LLM Trading Strategies Update: Plus a Prop Firm Plot Twist (Part 5)
My Prop Firm Journey Takes a Turn
Important: Trading futures is extremely risky. You should only use risk capital to fund live futures accounts and if you do trade live, be prepared to lose your entire account. There are no guarantees that any performance you see here will continue in the future. I recommend using ATS strategies in simulated trading until you/we find the holy grail of trade strategy. This is strictly for learning purposes.
I’m not just developing strategies—I’m on a quest for the holy grail of automated trading. Questions? Check the FAQs first, then feel free to reach out directly: AutomatedTradingStrategies@protonmail.com.
This is a follow up to a post I wrote a few weeks ago titled The Prop Firm Paradox: Win Too Much, You Lose. It was a post warning you about the consistency rule that Topstep has for the funded trader evaluation and how to avoid getting an increase in your overall profit target—something I was unable to do. Thankfully, I was able to clear the new target set, but I learned a few things in the process I want to share with you.
First, as a quick recap, this is Part 5 in a series about using LLMs to create automated strategies to pass prop firm evaluations. I selected Topstep based on three primary features:
they have a good reputation for reliable payouts,
they welcome automated trading; and,
they have an ‘end of day’ rather than an intra-day trailing stop.
These were critical in the determination and Topstep won.
Fast forward, I’m two months into this thing and Topstep announces that it’s only funding traders that use TopstepX, its own platform. Here’s an excerpt from the site:
I won’t bore you with my initial reaction, my initial, my initial reaction, my initial reaction—as you can see, I’m still a bit stuck. Needless to say I was not happy, but it is what it is. So we’ll have to abandon Topstep ISO a new Prop firm.
Meanwhile, I’ll give you a quick update on where I am/was in the process. The first account was $50K; it passed with no issues. So I decided to try for three $150K account.
Account #1 made it past the $9K profit target (green dotted line below). Here’s an overview of daily performance:
Instead of a funded account, however, I was told that I needed to make another $2,205 to satisfy the ‘consistency’ rule. The rule states that your best trading day can’t be better than 50% of your total. Here’s a snapshot of my Dashboard and the rule.
Do you see the warning telling me to split the remaining amount into two or more days to avoid another increase? The note implies that I need to take another two days to comply with the rule, but this is false. Here’s the calculation for determining the minimum profit needed to meet the consistency rule:
Minimum extra profit needed = (2 × Best-Day Profit) – Current Total Profit
Minimum extra profit needed = 2 × 5,602.56 = 11,205 - $9,000 = 2,205
I wanted to test the rule so I made over $2,205 on the next trade. Here’s the trade report:
Thankfully, I was right. It did not take two days because the rule is connected to your highest day. As long as you don’t make a new high you should be okay.
I’m not sure what they’re trying to pull here. By all account this is a reputable firm, so I’m willing to say that the “split it up” note is a one‑size‑fits‑all caution to help you manage your evaluation. That said, it’s also a confusing and border line misleading message that could cause a trader to extend the time it takes to pass, which increases the cost of the evaluation.
The main takeaway here is that you only have to split it up if your next win becomes your biggest‑ever day. In my case, $2,518 < $5,602, so I was in compliance even though the note suggested otherwise.
Here’s the final Dashboard:
As you can see, I finished up $11,642.
So that’s one $50K account and one $150K account funded.
What about the other two $150K accounts?
I failed one, and I’m still working on the third one, which is up $7K ($2K more to go). I’ll keep you posted on the progress made.
New Prop Firm Hunt
So we’ve had a bit of a setback, but I think we’ll be better off for it in the end. I know a few of you have suggested Apex, even though rules state that automated trading is not allowed. I’ve also heard good things about Tradeify.
It’s going to take a minute to get my head around our options in this space. I’ve created an html based Prop Firm Cheat Sheet that I’ll be referencing/updating over the next month or so. You can download it now at the bottom of this post. I’d love to get your feedback.
In Part 6, I’ll provide an update on platform reviews, and I’ll tell you my favorite method for finding winners. It was found with a reinforcement learning algorithm trained on trade data from January to June of this year.
LLM Strategy Update
In case you’re new, this experiment started because I wanted to see if an LLM could create a strategy to pass a prop firm evaluation. A proprietary ("prop") trading firm trades its own capital, not client funds. Traditional firms hire salaried traders; modern "remote" firms allocate capital to independent traders and split the profits instead.
My approach: I interviewed several leading AI systems (Claude 3.7, Claude Opus 4, Manus, Grok3, OpenAI3, OpenAI4.5, and OpenAI 3 Pro) using the prompting techniques described in Part 1. Each model developed a strategy tailored to specific prop firm parameters. The code revealed risk management features I hadn't considered before. I am particularly interested in Strategy 101 (also available for download at the bottom of this page), a strategy created from the best features of top performing strategies. Unfortunately, it was just enabled this week, but it’s off to a great start.
Here's a quick overview of how each LLM-created system is performing so far. The first column is the strategy, followed by net profit, start date, number of trades, Max MAE, and Avg MAE. You will also notice a few ATS strategies on the table that fit the "evaluation strategy" profile.
Note: Claude Opus 4 and OpenAI 03 Pro LLM are the latest/most advanced models available. They were released after this experiment started, which is why they have a later start date. Strategies Claude Opus 4, OpenAI 03 Pro and 99 have not been published yet. Strategy 100 is available for download below along with the other strategies published from previous posts in this series.