A companion to "Smart Money on Polymarket" (Paper I), this paper asks a sharper, methodological question: what is the alpha? We argue and document that every profitable wallet in our N = 273 panel is running the same underlying strategy-probability mispricing arbitrage-differing only in (i) which part of the price spectrum the mispricing lives, and (ii) what generates it. We formalise the unifying framework, decompose the eleven empirical regimes documented in Paper I onto four alpha sources-slow information assimilation, behavioural biases, microstructure inefficiency, and cross-market inconsistency-and show that aggregate profits map cleanly onto these sources. The slow-information bucket alone accounts for ∼ $140M of the $177M total realised PnL in our panel, dominated by Politics specialists holding underpriced high-probability YES shares to a $1 resolution payoff. Behavioural-bias arbitrage (long-shot, contrarian) is small in absolute PnL but the highest-Sharpe bucket. Microstructure (HFT/MM) and cross-market (sports model, weather model) buckets contribute moderately. The single useful insight for a practitioner is unambiguous: Polymarket is not a casino and not a high-frequency arbitrage venue; it is a venue where positive expected value comes exclusively from being correctly contrarian to a measurable mispricing in the implied-probability surface.