Real backtesting means flipping through hundreds of charts, marking every setup by hand, arguing with yourself about whether a wick "really" counted. Sentio replaces the whole ritual. Define a model — pick a built-in or describe your own — and hit validate. Sentio sweeps the cached history, lists every bar where the model fired, and lets you watch each one back with the exact structural geometry the detector saw. Watch, refine, re-run. Minutes, not weekends.
This is the workflow that replaces "spend Sunday clicking through charts." Each cycle takes minutes — refine one parameter, hit validate, watch what changed. Iterate until the model does what you actually want it to do, then activate it live.
Pick a built-in model or describe your own in plain English. Sentio's AI compiles the intent into a deterministic DSL — no code, no charting plugins, no CSV files.
"When ES sweeps PDH on 15m and NQ fails to confirm, wait for a bearish displacement and enter on a 5m FVG retrace."
One click sweeps the cached ~3,000-bar buffer, runs the detector on every bar, and lists every historical instance where the model would have fired.
Every fire is a replay. Scrub bar-by-bar, see the sweep, the gap, the displacement drawn on the chart at the instant it triggered. Judge each one for yourself.
Tighten a threshold, widen a session, add a confluence check. Re-run validation in seconds and compare the new trigger set against the old side-by-side.
The replay engine is not a chart with a slider on top. It re-runs the detector and the discipline layer on every bar you step to, so what you see is what the live scanner would have seen at that instant.
Advance one candle at a time. Every model re-evaluates on each step, so you see the exact bar it activates, gates, or invalidates on — no post-hoc rationalisation.
The chart pairs the execution frame with its context frame automatically (5m ↔ 1h, 15m ↔ 4h, 1m ↔ 15m). Structure on the higher frame is drawn in the same view.
Sweeps, FVGs, order blocks, breaker zones, session H/L levels — every structural primitive the model uses is drawn on the chart at the bar it fires. Nothing to imagine.
Custom models built in the DSL render every primitive from the spec as a labeled overlay. You can see exactly which condition matched on which bar.
Your discipline rules replay alongside the setup. Watch a rule trip in real historical time, see the gated candidate, and decide whether the rule needs tightening or loosening.
Each model timeframe carries a cached ~3,000-bar buffer, refreshed at session boundaries. Enough history to validate an intraday model over several months of the actual session you trade.
The point of replay is not to admire the chart — it is to build conviction that the model does what you think it does, and to catch the setups where your discipline rules are too tight or too loose. Every replay run can end in a journal entry that feeds directly into the tuning of the model or the rule.
Load any built-in or custom model on any of its supported timeframes. The buffer is loaded from cache — no network wait.
Jump to a specific bar via the scrubber, or step through with the arrow keys. The chart re-renders instantly.
Every bar, the detector re-runs. Overlay geometry appears the instant a condition matches — sweep, gap, displacement, breaker.
A side panel shows the full check trail for the current bar: which primitives passed, which failed, and the parameters that produced the verdict.
Save the setup with a note — "this fires cleanly on the London-open sweep, but the NY re-open version is trash" — for post-session review.
Replay is included on every plan — free forever.