, which actually performed worse. So by generalizing our performance, we're actually able to decrease the drawdown and increase the Sharpe ratio across this strategy. So what is this strategy? I'm going to show you right now. But if you're not subscribed to the channel, feel free to do that really quickly. I like to share really high impact content with you guys. And if you feel like it, please like the video, hit the bell notification icon, and drop any questions you have down below. I do my best to answer all questions that you guys have. So what

1:31 is this strategy? This is a visual representation of it on the candlestick chart here. So we have basically, we do like a pivot high pivot low strategy. This is a variation on this where we also have an ATR as confirmation what you see. Now I have several other strategies to show you here. These are a couple other versions of this. And all of these have different performance. So I'm going to go through these in this video. But before I do, and I want to show you all the dashboards that come along with them, because there are varying results with each of these trade setups. And obviously, I have it automatically optimized through my back end. Now Claude's doing that automatically. If you haven't watched my previous video showing you guys how to set up that for yourself from scratch, every step in the tutorial, go back and watch my previous videos already. And then you can set something like this

2:18 up so you can have Claude run overnight. And people ask me, okay, well, does this use a ton of tokens? It actually doesn't because the engine's already built. Claude just runs it sits back and goes to sleep and then it checks in every 10 to 20 minutes so it's not using that many tokens it just has to look at the results and the script that it builds itself is doing the ranking for it so it does it has to do minimal labor it just has to you know manage the moving parts for that so we have our top 10 here from overnight so we have the pivot break uh and we have a stability score well that stability score is what i was maximizing here so it's maximizing the pass rate in the simulation. So it has a 98% pass rate, meaning when we go into these simulations, here, I'll show you what they look like here with these are the results. But just imagine here, original candlesticks, and then one of these, and then

3:05 another one and another one. So all of these are random simulation that you have to put your bot through if you're stress testing it. Why do you want to do that? Well, because look at this, you know, this is the median graph. This was just the, this is the middle of the, of, of, the rankings, an example of what the profit graph would look like. So what does this have? Well, this is just a random variation of candlesticks, and it still survived. You can see the graph is obviously still profitable, 500 profit, and the win rate's 40%. So did the win rate come down compared to to the training data, yeah, it came down a little bit, 41% compared to 43%. But do you see how that is, you know, minimal distortion? It's not as clean of an uptrend as this. But when

3:50 we come back to, you know, The out of sample data when we have real price structures, not fake ones that we've just randomized, still look really good. And a Sharpe ratio is beautiful. That's, you want a Sharpe ratio typically above one is my target. And yeah, Sharpe ratio is really good, but we also want to minimize drawdown in tandem with that. So we look at Sharpe ratio and drawdown together. So 10% drawdown, really good. And then as a byproduct of that, you know, your net profit is going to be higher. So risking only $28 to make $413.00. That's a really good risk to reward ratio, and that's why the Sharpe ratio is so incredibly high and devastating on this. So back to the visual representation, this is just, I'm plotting it. Now, I want to show you guys what most people do wrong. Most traders who come into this are looking at TradingView and AI

4:39 can hallucinate, but humans can hallucinate. So let's just take an example. This is just a simple RSI strategy that I have on screen. So let's say we want to buy, and here is our long as an example. We want to buy when the RSI crosses above this oversold area. So once it crosses up, we want to buy. When it crosses below it, we want to sell. And look at this, this is a really clean trade. Beautiful, right? Really beautiful. So 22% trade looks good, right? Well, trade tactics, if we aggregate this out, let's go forward. Let's try a couple more, right? Let's see how well we do. We long here and then we actually closed above. Look, we long as soon as we get a little cross above here and we take profit. Well, you know, there's some run up here or some drawdown in that trade. And yeah, 10% trade. So it looks solid, right? Well, you know,

5:29 we could go forward. We might find some losing ones. Here's another winner, right? So look, that's a 100% win rate so far. Long as soon as we cross above here. Take profit once we cross above the max RSI here. And we cross under. We take our profit. Look, that's amazing. And yeah, we could do that again. And here's where it doesn't look so good. We cross up here and we just hold all the way down and we're taking our first loss. Now, You can do that and kind of cherry pick the trades and a human will always have bias. They'll always want the strategy to win. But look at the strategy report. We plug this into a trading view strategy to expand this graph and it's minus 93%. The strategy just doesn't work, right? I can even like, let's do the entire history back test showing how it works on all of it. And it's a 60% win rate. What's

