← Academy home
Deep Dive · Dossier 4 of 4

Sentinel — The Parameter Dossier

Sentinel already stood the longest trial of any strategy in this system — four deep-dive parts of anatomy, forensics, stress tests and verdict. This dossier doesn't repeat that work; it consolidates the parameter evidence into the same format as the other three, so every roster member carries identical paperwork into the forward campaign.

S.1The defendant

Rules 2 gatekeepers (choppy filter, WAE power gate) + 7 voters in unanimity · full anatomy in Part A

Roster config USDJPY only · H1 · dead zone 20 pips · pip_value 0.01 · ATR×1.5 stop, 1:3 target · 3-bar cooldown · magic 7702

Evidence base ~1,876 trades over 5 years — PF 1.16 · +18.8% · max DD 4.2% · the deepest sample on the roster by a factor of 25

S.2Where each number stands

Every roster parameter, its origin, and the evidence it now carries:

parameteroriginevidenceverdict
dead zone 20 pipsRound-4 tune (EA default 30)total plateau 10–45 (Part C)keep — any value works
SMA period 50indicator conventionhilltop 50–65; 30 and 80 both worse (Part C)keep — well placed
slope window 3was hardcoded; now a parameterplateau at 2–3 (PF 1.16 both); 4–5 degradekeep
ATR×1.5 · 1:3roster risk suitbroad healthy region; only 1.0×2 corner fails (Part C)keep
pip_value 0.01the JPY trap, fixed in Round 4"asleep but innocent" — power gate mostly redundant with direction voters (Part B)keep — now symbol-derived on the bench
the 7-voter committeeMQL5 EA designNonLagDot & WAE-direction individually shadowed; WAE power gate matters (Part C)keep — redundancy is harmless armor

Note the contrast with the other dossiers: Sentinel is the only strategy whose parameters were all re-examined under exam conditions before this series began — and every one sits on a plateau, not a peak. That is what "why these parameters" is supposed to sound like.

S.3The exit question, revisited with this series' lens

The dossier series taught us to distrust exit stories, so Sentinel's got the full treatment this week: its EA-native trail (leave when Heiken Ashi or NonLagDot flips, ignore other dissent) was finally simulated properly. Result: native trail +19.2% · PF 1.15 · DD 5.1% versus the roster bracket's +18.8% · PF 1.16 · DD 4.2% — a statistical tie, with the bracket keeping slightly better drawdown and a defined target for risk sizing. Unlike RevertX (strangled by the wrong exit) or BreakoutGuard (saved by it), Sentinel is exit-robust: bracket, trail, or raw signal exits all land within a point of each other. The alpha is in the signals; the exit merely shapes the risk. Roster bracket stays.

S.4Per-currency improvement? Asked and answered

The dossier series' central question — "should this strategy differ per currency?" — was Sentinel's founding question. Parts B and C answered it more thoroughly than any sweep could: the edge does not transfer to any other pair at any tested setting (nine deep-red cells on the edge map), equally-drifting yen crosses lose anyway (the drift theory was falsified), and no per-pair parameter change rescues them. Sentinel is not a strategy with per-currency settings; it is a single-pair specialist whose "per-currency tuning" is the pair list itself: USDJPY, full stop.

S.5Dossier verdict

Sentinel — ruling

All parameters confirmed; nothing changes. Every number sits on tested plateaus with the roster's deepest evidence base, the exit is robust to specification, the real swap costs are immaterial (5-yr +18.8% → +17.0%), and the guardrails — kill criteria, pain contract, wobble bands — are already machine-readable in the roster file. Sentinel's open questions are not parametric but environmental: the 2026 river is running slow (backtest PF ~0.76, already inside the review band), which is exactly what the forward campaign's guardrails exist to adjudicate. One human task remains: El's eyeball check of WAE against an MT5 chart.

This closes the dossier series. The combined docket — seven items across four strategies — now goes to El for decisions before the forward campaign: pair narrowings (RevertX → GBPUSD; BreakoutGuard → USDJPY+GBPUSD), the Wilder-RSI variant, the RSI·USDJPY candidacy, the moving take-profit engine capability, RevertX's native-exit exam, and BreakoutGuard's code-vs-docstring reconciliation.