ICT Strategy to Algo: How to Prompt Claude (It’s Not Perfect) | Pt. 3

Summary

Presenter Ruben reviews the performance of his prop-firm trading accounts and demonstrates how to use Claude to turn an ICT-style fair value gap (FVG) idea into a testable Pine Script strategy. His central message is that AI-generated code is a starting point—not a finished or necessarily profitable system—and must be checked against charts and data.

Trading results and account management

Ruben says he had two funded accounts, received payouts from both, and later lost both after a series of losing trades. On a $25,000 account, he made about $1,000 and withdrew roughly $500.

He says his system’s rule is to withdraw when a payout opportunity arises, and that he personally prefers withdrawing profits immediately. He does not give a universal recommendation; he says the decision depends on each person’s circumstances and is personal.

He compares two systems:

• A 15-minute FVG system with a volatile, uneven equity pattern.
• Flipper, which he describes as smoother and says recently passed funded-account verification. Its chart included a substantial neutral period.

Ruben emphasizes that both winning and losing periods are normal, and that prop-firm rules differ from those for personal accounts. His own spending guideline is about $350 over two weeks, allowing him to take on roughly two accounts per week while limiting costs and pursuing potential returns. He presents this as a personal guideline, not a general recommendation.

Prompting and testing framework

Ruben’s process for developing an algorithmic strategy is to:

1. Specify the setup precisely. Define the FVG type and timeframe, what qualifies as a valid break or retest, and the entry trigger.
2. Define risk controls. Specify the stop method (for example, below the FVG, at a swing low, or at a fixed distance), take-profit method, risk/reward ratio, and position risk.
3. Set operating constraints. Specify trading hours, maximum risk per trade, and maximum trades per day.
4. Ask Claude to explain or illustrate concepts—for example, an ICT-style bullish or bearish FVG—if it is unclear whether the model understands the rules.
5. Generate Pine Script, load it into TradingView, and inspect the chart. Check whether entries, stops, and exits appear as intended.
6. Iterate. Ask Claude to correct missing or inaccurate chart behavior, then retest.

Ruben cautions that generated code may not match the requested illustration and should not be trusted without verification.

He contrasts vague instructions, such as “buy pullbacks to the fair value gap with a short stop,” with detailed specifications. Claude may fill in missing rules using its own assumptions—for example, by adding an EMA filter—which may not suit the trader or produce a profitable strategy.

Key numbers and examples

• Account results: A $25,000 account generated about $1,000 in profit, of which roughly $500 was withdrawn. Ruben says he later lost both funded accounts.
• Risk/reward ratios: Systems are described as having 1:1 and 1:1.5 ratios. Prompt examples include a 1:2 ratio.
• Risk limits: Prompt examples include a maximum of $200 risk per trade and five trades per day.
• Fixed-stop examples: A 20-pip stop with a 40-pip target, or a 20-pip target.
• Claude plan costs: Examples discussed range from about AUD $300–$350 to $200, depending on the plan. Ruben recommends starting with the free plan and, if purchasing a paid plan, using it intensively.

Ruben says his own profit was about $1,000 while also acknowledging the challenges and losses.

Cautions and disclosures

Ruben repeatedly says Claude is not an “ultimate truth.” He advises traders to scrutinize its assumptions, verify its code and chart output, and keep refining the system. He also notes that profitability depends in part on risk management and that prop-firm accounts have different rules from personal accounts.

He says he will share both wins and losses for transparency. No formal “not financial advice” disclaimer appears in the subtitles, though he explicitly says he cannot make the cash-out decision for viewers.

Instruments, platforms, and sources

• Trading approach: ICT-style fair value gaps, including inverse FVGs. The subtitles do not identify a specific ticker or traded contract.
• Platforms and tools: Claude, Pine Script, TradingView, Quantis, Lucid, and Trade a Fire.
• Presenter and source: Ruben is the presenter; ICT teachings are referenced as the framework for defining fair value gaps.