← THE CHARTSPOD LEDGEROPEN TERMINAL →
AI · BACKTESTS

AI Crypto Backtests: Check Fills, Fees and the Number of Trials

An AI can run a backtest in seconds. Whether the result means anything depends on fills, fees and how many variations it tried.

By ChartsPod Editorial Team · Reviewed by ChartsPod Research
Published · Updated · 5 min read

Short answer

When an AI agent runs a crypto backtest for you, check four things before trusting it: the fill model (next-bar open with slippage, not the signal bar's close), costs (fees, slippage and funding for perpetuals), the number of trials (how many parameter sets or rule variants were tried before this one) and the sample (trade count, period, markets). Ask the agent to report all of these and to show the trade list. A strong curve from many trials with optimistic fills is likely overfit.

Four checks

CheckAsk the agentRed flag
Fills'When is each trade filled?'Fill at signal close
Costs'Which fees, slippage, funding?'None or tiny
Trials'How many variants did you test?'Many, only best shown
Sample'How many trades, which period?'Under ~30 trades
Auditing an AI backtestAuditing an AI backtest. Steps: 1. Rule stated — Exact logic; 2. Fill model — Next bar + slippage; 3. Costs — Fees, funding; 4. Trials — Count all; 5. Trade list — Inspect1Rule statedExact logic2Fill modelNext bar + slippage3CostsFees, funding4TrialsCount all5Trade listInspect
Each step can turn a great result into an ordinary one. Original ChartsPod editorial graphic.

Example

Illustrative: the agent reports +64% for a breakout rule. After asking: fills were at signal close, no fees, 40 variants tested, 22 trades. Re-run with next-bar fills, 0.1% round-trip fees and funding: +9%, and it was the best of 40.

Illustrative return after each correction (%)Illustrative return after each correction (%): As reported +64 %; + realistic fills +28 %; + costs +9 %.ILLUSTRATIVE RETURN AFTER EACH CORRECTION (%)As reported+64 %+ realistic fills+28 %+ costs+9 %
Then discount for 40 trials and 22 trades. Original ChartsPod editorial graphic.

Prompt

Illustrative: 'Backtest this rule on venue A BTCUSDT perp 1h for 2024–2025, signals on closed bars, fills at next open with 5 bps slippage, 0.05% fee per side, historical funding at settlement times. Report trade count, net return, max drawdown, profit factor, and list every variant you tried.'

Related: How to Backtest Crypto Strategies Without Lookahead Bias, Parameter Optimization: Report All Trials, Not Just the Winning Curve, ChartsPod Backtesting: Set Fees and Slippage Before Comparing Results.

Conclusion

Make the agent report fills, costs, trials and sample size, and judge the backtest only after those are realistic.

Run an AI-assisted backtest with explicit costs and fills.

Open the ChartsPod terminal

Compare AI Crypto Backtests across Binance, Bybit, OKX and Hyperliquid markets.

Open markets in ChartsPod

Questions readers ask

Can I trust an AI backtest?

Only after checking fills, costs, trials and sample size.

Why ask how many variants were tried?

The best of many trials looks good by chance.

What fill model is realistic?

Next bar's open plus slippage for market entries.

Sources and methodology

ChartsPod's AI agent (69 tools, visible tool calls, bring-your-own-key, no trading) and MCP endpoint were checked on its public pages on 17 October 2026. Prompt examples are illustrative templates; outputs vary by model and data.

  1. ChartsPod: product overview, pricing and FAQ — free terminal, venues, modules, no order placement; checked 17 Oct 2026.
  2. TradingView Pine Script: Strategies — simulated fills and assumptions.
  3. ChartsPod documentation — guides for order flow, derivatives, alerts, backtests, journal and workspaces.

This article explains research tools and workflows. It is not financial advice, does not recommend buying or selling any asset and does not predict prices. Crypto derivatives carry a risk of rapid and total loss of margin.

CHARTSPOD EDITORIAL EDITION · RESEARCH AND EDUCATION ONLY · NOT FINANCIAL ADVICE.