Category: Trading AI
Tags: AIautomationGPT5strategiestrading
Entities: BitcoinGPT5Lux Algo AI Super Trend Clustering OscillatorOpenAIPinescriptsignum.moneyTrading View
00:00
That's right, guys. OpenAI just launched GPT5 and I've tested it for trading strategies.
I've given it a very hard time. So, if you watch this video, you will know exactly what is good for and what it's not good for.
Can it convert an indicator to a trading strategy efficiently? Can it improve a trading
00:16
strategy by itself? And also, can it create a completely new strategy from the scratch?
And if you watch the video until the end, you will know how to use this new AI model to your advantage. As you can see, I'm not in my usual studio environment because I'm traveling.
This
00:31
is my hotel room and this is a corner in that hotel room and I'm here to give you the latest information when it comes to AI trading strategies and automation. And with all of that, we already helped millions of people.
I mean, we have millions of views on this channel. We
00:46
have 210,000 subscribers, which I'm very grateful for, right? Thank you very much.
And if you appreciate that I'm bringing you this updates even while I'm traveling, then massage the like button, subscribe to this channel, click the bell as well to not miss out on this time-sensitive updates. And now, let's
01:02
take this AI for a spin. First of all, this is the GPT5 model, and I asked it, how are you better than the 03 model?
Because, you know, I like the 03 model. It's very powerful.
And funny enough, it seems to know who I am, right? Because the first thing it says, it's stronger
01:19
in reasoning and coding. And it even mentioned Pinescript and Trading View as if it knows the history that we had together.
It even says that it's good at tweaking Pinescript and also converting an indicator to a weekly strategy. Basically, it says that it's good at
01:35
Pine Script. So, let's check it out.
Right. The first challenge I want to give it is to convert this indicator to a trading strategy.
This indicator seems to be quite interesting. By the way, this is called the Lux Algo AI Super Trend Clustering Oscillator.
Wow. Let's
01:50
go. Right.
And this one is interesting because I found that whenever there is basically a blue uptrend here, then also the price goes up and then when there is a weakness in the chart, right? So you have this red cloud here in the indicator, it gets weaker.
You could get
02:06
out, right? To protect your profits and so on.
And the first thing I said is basically take this indicator, right? Because there is code for this.
And I can just click here on these code brackets and then it gives me the code. So I can take the whole code, I can go to the AI and I can tell the AI to
02:22
convert it into a trading strategy. Now I have two options, right?
I can leave it to the AI to decide when to buy and when to sell, which I did, but I can also tell the AI when I think it should buy and sell, which I also did. So let's see which one is better.
And if you are
02:37
new around here, I do have a video about how to convert a Trading View indicator into a strategy with 1.4 million views. Okay, so if you want to dig deeper, watch this video afterwards.
Okay, so I took the entire indicator code, right? Copy the whole thing and I gave it to the AI here in the prompt and the
02:54
prompts are all down below. You will find all of them.
But I also given it the Bitcoin chart data which I always export from here. Export chart data because I wanted to have the same data that I am looking at.
Long story short, I told it please convert this indicator
03:09
to a trading strategy. I should have put a space in between as you see fit.
Right? So I didn't tell it when to buy, when to sell.
find attach the data right from the Bitcoin USD one-day chart and I gave it some limitations the way I trade right long only 2018 to the future one
03:26
day chart 100% of equity per trade is better to compare like this 0.1% commission one tick slippage okay cool and then it went ahead and calculated the whole thing and it gave me this trading strategy okay I copied the whole code right and I've put it here right I
03:42
have a trading strategy for testing paste it there basically And I always put AI test in the beginning so that on the chart I know that this is the test version that I have. Save it, right?
And then let's see what it comes up with. Well, it's a pretty bad strategy.
Okay,
03:59
minus 32% on Bitcoin since 2018 is a pretty bad result. Okay, so the strategy is uh not great.
But the next thing I did is super interesting because I told it, okay, the results are horrible. I mean, of course, try a different buy,
04:15
sell logic. Buy when blue dot, sell when red dot.
I mean, I didn't really give it a lot of information, but when you go to the chart and you check it a bit more, then you can see that whenever there is a blue dot happening, right? Not just crossing above the line, but the dot is
04:31
happening, right? It's printed.
It's actually a good time to buy. And when there is a red dot, where is the next red dot?
There is a good time to sell, right? So, that was a good trade.