6:15 wrong? It's 60% win rate. Why are we down 93% on our account? So, That's what's important to do is to actually plug the logic of how you're entering trades into a real backtest. You know, whatever it is, if you're trading Fibonacci, if you're trading MACD or RSI or ATR, whatever you're doing, plug it into a backtest. See if it's doing well. 97% max drawdown. So if you actually did this strategy live on the real markets, you're blowing your account immediately, right? You might get lucky and be in the... like the 3% of trades where you're like actually able to like make money and not have drawdown. But even though if you have 60% win rate, the losses are staggering, right? And you might say, well, I'm a human and I'm going to be here and I'm going to decide not to get into that

7:02 trade. But you're just breaking your own trading rules then, right? Now, if you have a really fleshed out strategy, like the one we have here, and it tells you, it's telling you a signal to buy. Yeah, you could buy and you can manage it. And, you know, if you really start to see something a bot might miss, which the bot can see more information than you can, I can tell you that. But if you somehow did see something and you get out, maybe you reduce your run up, right? Run up is just how if the profit, if you hit 10% profit and it starts to come down and you take profit at 8% and you could have had 10%, but you lost 2%. That's 2% run up, right? So yeah, maybe you can like monitor the babysit the bot and or even like enter it manually yourself. and decide whether you want to take the signal or not. But that's what's so important, right? And by breaking up the price structures here, by randomizing them,

7:50 you can see there's totally different grids, totally different tops and bottoms and price and candlestick patterns and all of that. If our strategy holds up in that, right? Let's go to Let's go here to a different bootstrap sample from this strategy. This is the median. Look at that. Still 44% win rate. Sharps is holding strong 0.65, but still look how clean of an uptrend this is. It's really clean. And that's really why you want to do these tests, right? You want to withhold candles. You don't just want to take one slice of data here and do your backtest on maybe just one month of data. You want to maximize how much you're backtesting on to make sure that your strategy and your approach is robust because they aren't kidding.

8:35 95% of traders lose money. If you're doing prop firm trading, 95% of those people buy prop firm accounts lose money. So that's why this is crucial. This is super important. So moving on to the second strategy here, this in sample data, right? And I'm gonna just, I'm just gonna describe to you what these are. So the first strategy, this was the number one winner. This strategy, we have a stability report here. This is the dominant report. So these are essentially the same metrics as before, but my Claude was able to review this for me and then decide which ones are best, rank them, and then show me it. So I set this all up overnight. I told it to, you know, manage this, manage my trading system and just build different strategies overnight. That's all I did. I have an engine. I told Cloud to use it. You can even have Cloud research what it's building to test published, peer-reviewed,

9:23 trading, using as many published reports as it can to try their strategies to see if they actually work to put it to the test. And that's why it's crucial. So if you want to pause and read, this is just the summary. What was done overnight? The mission was run, and this is what I prompted, running just a one layer strategies so we're not adding a bunch of take profit levels we're not adding a i'm just calling that a backbone just the base of the strategy and the goal was just to find them that had the maximum performance in mounting monte carlo simulations and what we ended up getting was something that performs not only in monte carlo but it performs well in the real back tests so that's how we ended up running it so for example hard fails, right? In these simulations, zero candidates survived. We tried super trend, we tried a PSAR, we

10:12 tried a KUSAM strategy, we tried all kinds of things. And then we also had other strategies. So we know that all of these just didn't work. Chandelier type strategies that try to take profit. But we know mean reversion had really high drawdown, mean reversion. There's HMA strategies, did not work. We tried all of these strategies. So it's like you're working on, it's like you're on steroids, right? You're trading, you're backtesting, and you're able to quickly hone in on what's good. That's why we do this kind of workflow, right? And we're leveraging AI, but AI is not able to hallucinate because we have the real dashboards. We have... a real trading engine that has objective metrics that Claude just reads off. It runs the engine, reads it, and then it'll build something else if that doesn't work. So it's almost like machine learning, but leveraging AI

11:00 to do that automatically. So it saves you a lot of time. So we have multiple different of these pivot break strategies because it really liked the top ranked one, this pivot break. There was a pivot break one, which essentially it just looks for previous pivots. And then when we break above it, we enter and et cetera. And that it really likes these kinds of strategies that actually look at price structure, but it's a more generalized price structure so that it can, you know, even when you randomize the candlesticks or you have new data that's unseen, it still works in real time. So we tried different filters on them, right? We looked for like an RSI confirmed variant that performed better. And these were able to really outperform. They're extremely dominant. So, you know, you can see here that this is the top This is the top 10. There's a bunch of different types of this strategy that I

11:46 really like. There's also some runners up here. There's an MFI signal. That was actually pretty good. So obviously, that's an honorable mention. I'll show you that as well here. So we can look at all these dashboards, and I'll just kind of show you why this one was so dominant. And if we go to these other ones, we can go to the bootstrapping, and that's where I'm really interested. This is just the median bootstrap, and it's still profitable. It's not bad, right? It's not as dominant as our other ones. Sharp came down. you know, we really like to see what is the auto sample data. We withhold data, we only show our algorithm it after, and then, yeah, this one's not as good as the first one, which is why Claude ranked it lower. Auto sample has 25% drawdown. And it's not that nice, clear, crisp, you know, upsloping graph that we like to see. So when you're building a dashboard, yeah, you can literally screenshot this, post it into your Claude, ask it to build

