paper mode · master kill switch: WB_PROPHECY_MASTER_ENABLE
Same as oracle_pythia_quality but fires only in 0-2h or >72h bands (gate-blocked complement)
Empirical: Pythia 0-2h band 70.5% WR (n=44) vs 12-72h 57.2% WR — gate may hurt
60.8%
n=199
+$124
no edge detected
Same as oracle_pythia_aggressive but in gate-blocked complement
Inherits Pythia A OOS 91.7% (n=229, +$0.20/$1)
77.2%
n=206
+$34.33
no edge detected
Same as oracle_cassandra but in gate-blocked complement
Inherits OOS 75.9% (n=29, +$0.425/$1)
—
n=0
$0.00
warming
BUY NO at 1-3¢ when p_mkt ≥ 0.95, no gate; same toxic-station skip
Inherits Sibyl backtest 96% WR · MARGINAL fee survival
0.0%
n=4
−$19.97
insufficient n
BUY YES at 1-5¢; uses include_post_end_hours=12 to recover docstring intent
DISABLED 2026-04-24 · ColdMath misclone (clones LOSING leg only)
0.0%
n=43
−$43.00
significant negative (fix or kill)
BUY NO at ≥95¢ POST end_date when temp observation has locked (5h buffer)
4 witnesses · 99.84-99.93% WR on $147k+ · NoonienSoong accelerating 2.68× in 2026
99.5%
n=186
+$3.80
no edge detected
BUY YES at p_mkt ∈ [0.15, 0.45] when p_pred − p_mkt ≥ 0.20 · standard Oracle window
1 witness (HC2 +$1.13/$1 on n=364) · single-witness flag
—
n=0
$0.00
warming
BUY YES + BUY NO same conditionId when sum < $1, simulate merge to $1 face
3 independent witnesses (Poligarch + neobrother + Lavincey)
50.0%
n=7384
+$1,230
significant positive
Why a separate generation? Two reasons.
Master kill: WB_PROPHECY_MASTER_ENABLE=false in .env disables ALL Prophecy scanners regardless of per-archetype flags.
See docs/strategies/prophecy/PROPHECY.md + per-archetype docs (HINDSIGHT.md, ECHO.md, FORGE.md, UNLEASHED_CLONES.md).