I Tried to Double $100 With an AI Trading Bot in 22 hours .. Here's What Happened transcript

I gave my cord AI trading bot one single challenge and that was to double my trading account. To make things worse, I gave it only 22 hours and I deposited $100. Post like these on X claim that Claude is absolutely printing money and I wanted to just see if I could do it for myself. I mean, this guy here is only 17. What the have I been doing all this time? Surely I can figure it out, right? So, I set up Claude Opus and gave it a 22-hour challenge.

Find its strategy and really trade it on live markets with real money. And oh boy, did it deliver. Look at these trades and wins in a row. But getting there was a little bit of a different story. Today, I'm going to take you through the whole process step by step, what I built. I'll show you how I actually tested it and the real live results. And at the end, we'll be able to see whether it actually made any money and did it manage to double that account.

But for that to happen, we need to buckle up. I hope you're ready. Let's get started right now. The road map. Okay, first of all, we're going to be looking for a strategy. We're going to leave that to Claude AI. We're going to give it some instructions and explain the math behind it. It's going to go away and back test those strategies and it's going to be pretty intensive. Then we're going to deploy it on live markets to see whether or not it is as good as what it seems and hopefully double our account.

Now, I've been trading bots and back testing strategies here live on YouTube for over 5 years. It really has become my passion. And in my experience, simply asking AI to trade, I've got to say, is pretty where it really has helped me is finding strategies, testing ideas, building those bots, and eventually managing the bots once they're live on real markets. But this time, it's going to be different. I asked Claude and multiple other AIs to trade live on the markets, and the results have not been the best in the world.

So this time, I used Claude with one single question. Can it predict one candle whether it's going to be higher or lower on the 15minute time frame? One single candle at a time. Let me explain. [snorts] Okay, so this is my broker and this is what they offer. All we have to do is predict one single candle whether it's going to go up or down. And when I say up or down, I mean that this candle here would be lower than this candle here, or would it be actually higher?

Now, we can go away and we can choose the direction of each candle. And my broker isn't the only one that offers this. There are plenty of brokers out there that offer this exact service. It looks very much like binary options back in 2010 where we have to predict the market is going to go up or down within the next 15 minutes. But this is where it gets a little bit more complicated. As you can see here, for every single trade, the winning trade only wins 1.8x, which means that we're going to have to do a little bit of math.

So, what does 1.x actually mean? It means for every single $10 trade, we are paid out 1.x. Meaning, if we put $10 in, we would get $18 return, which is $8 in profit. If we lose, on the other hand, we will lose the full amount of $10 here. And although there isn't no liquidation, it means you've lost your full $10. So, if I took a 50% win rate strategy, five losses, five wins, with five wins, I would have made a massive $40 on my account.

But then I take the five losses in a row and we lose $50, meaning that our account balance is now minus $10. And to make things worse, that is before the trading fees are applied. Now, this means that we're going to have to work out the win rate. Now, normally I kind of ignore the win rate. We're going to need to work out exactly the win rate that we need for Claude to be able to stay profitable. So, if we take P as our assumed probability of winning, we can see here that we are actually going to need with my broker a win rate of at least 55.56%.

Which is without a doubt a high win rate. And once again, to make things worse, that is before they take away their fees. After the fees of my broker, as you can see, which is 0.06%, 06%. That decreases my wins from $8 to $7.73 and increases my losses to $1027. And now we can see that I'm going to need a win rate of 57.04% to cover for those fees. And here on this table, you can see the win rate that's needed to actually become profitable before and after fees, which shows you how a 0.2 26 fee can absolutely explode your account without you even realizing it.

So yeah, all of this on one single slide shows you the input of $10 a trade. Whether we win or lose, we're going to need at least a trading strategy from Claude that has a 57% win rate just choosing one single candle. I think this is very possible and I think that Claude can actually crush this one. So I put together a prompt that you will be able to find down in the description down below. I'm going to give it to Claude and I'm going to give it 2 hours to go away and try and find a strategy.

It's going to be back testing that strategy looking for setups for us to trade. And I'm just going to show you the tools very quickly that I use to back test these strategies. Okay, the first thing obviously we're going to need is Clawude AI. I've got it already downloaded onto my desktop. You can download Claude AI. I'm going to be using the latest version called Claude Opus 5.5. May have changed. Don't worry about it.

You can also use chat GBT to this. It's very, very simple. Next, we're going to be using a free MCP server called tradingkit.com. If you go in, you can sign into this tool just here. Click on my account up the top here and follow the installs just here. Now, this MCP server is free of charge. It has a huge, huge community. It gives Claude the tools that it needs to go away and back test strategies just like this one here.

And the very cool part about this is we're all building strategies together. It is an open community. There is 1,78,000 strategies that you can actually get the code from. And if you're interested, guys, you can also join the school down below in the description where I go through and share all of my best trading strategies. Obviously, you don't have to. you can just follow along. But if you just want to make things a little bit easier, you'll be able to find the link down in the description.

