I Let Claude Code Build a FREE 5-Minute Scalping Bot… The Results Were Insane

I gave Claude some brand new model Opus 5A real trading account with one mission: to become a 5-minute scalper. I put 2 hours on the clock. The first hour to look for high probability trades, the second hour to actually take those trades. No help from me, and the thing that blew my socks off, it took multiple trades and won every single one of them.

trades. No help from me and the thing that blew my socks off, it took multiple trades and won every single one of them.

Now, I have been backtesting strategies here on YouTube for over 5 years. I've backtested over 2,000 strategies and am one of the top traders on multiple exchanges. So, I'm thinking it's like the same as you. No way a high win rate on such a low timeframe.

on such a low timeframe. I didn't expect that either and it's with my real money.

I didn't expect that either, and it's with my real money.

potentially the best part of this video. So, stay with me. I'll share with you every single trade, the truth behind the win rate because the truth matters way more than these numbers. Okay, so what are you actually going to get out of

Now, one of the trades did actually come close to a disaster before it was closed, and the fun part about the whole of this experiment was Claude was doing a second job in the background, which I think was potentially the best part of this video. So, stay with me. I'll share with you every single trade, the truth behind the win rate because the truth matters way more than these numbers.

mission to become a super scalper with a 2-hour window and in a secret mission, which I'll talk to you about in a minute. Once all of that's done, I'm going to go through the results with you, which you will also get access to

Okay, so what are you actually going to get out of this video today? I'm going to be sharing with you every single prompt. I'm going to be using free tools where possible, and where not, you most probably already have that subscription anyway. We're going to be giving Claude two missions, as I said earlier, one mission to become a super scalper with a 2-hour window and a secret mission, which I'll talk to you about in a minute. Once all of that's done, I'm going to go through the results with you, which you will also get access to as well. And then finally, we'll go through some real talk, like in every single one of my videos, going through what I think about the actual results.

And after 5 years of actually trading automated systems, do I think this one is any better? Guys, before we go straight over to my computer, if you are enjoying this video and you are looking for AI trading, if you don't mind giving

And after 5 years of actually trading automated systems, do I think this one is any better? Guys, before we go straight over to my computer, if you are enjoying this video and you are looking for AI trading, if you don't mind giving a good old thumbs up for this video, don't forget to subscribe. I do videos about AI and AI trading every single week.

today. We're going to be using its latest model Opus 5. So, if you haven't already got Claude desktop in downloaded, you have to go to claude.com/download and download Claude Code. Now, once you have Claude open, you will need to go to

Okay, so as you can see, I'm already on my computer. We're going to be using Claude Code for this example today. We're going to be using its latest model Opus 5. So, if you haven't already got Claude desktop downloaded, you have to go to claude.com/download and download Claude Code. Now, once you have Claude open, you will need to go to the code section over on the side just here. And we're going to be using what they call an MCP server. Now, an MCP server might sound a bit scary, but all it is is just giving tools to Claude so that it can actually trade.

that it can actually trade. Now, I'm going to be giving two tools to Claude, and they're pretty simple to install.

Now, I'm going to be giving two tools to Claude, and they're pretty simple to install.

over here, the community has already backtested over 352,000 backtests on multiple coin pairs, including Forex, gold, and also and also crypto. Now, this tool is pretty simple to install. We just click on my account, then copy this line of code, go over to

The first is Trader Kit. Trader Kit allows you to backtest any trading strategy on TradingView without having to use any backdoors or anything that TradingView doesn't like. And then gives you the code for the strategy. If you go over here, the community has already backtested over 352,000 backtests on multiple coin pairs, including Forex, gold, and also crypto.

also going to be using another tool, something new that I built this week. This is another MCP server that gives Claude access to TradingView's strategy library. There's hundreds, if not thousands, of free indicators on TradingView. And unfortunately, when you

Now, this tool is pretty simple to install. We just click on my account, then copy this line of code, go over to Claude, and paste into Claude Code. Hey there, buddy, would you install this globally for me? Boom! Just paste that in and hit enter. It will go away and install TradingKit for you so that you can start backtesting strategies. I'm also going to be using another tool, something new that I built this week. This is another MCP server that gives Claude access to TradingView's strategy library. There's hundreds, if not thousands, of free indicators on TradingView. And unfortunately, when you ask Claude to backtest or try a trading strategy, it generally goes for the terrible RSI or moving averages, all of these indicators that are losing money.

