The Efficiency Trap: Measurable Ex Post, Not Capturable Ex Ante

This study consolidates and extends the contract-level efficiency analysis of Bixhili (2026) on an expanded frozen dataset of 420 resolved Kalshi contracts, up from 244. The design is deliberately adversarial: four trading strategies — mean-reversion, momentum, favorite/longshot, and a simplified, fully specified proxy of the APEX ZENITH system — are evaluated at a simulated entry point two-thirds through each contract's life, gross and net of Kalshi fees and a simulated 2-cent slippage. Three results emerge. First, naive strategies deceive: mean-reversion shows a mean return of +56.88% but a median of +0.0101 per contract — one cent — and momentum a mean of +8.35% against a median of -100%. Profits live in a thin right tail of longshot resolutions and are not capturable ex ante. Second, the APEX ZENITH proxy trades in only 9 of 420 markets (abstention rate 97.9%) and loses when it trades (mean -15.05%; -17.05% net of costs). Third, execution costs narrow every apparent edge and invert none of the qualitative conclusions. The “Efficiency Trap” hypothesis — that the institutionalization of prediction markets (NBA and S&P DJI partnerships) is eroding algorithmic edge — is consistent with these numbers but is not formally tested here: the required temporal-trend analysis was precluded by anomalous or missing date formats in the 68 newly added markets, a limitation declared openly rather than buried. The paper is a consolidation step, not an arrival point: its contribution is a fully disclosed, larger-sample replication of the V1.0 negative results, and a precise agenda — temporal-trend testing, out-of-sample validation, live paper trading, cross-platform extension — for the pre-registered V3.0 study.