Scoreboard

Every strategy backtest we’ve run, in one table — winners, maybes, and the dead. Profit factors are net of slippage on the stated window unless noted. Survivors held an edge through a full-year test with costs. In testing means promising on a shorter window with the year-long out-of-sample gauntlet pending. Failed means it died under honest testing. Failures stay on the board, because the failure rate is the base rate.

Survivors

Strategy	Market / TF	Result	Window
Opening Range Breakout (IB structure)	NQ / 1m	PF ~1.7	365d + slippage
IB bottom-quartile close, long	NQ / 5m	PF ~1.5 (NQ 5m only — died on 1m, 15m, and ES)	365d
Linear-regression stretch fade, short-only	NQ / 5m	PF 1.8 @ 90d → 1.3 @ 365d (thin)	365d

In testing

Strategy	Market / TF	Result	Window
Gap-down half-fill long	YM, ES, NQ / 5m	YM ~3.0 · ES ~2.6 · NQ ~1.8 (GC dead)	90d
IB break-retest (“IB Retrace”), cleanliness-gated	NQ / 5m	PF ~3.4 (day-of-week gate may be overfit)	~100d
Session VWAP reclaim, long-only	NQ / 5m	PF ~2.0 (ES ~1.3)	~100d
3pm power-hour continuation	NQ / 5m	PF ~3.0 (ES ~1.9; dead on 1m/15m and YM)	90d
First 5-minute FVG of the open, continuation	NQ / 5m	PF ~1.3, 66% win	~3.5mo
15-minute FVG first-touch bounce	NQ, ES / 5m	NQ ~2.1 · ES ~1.6 (YM/GC dead)	~3.5mo
BPR (balanced price range) rejection	NQ / 5m	PF ~2.0, 70% win (only ~10 trades)	~3.5mo
Gold VWAP bounce family (7 variants)	GC / 1m	PF 1.9–2.8, 41–54% win	in-sample
Gold BB + EMA + engulfing (best of 10 GC systems)	GC / 1m	Short side PF ~1.6	in-sample
Filtered opening-range breakout, long-only	NQ / 5m	PF ~1.1 (marginal)	~3.5mo

Failed

Strategy	Market / TF	Result	Why it died
“Magic hour” overnight sweep-reclaim fade	NQ / 5m	PF 0.86 (claimed 2.17)	No costs + cherry-picked hours in the original. Full autopsy
RSI mean-reversion, tight target / wide stop	NQ / 5m	1.40 @ 100d → 0.91 @ 365d	Fat left tail the short window never sampled
Tight-TP / wide-stop geometry (as a class)	NQ / 5m	Every variant collapsed at 365d	High win rate hides rare catastrophic losses; fits mean-reversion only
1-minute VWAP-cross trend flip	QQQ / 1m	PF 0.97 (long-only 1.09)	Regime-dependent; short side bleeds in an uptrend
Prior-day high/low sweep-and-reclaim	NQ / 5m	PF ~1.1	Too thin after costs
Multi-signal confluence engine	NQ / 5m	PF ~1.0	Stacked filters just chased; complexity ≠ edge
IB size-regime switch (trade IB differently by range size)	NQ / 5m	No effect	The regime split didn’t hold up in the data
FVG size filter (only trade “big” gaps)	NQ / 5m	Worse than unfiltered	Stop distance scales with gap size — filter raised risk faster than win rate
NQ shorts, breakout variants (as a class)	NQ / all	Negative edge	Fighting a market that goes up
Gap fill on gold	GC / 5m	Dead at every stop setting	Gold’s gaps don’t behave like index gaps

How to read this board

Rows move over time — that’s the point. When an “in testing” strategy clears (or fails) its year-long retest, it gets a verdict post and the board updates. A few patterns worth noticing across the full list: almost nothing survives the jump from 90 days to a full year; almost nothing transfers between instruments (what works on NQ dies on ES, and gold is its own animal); and every high-win-rate system with a wide stop eventually met the fat tail it was hiding. Numbers are rounded; exact parameter sets aren’t published for live survivors. Everything here is subject to the risk disclaimer.