The wonderful community of TradingView developers have built thousands and thousands of indicators, so you can get out of the normal RSIs and actually find something a little bit better. This tool here is a pretty simple as well. We just

The wonderful community of TradingView developers have built thousands and thousands of indicators, so you can get out of the normal RSIs and actually find something a little bit better. This tool here is pretty simple as well. We just go here, we copy the whole of the readme, and ask Claude to install that as well. Okay, so now we have the two tools. TradingKit will also ask you if you want to live trade when you do the onboarding process, and ask you to connect over to your exchange. Now, personally, I'm going to be using my Bybit account, and I've set up a sub-account, as you can see here, with just under $2,500.

onboarding process, and ask you to connect over to your exchange. Now, personally, I'm going to be using my Bybit account, and I've set up a sub-account, as you can see here, with just under $2,500.

Now, the reason I'm not using my full account is because I like to diversify my portfolio, have multiple sub-accounts, and this is just an experiment. So, once you've done the full onboarding process, you will be also able to do exactly what I'm doing here today. Okay, so the first thing that I need to do to make sure that everything's working is make sure that we're actually connecting. So, hey there, buddy, could you go away and get my Bybit balance?

here today. Okay, so the first thing that I need to do to make sure that everything's working is make sure that we're actually connecting. So, hey there, buddy, could you go away and get my Bybit balance?

Okay, that took a couple of seconds, and as you can see, we have exactly the balances I said earlier. Everything is activated. We're on my Bybit account on a unified account. So, we're ready to go. I've already set up a prompt. Let's go through the prompt very, very quickly. Now, this prompt I actually have changed very slightly in comparison to my previous videos, after seeing how Claude reacted a bit connivingly last time, where it let the other LLMs just trade and sat on the sidelines just letting the other traders lose. I have actually changed the prompt. As you can see, I've set it up as a competition instead. You're an autonomous agent cryptocurrency trading competition. You're trading against your opponents. It has two phases. The first phase is to research and strategy development, allowing it to go through and scan the markets. Now, I think this is essential because instead of just going away and randomly trading, it's using its own logic to build ideas around current market conditions and is even allowed to use indicators to try and find better setups. As you can see, the prompt is very, very complete. And if I go down, you'll be able to see that phase two is actually where it is trading. We also asked it to create a log so that we can actually go through the trades and understand the real decisions behind those trades. So, I think the time has come. Let's let this thing trade for two hours. I'm going to put it in the background. Like I said, it's going to go away and create a strategy, and then we'll start trading in 1 hour. And here we go. Boom. Right, while that's gone away, I told you about something that it actually is doing as a side mission. Instead of just trading for two hours, I've actually asked it to go away and also create a strategy.

go through the prompt very, very quickly. Now, this prompt I actually have a changed very slightly in comparison to my previous videos, after seeing how Claude reacted a bit connivingly last time, where it let the other LLMs just trade and sat on the

