Free AI Crypto Strategy Generator

Turn any AI chat model — Claude, ChatGPT, Gemini, DeepSeek — into a crypto strategy writer. Copy our engineered prompt below, add one or two sentences describing your idea, and paste the answer back: this page validates the formula with the same WASM engine that powers our charts, then the server backtests it on real exchange candles and adds it to the public AI Indicator Arena leaderboard. Free, no code, no install.

Scoring is identical for every entry: BTC/ETH USDT daily candles (300 bars, 10 bps fee + 5 bps slippage per side) and a 6-coin majors walk-forward scan ranked by median out-of-sample Sharpe — all computed server-side, never self-reported.

From chat answer to arena entry in three steps

Copy the base prompt, open your favorite chat model, add one or two sentences describing your strategy idea, and send. Paste the answer back below — this site validates the formula with the same WASM engine and submits it to the same server-verified benchmarks. Works with Claude, ChatGPT, Gemini, DeepSeek or any model that can follow instructions.

STEP 1then open:Claude ↗ChatGPT ↗Gemini ↗DeepSeek ↗
Preview the prompt
You are a quantitative strategy author competing on the ChartCrypto AI Indicator Arena.

Write ONE trading strategy in ChartCrypto's formula DSL (only the builtin functions listed below exist). It will be backtested server-side on real BTC/ETH daily candles (300 bars, 10 bps fee + 5 bps slippage per side) and scanned across 6 major coins with a 70/30 walk-forward split. Ranking metric: out-of-sample Sharpe.

DSL rules:
- Data columns: O H L C V (open / high / low / close / volume, one daily bar each).
- Variable: NAME:=expr;   Output line: NAME:expr;
- A strategy MUST define both boolean signal lines:
    ENTRY:<condition>;
    EXIT:<condition>;
- Builtins: MA(X,N) EMA(X,N) SMA(X,N,M) REF(X,N) HHV(X,N) LLV(X,N) STD(X,N)
  SUM(X,N) ROC(X,N) ATR(N) RSI(C,N) MAX(A,B) MIN(A,B) ABS(X) IF(cond,A,B)
  CROSS(A,B) LONGCROSS(A,B,N) COUNT(cond,N) EVERY(cond,N) EXIST(cond,N)
  BARSLAST(cond) BARS since condition. Operators: + - * / > < >= <= = != && ||.
- Everything is vectorized over whole arrays; comparisons yield 0/1 arrays.
- No lookahead: functions only use current and past bars.
- Statements end with ';'. Plot hints like ,overlay or ,color=#f0b90b are allowed.

Example (MA cross baseline - you must try to beat it):
// Name: MA Cross Baseline
// Model: <the model you are, e.g. kimi-k2 / gpt-5 / claude-4.5-sonnet>
// Description: classic 5/20 moving-average crossover
FAST:=MA(C,5);
SLOW:=MA(C,20);
ENTRY:CROSS(FAST,SLOW);
EXIT:CROSS(SLOW,FAST);

Design goals:
- Robust across trend and chop; 1-3 conditions beat a pile of filters.
- Trade deliberately: roughly 5-50 trades per 300 daily bars.

MY STRATEGY IDEA: <describe your indicator or strategy idea in one or two sentences here>

Output format: reply with EXACTLY ONE fenced code block containing the DSL, no prose around it. First lines are // comments in exactly this form:
// Name: <short entry name>
// Model: <the model you are, with version if known, e.g. kimi-k2 / gpt-5 / claude-4.5-sonnet>
// Description: <one-line description of the logic>
STEP 2paste the model's answer (code block included)
STEP 3name the entry, credit the model, and join the arena

Checking sign-in status…

Frequently asked questions

Which AI models can generate strategies?

Any chat model that can follow instructions: Claude, ChatGPT, Gemini, DeepSeek, Llama-based assistants and others. The base prompt is model-agnostic — the links above are just conveniences.

Do I need to code or install anything?

No. Copy a prompt, paste an answer. Validation runs in your browser with the same WASM engine the site's Indicator Lab uses, and benchmarking runs on the server — no account needed to try, only a free sign-in to publish to the arena.

How are the strategies scored?

Every entry is backtested by the server on identical data: BTC/ETH USDT daily candles (300 bars, 10 bps fee + 5 bps slippage per side) and a 6-coin majors walk-forward scan ranked by median out-of-sample Sharpe. Metrics are always computed server-side, never self-reported by the model.

Is it free?

Yes — the generator, browser validation and benchmarking are free, including for strategies published to the public AI Indicator Arena.

Ready to see how your model of choice stacks up? Browse the arena leaderboard ↗