--- title: 'Fable 5 vs SQX: The Ultimate Battle Between AIs' source: 'https://youtube.com/watch?v=T6WLF1eZqpo' video_id: 'T6WLF1eZqpo' date: 2026-07-28 duration_sec: 933 --- # Fable 5 vs SQX: The Ultimate Battle Between AIs > Source: [Fable 5 vs SQX: The Ultimate Battle Between AIs](https://youtube.com/watch?v=T6WLF1eZqpo) ## Summary This video pits two artificial intelligences against each other in a trading strategy challenge: Claude's Fable 5.0 model and Strategy Quant (SQX), an older AI the creator has used for years. The goal is to determine which AI can generate the most profitable gold trading strategy, with the creator backtesting both on historical and out-of-sample data. ### Key Points - **Setup: Two AIs Face Off** [00:02] — The creator introduces Claude (Fable 5.0) and SQX, promising a comparison to see which creates the best gold trading strategy. - **Claude Receives Gold CSV Data** [02:21] — A CSV file with one-minute gold candlestick data from 2018-2024 is uploaded to Claude, with instructions to create the best one-hour trading strategy. - **Claude's Strategy: Simple Moving Average Crossover** [03:48] — Claude proposes a simple strategy: buy when the 50-period SMA crosses above the 100-period SMA, exit when it crosses below. Only long positions. - **In-Sample Performance of Claude's Strategy** [05:44] — Backtest results show net profit of $630, 236% return, 462 trades, 20% win ratio, profit factor 1.30. - **Creator's SQX Strategy Added for Comparison** [07:37] — The creator uploads his own SQX-based strategy and asks Claude to backtest both on the same data. His strategy outperforms Claude's in the in-sample period. - **Out-of-Sample Test on 2025-2026 Data** [11:02] — Both strategies are tested on new data from Jan 2025 to Jul 2026 without parameter changes. Creator's strategy leads 90% of the time. - **Final Comparison Metrics** [13:01] — Creator's strategy: profit factor 1.74, win ratio 52%. Claude's: profit factor 1.30, win ratio 59%. However, Claude's strategy is simpler and likely overfitted. - **Key Insight: Associative vs Brute Force AI** [13:44] — Claude's associative model risks over-optimization and may fail if widely used. SQX's brute-force search finds non-obvious combinations that may be more robust. ### Conclusion While both AIs generated profitable strategies in simulation, the creator's SQX-based strategy proved more consistent in out-of-sample testing. The video highlights the trade-off between easy-to-understand strategies from LLMs and more robust but complex strategies from specialized AI tools. ## Transcript two top one and top two artificial intelligences for trading. In one corner of the ring we have Claude with his Fable 5.0 model, which is free for those of us who pay a subscription until mid-July. And on the other side of the ring we have the artificial intelligence called Strategy Quan, which is the one I have been using for the last 7 years to create my own profitable strategies and are the ones that are generating financial markets today. Well, the time has come to confront these two artificial intelligences. More modern AI as we know it, such as LLMs, we know it, such as LLMs, and a slightly older AI that reaches . Which of the two will create the most profitable strategy for me? We'll see it in the next few minutes. But first of channel because you'll receive notifications when I upload this type of the description of this video, in the first comment, you have an invitation to the miniseries Trading with Artificial Intelligence, which I will be launching very soon, learn how to create your own profitable trading strategies with sign up for free in the first comment and in the link in my description. We're going to fight this battle within the Cloude ecosystem, downloaded my best trading strategy, which I've been using since the beginning of 2025. If you don't know, I 've uploaded many videos to my YouTube channel very profitable gold trading strategy that has allowed me to make a lot of money going to give to Cloud so he can backtest it and give me the results, but I'm going to compare it against the best trading strategy that Cloud can give me with his Fable model. Right now I have Opus 4.8 for certain tasks, but I have Fable 5 included until July 19th and it hasn't Therefore, what I'm going to do first of all is create a strategy that according to Fable is very profitable, but given a set of data, which is how I do it, I say to him, "Come on, create for me the best trading strategy from 5 years ago until today." No, no, no. In other words, I give it a certain number of years, and I reserve other years to see what I would have done once that strategy was created. This is a very basic validation that we do in our methodology. So what I'm going to do is add files, and here I have the gold data from 2012, from 2018 to 2024, which is how I think of it today, okay? In other words, now with this CSV you have the one-minute candlestick data , although I'm going to tell you to create it for one hour from 2018 to 2024, and I want you to create write to him next time. Instead of writing it, I'm going to send you an audio, and it's with this one here. Okay, I just attached some data in puncsv format which includes the data for the gold asset in CFDs. These are one-minute timeframes; they are one-minute candles that include opening, closing, high, and the best trading