So I went to the eye and said, "Hey, please do it like that." Okay. And the AI gave me new code.
So let's copy this and
04:47
let's put it here. And again, AI test.
So I remember that this is the test. And let's see which result it gives now because now suddenly it gives,200% profit with 37% max draw down.
And it's not the best thing, but it's quite okay.
05:03
And it somehow added this zero line which yeah, which messes up the scaling. But let's see if it did the right thing.
Okay, let's see if it respected my request. So, here is the the dot, right?
It's printed and on the next candle open, it buys perfect, right? And then
05:19
here is the red dot, the first red dot since then and it sells. So, this is the logic I wanted.
And it's not too bad actually for the fact that it took me 5 minutes to come up with this, right? Converting an indicator to a trading strategy definitely works and it gives
05:35
you a good foundation to build on top. And if you're interested in improving a trading strategy and how to do it step by step, watch the video that I will put up here.
And for the nerdy people watching this video, it says here that the strategy might have a look ahead bias. Of course, you need to take this
05:50
seriously. Okay?
And you can ask the AI, hey, what does it mean? But as you can see here, the condition to enter is only when the bar is confirmed, right?
So there is no look ahead bias in this particular case. The next intellectual challenge for GPT5 is to take an
06:07
existing trading strategy that you like, for example, and improve it. Okay, make it better.
Of course, you have to tell him what better means because it's an AI, right? You have to give it good instructions.
And the way I did it is that I took the Gshian channel strategy code, which I have, which you don't
06:22
have, but it doesn't matter. You can take any strategy that you have access to, and I wanted it to make it better.
But make it better in this case means that it should avoid trading in a sideways market, right? Because you can see here it loses money a lot of times
06:39
because this is a a long sideways chop. And I wanted to take only the nice trades, right?
Where the trend is clearly up, right? Not sideways.
So I told the AI to do exactly that. I went to GPT5 and I gave it a couple of
06:54
informations, right? not only the code but also the screenshots of my inputs and the properties that I'm using to be sure that it uses the same settings that I use just improve the strategy right to make it better again I gave it all the data of the Bitcoin chart which again
07:10
you can download by coming here and clicking on export chart data right and then I told it to improve the strategy because it trades in sideways markets and loses money right so the idea is to avoid such trades and to make it better and better means 4% profit and lower
07:26
draw down. I told it to consider all the attached images and data.
I gave it the code of the existing strategy and then it went on to create a better version. The problem was that it created some errors, right?
So I had to tell it, hey listen, I have this error. I just give
07:42
it a screenshot, right? And then I say, hey, fix the error and give me the full fixed code.
Then it gave me fixed code. But then there was another error because it always uses adx and this function does not exist in pine script.
It's very annoying because it does it all the
07:58
time. Okay.
Anyway, gave it that screenshot, fixed the code. Long story short, I then went to Pinescript.
I pasted in the new code, right, that I obviously got from the AI and this is the new version it gave me, right? So, let's forget the total P&L for now.
The
08:15
draw down is lower than the previous version. and let me switch between.
Right? So you see the previous version had 30% draw down.
Total trades is less. Okay, cool.
Right? Because it trades less in sideways markets.
And also the profitable trades went up by well 3% points. Not huge.
And the profit factor
08:32
went up a little bit. But in total this new strategy has less profit.
But still at least the AI managed to improve the draw down because that's what I told it right that a better strategy means either or both, right? either I have more profit or I have lower draw down.
08:49
But what is a bit disappointing is that when you look at this area here, this is the sideways market that I don't like. Right?
And I gave it the instruction to eliminate these trades, right? And if you compare with the old version, all those trades are still there.
But what it eliminated are actually the
09:06
profitable trades. You see this one here disappears in the new version, right?
And also this one here disappears in the new version. This is very disappointing.
And I think that this model is not any better than 03 before for this type of use case. And I really expected the AI
09:22
to do a better job. Of course, I can now continue talking to it, give it more screenshots, tell it, hey, this trades here is the are the ones I don't like.
But it is disappointing. Let's go to the next challenge.
Maybe it's better there. Is GPT5 better creating completely new
09:37
trading strategies? And the idea here is, let me tell you the secret.
The idea here is that I'm hoping that in the training set of the AI, there are very good trading strategies hidden, right? I want to get it out of it.
Okay? So, I give it the most freedom possible to
09:52
create a good trading strategy. So, here's the thing, right?
