Context searches hundreds of variations of your strategy across months of real history, validates each one out-of-sample, and ranks the robust survivors. Promote a winner to shadow for live forward-validation before any real money follows.
Optimization
⏳ This runs on the server — you can leave this page or switch tabs and it'll keep going.
🗂 Past runs
⚙️
No optimizer runs yet
Pick a strategy and hit Optimize to search for better parameters.
✦ How it works
Context searches hundreds of variations of your strategy, proves the survivors on data they were never tuned to, and only then lets one near your money. Here's the flow — and the words behind it.
Search→Walk-forward→Holdout→Promote→Shadow-validate→Arm live
Search
Tries hundreds of variants of your strategy — which gates are on, how strict, and the stop / target / trail levels — using random or genetic (evolutionary) exploration.
Walk-forward
Each variant is tuned on one slice of history and scored on the next, unseen slice, sliding forward. It catches strategies that only look good in hindsight.
Holdout
The finalists are judged once on a final, completely untouched stretch of data — the true out-of-sample test.
Robust score
Ranks on profit factor + Sortino (risk-adjusted return), penalized for drawdown, with a minimum-trades floor — scored on the pessimistic 25th percentile of a Monte-Carlo reshuffle, so a lucky ordering can't win.
Deflated Sharpe
The overfit guard. It discounts the winner by how many variants were tried; below 0.90 the banner warns you it may just be the luckiest of many draws.
Regime
Optionally finds the best strategy per market state — UP tape, DOWN tape, or CHOP — instead of one brittle global answer.
Shadow validation
A promoted strategy paper-trades live going forward — real signals, zero hindsight. The final proof before any real money follows.
Arm to live
Once shadow has proven it, one click makes the optimized profile your active strategy. Your live signals adopt its tuned confirms, exits and ADX gate — your entry trigger is unchanged. Live cards badge it Optimized params so you always know what's running.
1Pick your strategy, symbol, a date window (use months of history) and a method, then hit Run optimization.
2Watch the leaderboard fill in. The green / amber banner up top tells you whether the winner cleared the overfit guard.
3Hit Promote → Shadow on a row you like — it begins forward-validating on live paper data immediately.
4When it has earned its keep in shadow, hit Arm to live on that same row — the optimized profile becomes your active strategy and your live signals start using its tuned guardrails. Entry trigger stays the same.
How shadow mode tests it. A promoted strategy never touches real money. Every signal it would fire is paper-traded live, in real time, with no hindsight — entries, exits and running P&L tracked exactly as if it were trading. Over days to weeks you see its real win rate, payoff and drawdown on data it has never seen. Only once it has earned its keep do you Arm to live — and even then your entry trigger never changes; only the tuned guardrails (confirms, exits, ADX gate) come along, and every live card it produces is badged Optimized params.