Because it's all great having random trades that might actually work, but previous models have also been fantastic at creating and engineering new strategies, and this is where I believe the real edge is. Building multiple strategies, leveraging the time to allow AI to build and backtest strategies was something that used to cost a lot of money and now can be done in a couple of seconds. So, one of the most powerful things about Trade Kit is that you can actually backtest strategies. It has data for forex, crypto, and gold, and all of those are great things. And you can actually go through other people's strategies and see the code behind those strategies. We're working together as a community to find strategies together, and over 300,000 strategies have already been backtested. So, I'm going to be giving Claude a side mission. That side mission is to go away and find the best possible strategy on the 1-hour time frame. I've used BTC USDT on the 1-hour time frame as my goal today. I've asked it to use my border plate and to search through all of these indicators to try and find something that's actually profitable and not use the RSI and the EMAs and all of these absolutely outdated indicators that will actually lose m

phase is to research and strategy development, allowing it to go through and scan the markets. Now, I think this is essential because instead of just going away and randomly trading, it's using its own logic to build ideas around current market conditions and is

even allowed to use indicators to try and find better setups. As you can see, the prompt is a very very complete. And if I go down, you'll be able to see that phase two is actually where it is

trading. We also asked it to create a log so that we can actually go through the trades and understand the real decisions behind those trades. So, I think the time has come. Let's let this thing trade for two hours. I'm going to

put it in the background. Like I said, it's going to go away and create a strategy, and then we'll start trading in 1 hour. And here we go. Boom. Right, while that's gone away, I told you about something that it actually is doing as a

side mission. Instead of just trading for two hours, I've actually asked it to go away and also create a strategy.

Because it's all great having a random trades that might actually work, but previous models have also been fantastic at creating and engineering new strategies, and this is where I believe the real edge is. Building multiple strategies, leveraging the time to allow

AI to build and backtest strategies was something that used to cost a lot of money and now can be done in a couple of seconds. So, one of the most powerful things about Trade Kit is that you can

actually backtest strategies. It has a data for forex, crypto, and gold, and all of those are great things. And you can actually go through other people's strategies and see the code behind those strategies. We're working together as a

community to find strategies together, and over 300,000 strategies have already been back-tested. So, I'm going to be giving Claude a side mission. That side mission is to go away and find the best possible strategy on the 1-hour time

frame. I've used BTC USDT on the 1-hour time frame as my goal today. I've asked it to use my border plate and to search through all of these indicators to try and find something that's actually profitable and not use the RSI and the

EMAs and all of these absolutely outdated indicators that will actually lose money. So, I'm going to hit enter as well. I'm going to give it 2 hours to go away and find a strategy, and then I will come back with the back-testing

results. And only that, the live trades of the 5-minute scalper. Okay, guys. I just thought interrupt this video very quickly. We have taken our first trade after an hour of actually back-testing and looking for actual setups. We've taken a trade on

DEX USDT. We've taken $50 of collateral used and 9x leverage. And very, very surprisingly, the trade actually has been closed out very quickly. So, I'm not sure what happened there. I'm just going to leave it to trade, and I'll

come back when the full 2-hour period is actually over. In the meantime, we are actually building strategies out. We've already done 130 back-tests on using 16 agents all in the same time. Pretty impressive. As I said earlier, we're

using Opus 5 on extra. It allows us to open multiple agents. And as you can see, I'm not going to give away the actual back test results here, but they look pretty insane and I can't wait to share them with you. Okay, the 2-hour

period is over. We've managed to back test multiple strategies and also do of a 5-minute scalping strategy. I think the best idea is is first of all, let's have a look at the back tested trading strategy and then we'll go over to the

live trading results. As you can see, we've done over 130 back tests creating an amazing back testing results of 1,968%.

As always, we have asked a Claude to log all of the back test to our personal panel which is included in the actual prompt. You'll be able to see the back testing results on an equity curve with all of the results as you would expect

from TradingView and even better if we roll down a little bit further here, you'll be able to see the Pine Script just here. So, let's go over to TradingView. Let's have a look to make sure that this trading strategy actually

really is as good as it seems. Drumroll, please. And boom! As you can see on the back test results, we have exactly the same equity curve as we did with a Claude. The P&L is 1,968%.