Again, I go to the I give it the chart data. I think that's important because it has to have the same data as I have in Trading View, right?
And then I said, "Hey, I need the best trading strategy you can come up with for Bitcoin USD one day chart data
10:09
attached long only 2018 to 2069 because it's funny, right? Sometime in the future." Yeah.
One day chart, 100% of equity per trade, 0.1% commission, one tick slippage, and the goal is to have a super high net profit and a max draw
10:24
down that is below 30%. Let's go.
It did come up with something, right? And the thing is that the first version that it gave me, I dropped it in.
Yes. It had of course coding errors.
So I gave it again the ADX problem. Yeah.
I gave it a screenshot. I told it to fix it.
And
10:41
then it didn't give me the full code, right? And I said, "Hey, give me the full code or the entire code." Cool.
It gave me the entire code. Then there was another issue with this function.
Again, just give it a screenshot. You don't need to understand what the issue is.
Just make sure it can see the code around it and the error itself. Right?
10:58
And so, okay, give me the full fixed code, right? It gave me another code.
There was another coding error, okay? And so on and so on.
In the end though, it gave me an interesting trading strategy that I want to show you right now. This is the trading strategy it gave me.
And as you can see, what I
11:15
really like about it, the max equity drawdown is just 15%, 15.46. That is amazing.
Profitable trade 61%. Oh my god, that's so good, right?
profit factor 7.1 and now you would say yeah but the net profit is not that high
11:32
compared to gshian channel that is correct but gshian channel and here's the trick how to think about this has a draw down of 30%. Right?
So now check this out. If you basically take this strategy from the AI and you 2x it,
11:48
right, you leverage it 2x, then you have 30% draw down with 2,400% profit, but 60% profitable trades. Way better.
Okay? So, you see what I mean, right?
So, you have a higher chance of success even though the net profit is lower. But when
12:05
you have a super super low max draw down, you can leverage the strategy 2x 3x. Okay?
Don't go above that. Right?
and then you basically take advantage of the entire thing. Now, what you can also do, of course, is you can continue the conversation with the AI to improve this
12:21
strategy in front of you even further. And that's exactly what I recommend doing because you will never get the perfect strategy from the AI and that's it.
You will have to improve it. And I have a video for exactly that, a step-by-step guide on how to improve a trading strategy, which steps to take,
12:37
right? How to use the AI to help you do that.
Right? This is the video up there.
I think I linked it already. You will find it.
It's a very important video because that's the only way, at least right now, to get to a way better trading strategy than the AI can give you from the scratch. Overall, I think
12:54
GPT5 is not much better than the 03 model. I think what happens in the background is really that the AI is choosing from the existing models.
You remember the models from before, right? 403, the mini versions and all these things.
So GPT5 seems to be choosing
13:10
from the existing models depending on the task at hand and it basically just uses all three anyway, right? Because that was the best one for coding or at least for our use case.
So I don't think that this model has more knowledge or is smarter than before. It's just a consolidation of all the models that
13:27
were there before to simplify it for everyone. If you want to get started with trading strategies that already prove themselves, go to signum.money.
We have the gshian channel strategy for the one day chart, the Ichimoku strategy for the 4hour chart, and of course when you click here on get access, you can get
13:43
access, but you can also see the videos representing those strategies and how to automate the strategies so they buy and sell on your exchange fully automatically. If your crypto portfolio is $25,000 or more, then you need more than just trading strategies.
Consider the autotrading masterass where I teach
14:00
you my entire process including my trading strategies of course which are the best in the industry. Click on these screenshots and you will see every single trades they took.
But you need the education to properly build a riskadjusted portfolio to know how to play the markets depending on the cycles
14:16
and how to rebalance how to use all of this to your advantage. Right?
Do your homework by the way. Check the reviews.
Watch the video here on the website and you will understand that this is a limited access program for thousand people only. We still have some seats because I just started the program one
14:31
year ago. This is the most advanced program that you can find and this is just for higher net worth individuals for people who have enough money to understand that risk management is the most important thing.
Okay. So, if this is for you, apply for access.
Click this
14:46
button. We're going to talk to each other.
If it's a good match, then you can join. As you have seen, GPT5 is pretty good at creating completely new trading strategies, but you still need to learn how to improve them.
And that is why this is the next video you should watch.
15:03
[Music] to do. Yeah.
[Music]