Okay, so let's go over to our account at the top here. We're just going to copy this code here. Whether you're using Claude or Open AI, it's the same. We're going to go over and we're going to start a brand new chat here on Claude. We're going to say, "Hey there, buddy. Could you please install this globally for me?" And I'm going to hit enter. Boom. And that's going to go away and install it for me. I'm not going to do it on my computer because I already have it.

It will then ask you to authenticate. Literally go back to your account up here. Copy this code just here into Claude like so. Hit enter once again and you're ready to go to back test those strategies. Hey there belly, could you do a very very simple back test on BTCUSDT using the MPCP server just to test that it's properly connected. And as you can see the connection is working perfectly. We have a net profit of 16% and 26 trades.

So that's neither here nor there. We just know that we actually have a connected back tester thanks to tradingkit.com. Unlike other videos, we're not actually going to be using Trading View today. This is just literally so that has a tool to go away and back test our strategies and it will then implement them in something that we will build a little bit later. Okay. Now, what I'm going to do is give Claude a prompt which I read earlier.

This prompt is explaining that it has 2 hours on the clock to go away and back test strategies. We're going to be using the odds of market, the 15minut up and down on Ethereum on these two markets just here after calculating the house edge. Now the house edge obviously as I said earlier is the fact that we are only getting 1.8x to each trade and taking away the fees before we actually calculate our profit in our pocket. goes through, explains everything, asks it to go away and use trader dev to back test those strategies.

Only strategies with a 57% win rate will be selected. Not only that, we're going to go away and build a mission control where you'll be able to see all of those strategies before we actually allow Claw to choose which ones to trade. Okay, so that's all ready. Off we go. And boom. That's going to go away for the next 2 hours. I will come back in 2 hours and we'll have a look to see what it is actually built. In 3 2 1 boom, time is up.

Here we go. We're going to go back over to my screen. Let's see whether it's actually made any profitable back tested strategies. As you can see, it has already popped up our mission control where it has back tested 1,667 strategies. In the past 2 hours, over 34,000 variants, different coin pairs, different time frames, only 1,161 actually go past and it went through a full validation period. And this is also included inside the prompt that we gave Claude.

Claude has gone away and optimized strategies, but has taken care to not show all of the data. This helps us not to overfitit these strategies. If we give all of the data to Claude fork all of the candles back in time, Claude will literally fit all of those settings to historical data and 99% of strategies when you put them into live trading will absolutely fail. So Claude has hidden around a third of the data and then comes away and back test strategies that is already tested on that data that it has never seen before.

So as you can see it has gone through multiple stages and validations before ending up only with 25 strategies with potential. Now currently for me I think that 25 strategies is great but I don't really want to be trading all of them. I'm going to ask Claude just to pick the top five so that we can actually go away and start trading these. Okay, that's amazing. Could you now go away and choose the top five strategies so that we can actually go away and trade them?

Right. Okay. Once that's done, what we're going to do is we're going to move over to the live trading part of this video. We're going to see whether or not these strategies are actually viable on live markets and see whether or not Claude can actually really make any money over a 22-hour period. Okay. So now we can see on our control panel we have a top five strategies that have passed the final test. There is 1 2 3 four five.

Just here we can see that the equity curves look absolutely amazing. And another part of the prompt that I didn't speak about earlier is that we are actually using position sizing. Now, this is probably the one of the most underrated things that creators on YouTube talk about, and that is when do we actually use a full amount of our position size or when do we actually use a little bit less? Most of the winning strategies used half Kelly on the bank roll.

And as you can see, the back test and the forward test look absolutely amazing. This $10,000 account traded all the way from 2024 would be now currently a staggering $341,000. The win rate has been checked on multiple time frames and has never gone below the 56% that we actually need to actually pass and actually be profitable. Same goes for all of the other strategies that Claude has chosen for us. And to me, the results look pretty pretty insane.

Okay, so now that we have these strategies, Claude knows exactly how to trade them. The next thing we're going to need to do is go away and tell Claude what to do. But there is one big problem. My broker that I'm using doesn't have an API for this. So, I'm going to have to literally tell Claude to trade it using compute, which obviously isn't the best way of doing this. Or alternatively, it can create its own scripts.

But that's pretty simple. We're going to go back over and we're going to say, "Hey, buddy. I'd love to actually trade this live. Please open my broker's website and work out a solution for me to actually trade this. I'd like it to run in the background on my computer without it interfering with my day-to-day work and it must be 100% automated. Going to hit enter. It's going to go away and open our browser. It's going to take about 15 minutes to come up with a solution.

You will have to probably answer a couple of questions about your broker, but Claude will always help you find a solution for these things. And boom. Right. Okay. So, after about 15 minutes of chatting with Claude about my actual solution, he has decided to build this Chrome extension. Now, if you want the Chrome extension, you'll be able to find links in my school down below. This Chrome extension allows Claude to take full control of my Bybit account.

