start_optimization_runStart an AI optimization run on one of your own strategies — repeatedly mutates the code, backtests each variant, and keeps only the improvements, validated on data no experiment trained on. Pass several pairs to run a multi-pair optimization that tests every candidate on every pair, so the result is not tuned to one chart. Charges your Strategy Lab wallet per experiment (plus a small charge per extra pair); multi-pair runs always show a quote first and are confirmed with its quoteId.
| Name | Type | Description |
|---|---|---|
strategyId | string | ID of your own strategy to optimize |
exchangeId | string | An exchange you own, used to source historical candle data |
symbol | string | Trading pair to optimize against. Defaults to BTC/USDT when symbols is not given; with symbols, it only picks the primary pair when it is one of symbols, otherwise it is ignored(default: BTC/USDT) |
symbols | string[] | Pairs to test on, primary first, on the same exchange account and market type. More than one pair tests every candidate on every pair; capped by your plan (Free 1 / Pro 3 / Trader 4 / Whale 5). Combinable with searchPresets once an admin prices the multi-pair preset search: every candidate is then tested on every pair and the search is priced per pair |
timeframe | string | Candle timeframe, e.g. 1h, 4h, 1d(default: 1h) |
maxExperiments | number | How many mutate-and-backtest experiments to run (1-30)(default: 20) |
scoringProfile | string | balanced, conservative, aggressive, or consistency(default: balanced) |
trainSplit | number | Fraction of the window used for training vs. out-of-sample validation(default: 0.7) |
userGuidance | string | Free-text steering for the AI mutations |
autoPromote | boolean | Auto-promote the result to stable if it clears a strict validation bar(default: false) |
acknowledgeCost | boolean | Confirm the wallet charge to actually start the run(default: false) |
quoteId | string | Multi-pair runs: the quoteId from the cost estimate the user approved, sent with acknowledgeCost true |
acknowledgedCost | number | Multi-pair runs: the estimated cost the user approved; used if the quote expired, and the run starts only if the fresh price is not higher |
Try asking your AI assistant any of these:
get_optimization_statusCheck progress of a Strategy Lab optimization run — current experiment, best score found so far, per
send_optimization_guidanceInject live, free-text steering into a running optimization — e.g. "focus on trend-following, not me
list_strategy_versionsList the full version history for one of your own strategies — every stored version, its validation
Set up the TradeStaq MCP server in under a minute.