Backtest
Replay the real deterministic spine over historical bars — no lookahead, conservative fills. A lab to test configs before risking capital.
idleRead before you trust a number
- Survivorship-biased. Uses today's watchlist, not point-in-time membership. Real-world results would be lower.
- Excludes the catalyst / LLM layer. Deterministic spine only (sentiment 0, event-risk 0, earnings-blackout off).
- Best case under known history. A backtest is the ceiling, not the expectation. Over-tuned results are curve-fit, not edge.
- Drawdown beside returns, always. A return without its max drawdown hides the risk taken to get it.
Configuration
Date range
Strategies
Regime gate (test the gate, don't assume it)
live: momentum in TRENDING, mean-reversion in RANGING
Symbols (blank = live watchlist)
Tunables
Run history
sweep #3alpha col verify2/2
sweep #2VERIFY regime sweep2/2
sweep #1lab sweep3/3
#11benchmark verify-0.34%done
#8VERIFY default-0.84%done
#7VERIFY mr ungated-2.05%done
#6VERIFY mr gated-0.84%done
#5lab run-0.79%done
#1lab run-3.69%done