Fading miscalibrated price bands
Historical data showed some price bands win less often than their price implies — so sell into them. The calibration was real; the trade was not.
The hypothesisIf markets priced at, say, 60¢ historically resolve YES only 52% of the time, that band is systematically overpriced — fade (sell/buy-NO against) anything trading in it.
The test
Band-level calibration was computed from a large archive of resolved markets: empirical win rate per 10¢ price band versus the price itself. Bands with a persistent negative gap became standing fade signals; a paper daemon (later briefly graduated) sold into them and settled every position at resolution.
The result
The strategy lost $1,345 net before being disabled. The calibration tables were correct about the past and useless about the present: aggregate band-level mispricing is dominated by composition — which kinds of markets happened to sit in the band during the measurement window. When the mix shifted, the gap moved or closed, while the fade kept selling. A fade also pays the spread twice at entry and holds unlimited-downside positions against every genuinely-informed buyer in the band.
Conclusion
Calibration studies are good telescopes and bad trade signals. A band that was overpriced tells you where an edge once lived, not that it still does — and by the time a gap is statistically visible in a historical archive, the flows that created it have usually rotated. We still compute these tables; they now gate other strategies' entries instead of generating their own.