bin/Custom/Strategies/SentinelTBarsEdgeProbe_v1_0_0.cs
| Family | Strategies |
| Version | 1.0.0 |
| Size | 121 lines |
| Scope | public — ships in sentinel-suite |
| Class | SentinelTBarsEdgeProbe_v1_0_0 |
| Namespace | NinjaTrader.NinjaScript.Strategies |
| Documented by | no doc tracks this artifact |
Rendered from the published copy in
sentinel-suite/src/, not the author’s private tree — so this page describes the file you actually have.
Sentinel TBars Edge Probe — the PHASE-A honest-fill test | Version v1.0.0
File: SentinelTBarsEdgeProbe_v1_0_0.cs | namespace …Strategies (base — strategies do NOT sub-namespace)
THE ONE QUESTION THIS ANSWERS: is the ~66% first-touch win rate on TBars 6/24 a REAL,
TRADEABLE edge, or an artifact of the excursion recorder's bar-level label?
METHOD — deliberately DUMB + COUNCIL-FREE, so it isolates the BARS (no 21-voter confound):
• Signal = the TBars brick direction (sign of Close-Open). Enter on a FLIP (default) or every brick.
• Barrier = a SYMMETRIC R target/stop, R = max(MinTicks, AtrMult × ATR(AtrPeriod)) in ticks —
this MIRRORS the recorder's ATR-scaled first-touch barrier, so a win here == firstTouch=+1 there.
• One position at a time (act only when flat) → each trade is an independent first-touch barrier test.
• Managed SetProfitTarget/SetStopLoss so NT's own fill engine resolves the barrier.
⚠ HOW TO RUN IT HONESTLY (the whole point): Strategy Analyzer ▸ your GC SentinelTBars 6/24 series ▸
**Order Fill Resolution = High, 1 Tick** (NOT the default bar-close fill — that would just reproduce the
optimistic bar-level label). Set Slippage to your real cost. Compare the win rate + expectancy to the
corpus 66%. If it SURVIVES → the bars carry the edge (advance to Phase B: variants + session overlay).
If it CRATERS (à la CompressionBase 81%→37.5%) → the 66% was label optimism, and we've saved months.
NO Sentinel plumbing on purpose: no SentinelCore, no card, no seam — a clean, independent instrument.
CHANGELOG
v1.0.0 (2026-07-13) — NEW. Phase-A honest-fill probe of the TBars edge.