The max drawdown is just under 30% which is a little bit high, but we could probably improve the position sizing to reduce the max drawdown. We've taken 111 closed trades and we have an amazing profit factor of just above two. The

equity curve looks brilliant, although a slight drawdown recently, it does seem to be recovering. Now, I did ask a Claude to go through the indicators that actually built this strategy because I believe that it is a pretty important to

understand what is actually going on. So, here is exactly what a Claude built. We are using the T3 which is one of my favorite indicators for trend direction.

When price is below the T 3, we're looking for shorts, and when above, we're looking for longs. We have multiple indicators to confirm our entries. The first being the squeeze momentum. Now, the squeeze momentum was a long time favorite of mine, and

surprisingly enough, when I did a test with the Kimmy not long ago, it also chose the squeeze momentum. To run it, we just go up to indicators up here and we type squeeze momentum, and there's this one here by Lazy Bear. Now, it may

be that it has had 114,000 downloads, and that may be the reason that the LLMs are actually using it because it is one of the most popular indicators. In second place on the actual leaderboard. So, if we go over

here, we can see we have waves above the zero and below the zero, and we also have these black crosses. To take entries, we're looking for the oscillator to go green, and we look for these crosses to be black. And when

we're looking for a short, we're looking for black crosses, and we're looking for a red oscillator. Now, this indicator looks very much like a MACD or something like that, but behind it, it is a pretty powerful. It uses multiple indicators

like Bollinger Bands to work out whether we're actually going to make a breakout. So, as I said earlier, it is one of my favorite indicators, and I'm not surprised that Claude actually chose it.

Now, Claude actually came up with a perfect recipe, believing that these two indicators made the perfect entry.

Claude then went away and worked out what would improve this strategy and make it a lot better, adding two new indicators which help us to stay out of the market. I did ask Claude to explain its choice, saying that the change of

volatility and the shaking volatility indicators were the real edge behind this strategy. If you go deep into it, it tells us that volatility is forecastable as direction certainly isn't, which I tend to agree with Claude. When you look back at

backtesting results, one big push in a direction often predicts another following big push. As it is a very very hard to work out working out whether or not it's going into a bullish or bearish momentum. It is a lot easier to work out

whether or not a market is actually going to make a move. As it says a bit further down, you can see that the T3 and the squeeze momentum are actually going to help us define direction. As the next two indicators, the change of

volatility and the shaking volatility indicators will help us to know when we should actually take our entries. So, by adding these two indicators, we can miss out all of the sideways moving markets and actually take entries when there's

actually going to be movement. So, we're looking for the shaking volatility indicator to be above zero for us to take an entry and for also the change of volatility to be above zero as well. So, these two indicators, it's

pretty simple. Just go over to TradingView, type in shaking volatility. It's this one here by David Tech. And guys, if you want access to all of these indicators, you'll be able to find a link where I'll be able to send you a

list of all of free indicators that I personally use plus the ones that we're talking about today. It's in the resource link which you'll be able to find in the description and in the pinned comment. Next, we're adding that

one here. And the final indicator is the change change of volatility by BJR117. It's had 202 downloads. Now, these two indicators actually have had a huge impact on the actual max so drawdown, decreasing the max drawdown from 37% to

the best settings that he actually found was around 21%. It also helped us increase our profit factor from 0.8 all the way up to two on best settings and cut down the trades from 208 to 106. So, these two

indicators alone have made a huge difference. Adding on top of that, we're going to be add the ADX, which also complements these two indicators. We need the ADX to be red above blue. Our final indicator that we're going to be

adding is the Stop ATR. Now, the Stop ATR is a little bit different from what I would use normally. This indicator is going to set the stop loss to the actual low of the indicator, and we're going to

be using it as a trailing stop loss just like this. Now, personally, for a long time, I've been against the trailing stop losses. Although they do often make more profitable trading strategies, the problem is is all of your rules actually

sit inside of TradingView instead of on your exchange, which just adds more risk to the situation. Unfortunately, I learned this the hard way. I had a couple of trades using trailing stop losses. There were some huge wicks on