Now, this may seem a little bit scary, but inside of this Chrome extension, there are multiple risk managements, including paper trading, alerts only, and locking into one single account. Now, I've put $100 over on this account. I'm going to set up my auto trade just here. Okay, so now it has access to that Chrome extension. We're just going to say, "Hey there, buddy. Could you connect to the Chrome extension that we built?

I have given you a 22-hour period to go through and actually trade live. Keep an eye on the results. Keep an eye on the actual trading. Use the five strategies that we actually decided to go live with. And your absolute mission is to try and double my account within 22 hours. Okay, we've hit enter. It's going to go away and start trading for us live using the wonderful Jarvis extension. It should start trading very shortly.

I'm going to put 22 hours on the clock and I will be back with the trading results within a couple of hours. I will tell you when we got our first trade and hopefully we will double this account. Okay guys, looks like we got our first entry. We have around 14 minutes left. So, we've already been in the trade for around 1 minute. We have invested $40. Potential return is $72. And our first trade direction is down. So, let's see what happens.

Currently, we are already down by $13. Hopefully, we push a little bit further down. So, we can bag our first win. Okay, there's less than five minutes. And I've got to say, guys, the first trade is not looking good at all. Um, we aren't far from the entry, but the market is currently pumping a little bit and yeah, isn't looking great for us. Very frustrating, but we're just going to trust the process and see what happens here.

Okay, the the rockets are just literally dumped on us. Um, we are at less than a dollar from our entry. I don't know what this is going to happen. This is really sitting I'm sitting on the edge of my seat here. I mean, it's it's only $40 entry, but yeah, less than 35 seconds to go. 76. No way, bro. It is pumping. Got 32. Oh, no. No way. No way. Did that not do that? No way. We just lost our first bet in the last second.

We just lost. Very unfortunate, man. And I can't believe that we didn't get that in the very last second. The market just pumped. Like I said, we're going to trust the process. We're going to continue. We're going on. We'll leave this to run for an extra 20 odd hours. And I'll be back with you guys afterwards. Okay, it looks like we were just had our last trade. I was just about to set the clock. 22 hours is almost up, but we now have an ETH on an upwards direction.

We're going to give it another 15 minutes on the clock to see whether or not we actually win. This is a trade for $107 and potential return is $192. As you can see, things have been going pretty well for the bot, but I'll go a little bit more deeper into that in a minute. We got 5 minutes left. I do not know what this one's going to do. is bouncing around on that candle. It's playing with my nerves. And yeah, this is kind of why I like bot trading [laughter] because I don't have to sit watching this all the time.

There we go. We're going back up. Fantastic. We're going to see what happens. Yeah, this is an amazing system, but very, very stressful. Come on. We're pumping. Little bit more. Come on. Three minutes to go. We are up $10 so far. We've got 1 minute and three seconds for this contract to close. Come on. 30 seconds. Don't go down now. 3 seconds. 2 seconds. Come on. Yes. $92 profit. It said to 85 over there. It miscalculated, but yes. $92 profit.

If I go over to settled, we can see $192 returned on inition. Initial investment was $107. Yes. Right. Okay. I'm going to calculate all of this and I'm going to give you a proper equity curve so I can tell you exactly how well this bot has actually done. But just for proof, let's have a look very quickly. As you can see, it's had a bit of a rough start, but we have had some absolutely amazing trades. Okay, here is our report.

Drum roll, please. Boom. Uh, okay. So, we've been trading for 24 hours. As you can see, just above, we have almost made $500 from a $100 account. You can see from the equity curve here. Claude has done such an amazing job of those wins. We can see our wins and losses. We took our first loss, which almost took us almost to zero. And then we had win after win after win, a loss, and another two wins. Absolutely insane. Insane that this has been able to do this and to be honest, it just feels absolutely too good to be true.

Now, I've been trading as an algo trader and an AI trader for the past 5 years. And if there's one thing that I have noticed is if it looks too good to be true, it most probably is. So, I'm going to make this into a full series. I'm going to come back on Thursday this week and give you a full update to whether or not this actually is still continuing to make money. And if you're interested in building your own version, you'll be able to find the prompts down the below in the description.

Like I said also earlier, if you don't want to build your own, you're more than welcome to come and join our community on school. There is no obligation whatsoever. You can go through this video and build your own. But if you just want a bit of a shortcut or helping hand, we are here down in the description down below. Guys, it has been a long, long time since I haven't been this excited about a project. Look at those returns.

I am so, so excited by this project. I can't wait to get back to you on Thursday to tell you whether or not this is really as good as it seems. If there's one thing that I've learned from algo trading, the law of large numbers is a rule that we should always put in place. So, I'm going to give it enough time to breathe and actually really trade to make sure that this wasn't just a lucky streak. Without further ado, guys, take care, keep safe, trade like a boss, and I'll see you in the next one.

Goodbye. >> Come on, David. And once again, [music] he's gone to a new strategy. He'll be back in just five minutes. RSI and MACD. If he doesn't [music] run into any bucks, there will be a new strategy.