strategy that you consider to be the most profitable in this data set. It was a dataset from 2018 to the end of 2024. I don't want you to deviate from this data. I don't want you to go looking for gold data outside of this data range on the internet. I want you to know that with this data only, obviously to develop strategies you can go online and wherever you consider, but for data you can only get information from this CSV. Give me the best one-hour gold trading strategy you can find for this dataset. Now what I've said is really getting in the way. Here we have the prom, look, Fable 5, high reasoning, I'm going to hit send. Okay, the first thing this is going to do is break down this CSV file, it's going to overread it, and then it will take, I think, quite a few minutes to deal with this in the last few days and it's very fast, but anyway, I'll stop it and we'll come back when it gives me the results. I think it took, if I'm not exaggerating, about 20 surprised, to be honest. He had to vectorize it to speed things up; it's a technical matter. And he tells us, "Hey, done Ramon, work exclusively with CSV 2.48 meg, one-minute candles, such, H1, costs, blah blah blah blah blah blah blah H1, costs, blah blah blah blah blah blah blah blah, gold strategy from 2018 to 2024, final equity and trade list." Let's look at the strategy. He tells us, "Main strategy, moving average crossover, longs only." I didn't just say long ones, you know, if you don't know me, I'm all about gold, but long ones. And look, only long ones. SM rules are quick, damn, very very simple. Uh, the 50- period SMA in H1, which is the fast one and the 100-period slow one, a crossover. When the 50-period moving average crosses above the 100-period moving average, exit at the close of the candle in which the SM50 crosses below. It's a cross of stockings, a [ __ ] cross of stockings stockings, a [ __ ] cross of stockings , man. Fable 5 just gave me a strategy from a freaking moving average crossover, from a freaking moving average crossover. I go here and go to settings, I go to usage and it has and go to settings, I go to usage and it has used 28% of the Fable session. Everything at 28%. Well, that's something. Hey, if he comes up with a simple strategy, as silly as this one, is it because strategies that are truly silly and simple work better in trading? As I've always said in all my videos, we'll see, because mine is a little more complicated, but Hey, look, what interests me is the complete run from the 18th to the 24th, a net profit of $630. I don't know what risk he took. What interests us is Rodan's 236 return. I'll compare it with my strategy now. strategy now. 462 trades in a good while at 20%. Heads up, the win ratios are very low, okay? Va win ratios are very low. Profit Factor of 130. That's what I got out of it, huh? Wallet accessory. Nothing, nothing. Here he is, already coming up with new strategies, like that. What was tested and what was Okay, I'm going to tell you the main strategy, the one you told me, okay? The final equity gives us this, it gives us a combined portfolio, the red one, the black one doesn't interest me, eh, I want the red one, which is the main one, it's the 100-day MA50, only longs, the main one, the red one. This is the one that considered me the best, okay? Uh, number three, okay, what I'm going to do now is add my strategy. I have it here, it's this one right here, okay? And I'm going to tell him, wait, I'm going to close message. I just gave you a strategy that is mine and that I have been using for more than a year in my real trading accounts. I want you to take the main strategy. Forget about any other Forget about any other main one, the one that for you has been the best in that time period. Now take the strategy I've given you and do the same backtesting. Compare your main strategy with this one during the years I've given you, the years I've given you, which is primarily from 2018 to 2024. we'll look at other years and I'll give you that data. But for now, let's compare those two, your main one and mine, from 2018 to 2024. There's the text. I'm going to send it to him and see what he comes up with. After another half Cloud was a [ __ ] jealous one and we're going to see him on the track now to see who's winning. I've gotten hooked too, you know? Look, eh, I'm up here strategy, my strategy was much better, eh, let's see if I can show it up here better, wasn't it ? No, then I gave him my details. Well, in the end, I think my strategy was much better this way. Okay, my better. He tells me, "No, it's just that I gave him the list of three that I had he was doing, and now I've adjusted it." It was a matter of costs that they didn't have . I had picked up internet costs. Who knows. I told him the costs from the same broker where I did the backtest so we could compare them told him this. The swap, the fees, the spread, I should have told you all that in the initial promo so there would n't have been so much back and forth said to him, "Okay," I said, "With these costs that you now know, I'll let you change the strategy because I was playing with an advantage." Well, I'll give him that advantage, right? But I'm going to remove this from here so we can see it better. Don't deviate from the data set I gave you at the beginning and don't copy mine, that is, don't come looking for other damage because I'm