the market, and unfortunately, the alerts from TradingView took too long to get me out and made my stop losses a lot bigger than they should have. So, generally, what I try to do is put my stop loss directly on the exchange using

market orders to get me out as fast as I possibly can. That makes the fees a little bit more, but as an algo trader, you set all of your rules in advance, you calculate how much fees are going to

cost you, and then you have to find the balance between risk and those fees and actually making a profit. Okay, so once again, that is the backtest of this strategy that Claude managed to find for me within a 2-hour period. The

backtesting results looks pretty amazing. I'm very, very impressed. So, now let's go back over to the scalping strategy and see how many trades it actually took and whether it was profitable. Okay, guys, so we're back here on the trading competition. The

current trading competition is over. It took a grand total of four trades in 1 hour and it won every single trade. Now, this is the first time that this has ever happened for me. Not only did it take four good trades, it they were all

profitable and it actually did execute all of its trades. You can see we took a trade a long trade on DEX, which I spoke about earlier. It was in the trade for around 8 minutes and made $1.01.

The next trade was a volume breakout. The trade was open for 6 minutes and made a grand total of $8.27.

Now, the next one was the trade that I was talking about that almost went disastrous. We took a trade on bank long USD. See, the regi using a regime expansion, the ATR tripled and volume was 7.16 x its normal volume. The volume died and

this single trade almost cost the full amount of profit. Luckily, it bounced back up and bank made a very small profit of 33 cents. Now, as you can see, Claude actually seems to have regretted not staying in the trade and it opened a

second trade. This one was open for 10 minutes after a pullback, which was the pullback that actually worried Claude in the first place and made us close out in 33 cents. Finally, it closed our fourth trade with $5.71

cents in profit. That is the highest win rate I've ever seen on any AI. As a matter of fact, most AIs that I have tested have never made a profit like that. Now, you do have to remember, guys, that this is just the AI's that

building their own tools. They don't have any access to any of my strategies or anything like that. They're just building strategies, trying to backtest on current conditions, and then take those trades. So, as you can see on my

screen right now, this is the proof of the live trades and all of the profit that they actually made. I'm pretty impressed, but let's not get overexcited. Let's have a bit of a real talk about these strategies. Now, I'd

love to know in the comment section, would you give AI your money and let it trade under these conditions?

Personally, I wouldn't. I like to see a full backtest and understand what's actually in that black box. The problem with AI is they've all been trained on things that they've learned from places like Reddit and books that they have

access to for their training data. So, if the stats are true of 95 to 99% of traders actually don't make any profit.

Unfortunately, all of these AIs and LLMs have actually been trained on that data, so there's not much chance that they'll actually be profitable either. On the other hand, we did have a good backtest result. This is where I feel as though

our leverage in AI is at the moment and how we can actually monetize and build profit from real algo trading. We can ask AI to backtest all day, every day, finding strategies using MCP servers to find real setups that would have cost

thousands and thousands of dollars only two years ago. Now, just for understanding, I wouldn't even live trade this trading strategy. Although the backtesting looks absolutely amazing, I would put this into a real forward testing period for 1 to 3 months

to make sure that the data and the live trading actually looks like the backtest. And then I'll put it onto a smaller account to make sure it actually works as I expected. Things like the trailing stop losses I've had to learn

as I've gone along and have cost me money. So I've just put rules in place that I don't continue to lose stupid money in the future to trailing stop losses and rules that just don't work.

So for testing purposes, I will actually be putting this into live trading. And if you guys want access to all of these strategies, I have backtested over 2,000 of them. There are 400 of them on the school. You'll be able to find the links

down in the description to that as well. Once again, guys, I hope you have a wonderful day. Take care. Keep safe.

Trade like a boss and I will see you in the next video. Goodbye. Come on, David. And once again, he's gone to code [music] a new strategy.

He'll be back in just 5 minutes. RSI and MACD. If it doesn't run into any bugs, there will [singing] be a new strategy.