12:33 your dashboard just like this. So it's really well organized. I really like, you know, having all of the dashboard here. There's the simulations. We have all this nice data that Claude can gather from simulations and then aggregate it. So it's really looking at the actual metrics. All of these metrics are one-to-one with TradingView because I just like it that way. It doesn't have to be programmed to TradingView, but if I'm converting the strategy back into PyneScript ever, then I'd just like to have it one-to-one. So I prefer that. So looking at this strategy, the This is the pivot break for strategy. This one, I think this was the, yeah, the RSI here. So here's the RSI and it's just like our trading view script. We have our RSI down here, but we tune it so that this RSI is robust for confirmation, but we wait for our secondary confirmation. And Shreddy really liked these kinds of strategies. This works

13:20 on Ethereum. It worked on Solana. These same strategies works on Bitcoin. So I'm assuming the next steps are, how does it work on NASDAQ? How does it work on trading gold or trading different things because they'll have different price structures. Some of them will be actually easier than crypto. So sometimes crypto is easier to trade. Like if you're on the long-term charts, zoomed out, like daily chart on Ethereum, it's really easy to trade. If you just buy, you're gonna make money in some of these trades. So it depends though, if you're coming down to like the five minute timeframe, there's so much noise. Look, this is just gibberish here, right? Like a bot looking at this, there's all these ladders of trades up and then down and really strong up and then chop. And then the head cut comes off here then it really starts to rebound way back up but look at all these liquidity sweeps that your bot

14:06 has it's like landmines for your bot but like i said in my last video if you haven't watched it a really good point that i was making in that video is well yeah the market's manipulated here look at this liquidity sweeping here well if your bot's training on data that has liquidity sweeps in it or market manipulation you've baked it in so your bot is trained to handle market manipulation that's what you want that's exactly what you want right so obvious uh This approach is dynamic and factors all of that in. And that's a very data-driven approach. Here's our third strategy here that didn't pass. as well. It got 40% drawdown in just the regular in-sample data. So, you know, out of sample, it might look okay. We'll have a look. Yeah, it's not good in out of sample. So look at the difference

14:53 here. I'll just flip back and forth. This is our strategy here. This is one of the versions that did not work out. And so it was a variant of the pivot break compared to the one that worked. So out of sample data here, really clean and up. And, you know, in this one, out of sample data, it did not like it. So if we go to the bootstrapping the median one. I'm guessing this one's not going to be good either, but it might be good. No, it's a lot of chop. A lot, a lot, a lot of chop. So there was a price structure here that it did not like in this version. I could obviously go to different bootstrapping charts here. I can pull up another one here. This was the best one. And it looks, you know, still, if this is as good as it gets, this is the best simulation that it had. But that doesn't look like a really good graph, right? As sharp as 0.18. So that's A red flag

15:42 when you have higher drawdown in your in-sample data, 42%. What's going to happen in your out-of-sample data? What's going to happen when you have different challenges? So looking at this one, the in-sample on our really good one was I think 20 or 30%, 24%. So it's better, but it really matters. The stability in complicated markets is really what matters the most. So that's the gist of the methodology here. I can show you a couple more charts. But yeah, it's just going to get more drawdown from here. So 60% drawdown. And if we go right to the other sample, it's not going to look good. Look at this chop, right? 75% drawdown, right? If you're really trading this, it actually ended up making money. But the chances that it was making money were low. I mean, the win rate came down. And if we look at this is one of the strategies here that was paired with it. And it

16:31 was using money flow index as a confirmation. And it just turned out not to be a robust technique to you. So that's how we know. We look at, if we have an RSI as a confirmation, okay, it works. Well, if we have MFI as a confirmation, no, there's no setup of this that works. This is not a robust technique, at least for ETH one hour, right? We can look at other charts, remains to be seen because what works on one coin or timeframe might not work on others. It really does depend. But, you know, we're going for the most broad generalized strategies, which tend to be the most safest. They won't make you the most amount of money up front, but, you know, over time, you can learn to rely on them. And it's a trust building thing with your bot where you're comfortable putting in more and more money as you trade, as you trade it more. So this is the whole thesis behind what I'm doing is very data driven. So here's another