saving this to make the real track later. Okay, so here blah blah blah I'll tell you my yes, everything is fine. Okay, give me the comparative summary of the equity Char, I I'm going to create a day trading strategy now high. I already know that the swap on gold is very high, but based on this, without me done a scalving strategy." And what has it gotten out of me ? ? First of all, look, the line, uh, the red line is [suppressed scream] their strategy, okay? The red line is their strategy and the blue line is mine. Notice that mine from 2020, 2024, or 23 is ranked, okay? It's a bit flat, it's ranged because gold does n't offer many buying opportunities either, and he's drawn me a line that looks like the Mask rocket going upwards, okay? And well, black is the much sense, but oh well. And maximum drug, huh, mine falls more, mine falls more than yours, okay? And a bit of a comparison. Here is the summary table. comparison. Here is the summary table. Cloud Asia Drift calls it n't know is that the 1 to 3 one has some incredible prech, okay? He does n't know this, but well, it's on top of backtesting and well, I have these numbers, okay? I have 953 trades and he does 1800 trades in a while and is better than me. Okay. Well, I honestly don't know if this strategy is viable in the real market, but oh well, it doesn't matter. Having said that, what I'm going to tell you is the following. Okay, I'm going to Okay, I'm going to give you the data for 2025 and 2026 now—that is , a year and a half into 2025 and half of 2026. I'll 2025 and half of 2026. I'll put it in here and say, "Great, now that you understand both strategies, I want you to test them without touching any of the parameters of either one: the one you created, the new one, and the one I gave you with the costs you already know, both the same, the same risk for both, using this dataset I'm giving you, which is from this dataset I'm giving you, which is from January 1, 2025, to July 2026, around the 7th or 8th. And compare the results only from that dataset. Don't want the results from this new dataset." There you go. I press enter. This shouldn't take too long because, well, just so you know, this is taking up my entire session today. I mean, today. I'm going to use it. I have 76% used. I hope this last test doesn't we'll have to move on to Opus 4.8. But anyway, what I was interested in is Then, when I do backtesting, I can do it with any other model. Let's wait and see what we get. Screw you, Fable, screw you. Look here who's winning from January 2025 to June 2026. Look, all the time, all the time, mine was winning , absolutely all the time. Then in the end, it's true that it recovered, but then they're right there. Technically, mine is ahead, but well, we'll call it a tie, but notice that 90% of the time mine is ahead. Here the drawdown recovers. Mine's a little higher. Mine's a little higher. Notice that here it's more mixed and notice that they come from an in-sample much, much higher. that I had this, that it's probably been over-optimized. Well, it doesn't have to be profitable here, but we should also discard the big sample. Years ahead, I we'll leave it here. However, I want you to understand that Fable 5 is all here, I have more data here, I have the MD which is [clears throat] the whole report. Uh, well, Ramón earns a little more, Profit Factors, above Cloude's strategy. I'm at 1.74 which is par, well, I have far fewer. Win ratio 52% of 59%. Maximum drop, I'm higher than him, okay? Return part Drown. It's true that he's much better. However, he's trading between 1 and 3. And since we don't have live data on the spray in CFD assets during those hours, it's true that it's not being reflected here. But even reflected here. But even so, both AIs are giving me strategies that are, well, in this case equally profitable, which is the simulation of real-world scenarios. What's the takeaway here? The takeaway here is that older artificial intelligence algorithms aren't biased by what the do the same thing with Fable, which is the give you the same strategy it gave me. Why? Because it's an associative model and it knows what works based on what the Therefore, if everyone watching this video does the same thing with Cloud, it's going to give them the same strategy, and it won't work anymore. What happens with tools like Strategic One that generates strategies by brute force and looks for combinations that, a priori, wouldn't make combinations that, a priori, wouldn't make associative or common sense, but it will test them and lead you to those branches of options and possibilities that you would never otherwise consider? The cloud is coming, and those options are likely to be the most durable over time. So you've seen both worlds: the most expensive artificial intelligence model in the world, costing the biggest tech companies trillions of dollars, is capable of creating strategies that, if you're lucky, could win you the same as tools you can learn for free today by commenting on the first post of learn for free today by commenting on the first post of this video. It's that miniseries I invited you to at the beginning, and I'm inviting you again now. We'll unprofitable strategies with AI tools designed for this purpose, tools we know the next video. Don't forget to subscribe because what we publish subscribe because what we publish here is pure gold. Ciao.