EGOLDSv4
Sign in
Back to strategy
Backtest replay · 500 closed theses

Tune the desk gate, see history flip

Pure deterministic re-classification over the last 500 closed theses. Adjust R:R thresholds per market regime and the minimum target distance in ATR multiples — every change re-runs the same logic the desk's Risk Manager applies in production. No LLM calls, no inference; your params + historical price levels = WR you would have realized.

New WR (your params)
70.0%
35W · 15L · 79N
Approval rate
25.8%
129 / 500 pass
Old flat 1.5 WR
62.0%
81 / 500 pass
Δ WR vs flat
+8.0pp
pop WR: 58.0%
Tune parameters
Min R:R · trending (ADX≥25)
1.01.21.31.51.8
Min R:R · chop (ADX 15-25)
1.21.51.82.0
Min R:R · high-vol (ADX<15)
1.51.82.02.5
Min target × ATR
off1.5×2.0×2.5×3.0×
Approved by regime: trending 123 · chop 6 · high-vol 0Reset to desk defaults
Approved theses · 129
AssetSideConvR:RTgt × ATRRegimeOutcomePnL
SOLshort621.75:13.0×trendingneutral+0.01%
DYDXlong522.62:12.7×trendingneutral-2.30%
ETHshort522.00:12.8×trendingloss-1.96%
PYTHshort741.28:1—trendingloss-3.76%
FARTCOINshort681.24:1—trendingneutral+0.32%
VIRTUALshort622.50:12.5×trendingwin+3.09%
ZEREBROlong641.21:1—trendingloss-7.41%
INITlong551.91:1—trendingloss-3.03%
ZEREBROlong551.91:1—trendingwin+32.17%
ZEREBROlong551.91:1—trendingneutral-1.88%
INITlong551.75:1—trendingloss-4.45%
INITlong551.75:1—trendingloss-4.45%
FILlong582.33:13.5×chopneutral-2.09%
INJlong621.53:12.6×trendingneutral-0.67%
TRXlong581.47:1—trendingneutral+1.10%
TRXlong551.83:1—trendingneutral+0.63%
TRXlong621.67:1—trendingneutral+0.90%
TRXlong581.24:1—trendingneutral+0.85%
GRASSlong551.67:12.5×choploss-4.24%
TRXlong551.33:1—trendingneutral+0.86%
TRXlong551.33:1—trendingneutral+0.42%
TRXlong552.27:1—trendingneutral+0.37%
JUPlong552.06:1—chopwin+9.24%
AEROlong582.00:13.0×trendingwin+12.80%
CELOlong551.44:1—trendingloss-5.19%
POLlong581.63:1—trendingneutral+0.48%
INJlong552.18:13.4×trendingwin+11.44%
POLlong551.59:1—trendingwin+1.63%
POLlong552.09:1—trendingneutral-0.13%
PYTHlong681.23:1—trendingneutral-3.41%
PYTHlong621.60:1—trendingwin+2.48%
MASKlong621.60:14.0×trendingneutral+6.75%
MASKlong652.25:14.5×trendingwin+2.29%
IMXlong581.50:1—trendingwin+4.83%
MASKlong651.75:13.5×trendingwin+4.36%
TIAshort551.67:12.5×chopwin+6.50%
GRASSlong582.00:13.0×trendingwin+8.54%
CELOlong551.77:1—chopneutral-0.80%
MASKlong621.40:13.5×trendingwin+4.36%
MASKlong681.40:13.5×trendingwin+6.12%
MASKlong581.75:13.5×trendingwin+4.04%
BLURlong741.24:1—trendingneutral-0.43%
MASKlong581.80:14.5×trendingwin+4.10%
MASKlong621.67:15.0×trendingwin+2.59%
AEROlong551.40:13.5×trendingneutral-0.13%
MASKlong622.33:13.5×trendingneutral+0.61%
SEIlong551.55:1—trendingloss-4.85%
GRASSlong621.40:13.5×trendingwin+14.40%
AEROlong581.46:14.7×trendingwin+2.03%
MASKlong581.50:13.0×trendingwin+5.10%

This replay applies only the gate logic retroactively. It does NOT re-run the LLM with new prompts. So if a thesis was generated with target too close (less than your ATR floor), it gets rejected here even though with a better prompt the desk would have aimed further. Use this tool to evaluate gate sensitivity, not to predict future LLM behavior. Static analysis · ~500 closed theses · ISR cache 30s.