17:18 one. This is the last one. This is the top five from overnight that I that it made. And this is a totally different strategy. And let's see how it performed in the out of out of sample data, which is the number one thing we want to look at. So sharp stayed the same, it was around 0.37 before 0.39. Now, and drawdown is around the same as well. But you know, that drawdown is a little bit high. So could we work on this set up potentially depending on how you know, this works in bootstrapping, if we take a look at the median bootstrap. So this is just the middle of the line bootstrapping, and it's having a very harder time, right? Sometimes it gives it really hard challenges, Sometimes, you know, it doesn't. So you want to have a high enough sampling of difference. And again, what I'm talking about here is these random, when I say bootstrapping, it's giving it random simulations like these ones here on the right hand chart. And you can see here, yeah, like the volume ratio. This is like, this is an example of what

18:06 this strategy was using was how high is a volume ratio for entries. And if it finds a peak, it'll enter. This is how the strategy worked. And it tested that idea out. And it, you know, it wasn't like the worst that we had here, but not the best of either. So from here, what would I do, right? Is this just a toy? Am I going to actually trade with this? Yes, I'm actually going to trade with some of these, like this one here, this winner looks very interesting. So, you know, from here, we could see if we take this version of the strategy with the winning configuration, if we could add to it, like if we add different take profits, are there any different types of take profits that work better than right now? Maybe we could produce our run up, have a better profit, lower drawdown as well, if we get out trades earlier, and if we We exit

18:51 them earlier at the right times. Can we have a higher win rate, higher Sharpe ratio? Yes. So we could build off this. And then we'll look at different coins. You know, we'll have Ethereum now. We might look at some other coins. I chose Ethereum just because it does have, it's harder to trade than Bitcoin. It has a lot of data on it, so it's good to train on. And then we could take this to see, okay, well, what about even meme coins? What about Dogecoin or any meme coin? What about any of the harder stocks to trade? And what about trading NASDAQ? So I might take this and see if it can work and hold up in those markets. And if it's reliable and if it's consistent across all these markets, then we know that this is broad robustness. So everything we do... is looking at the stability in simulations. That's essentially the point of this entire video is the

19:37 number one thing to take away from this. Look at stability in simulations because if it's not stable in simulations, when you run it through these random data frames, it's not gonna work. It won't hold up in real time. The chances that it works are low. It can work. You know, I'm not telling you that your strategy really sucks, but the chances that it is not robust are very high. And the chances that you're just either having a hallucination when you look at it to see if this trade worked, but what about the 10 other trades that you didn't look at? That's what you have to do with these back tests. And what about the versions of history that you didn't look at that you didn't even try? So there's a lot here to learn. And yeah, I just want to show you guys how I will trade this live once I have this up. is Signalswap.io and you know it's

20:23 free to use, free to sign up, connect your Bybit Apex Pro or Pynex account and then you can just link your signals directly from TradingView. You can connect it from if you're trading in Python you can even connect it as well. So you can connect it here and then if it performs well you know you have a dashboard it's easy to set up your bot in five minutes and I have other videos showing you guys how to do that but It's on the create bot page, so you can do that here. You'll find all your bots right here. And then from there, if it actually performs well, you can publish it to the marketplace. We have a whole community of people who are eager to publish to the marketplace. So this looks at real data and it does look at the stress test as well. is not fake data. It's not simulated data. It's using real trade data that the website gathered so that there's no foolery going on. So that's, and then

21:13 backtesting results as well will show up. So there's a backtesting area as well, which has very similar backtest to what I'm showing you guys today. So if you just drop PineScript in here, it will run a backtest just like this one here, and it'll show you the results. It'll do all the stress tests completely hands-off, automatic, spit it out to let you know, is it robust or not? And if it is, you You can attach that to your bot and if you ever publish it to the marketplace, it'll show up on the marketplace in the backtesting tab as well. So really cool to use. If you want that, it's optional, but when you create your bot, you verify that you have your backtest results. So you just attach the backtest that you ran to it, you create your bot, you create your signal, and then it's all linked up forever and it'll have that stapled onto it. So that's trust building as well. That was just a quick shout out to signalswap.io for you to

22:00 sign up. Go create an account now, create a bot, start testing as well. There is paper trading as well. So you can connect to Bybit Paper Trading without KYC. You do need a paid account on SignalSwap to do the paper trading, but you can set that up right now if you go and sign up. And yeah, there's no referral code because I'm a developer, so I helped build this website. But for backtesting, this will be up very shortly. It's not up right now. Just a quick note, but yeah, like next week should be up. Just working really hard on getting this feature up. But it is working in tests and it is really incredible what it's doing. And it's doing essentially what I'm showing you here today in the video here. It's doing these incredible back tests. So you can do that for yourself to figure out if your strategy is reliable or not. And yeah, I'm going to wrap up the video here. So just hit that subscribe button if you got a lot.

22:46 If you enjoyed the video, if you got a lot of education out of it, I'd really appreciate a like, a follow, comment down below if you have any questions. And I will see you next time. See you in the next video.