Normal prediction markets predict what will happen. Decision markets predict the consequence of a choice — here, how a candidate would do in the general election if their party nominated them. That’s the question a nomination actually turns on.
We take the odds from several live prediction markets and isolate that conditional — a candidate’s chance of winning the presidency divided by their chance of winning the nomination — then combine the markets with volume weighting and Bayesian shrinkage into a single estimate. It’s built entirely from live markets: no hand-picked coefficients, no thresholds, no polls. In theory, the most truthful number you can get.
It’s conditional, not causal — “if nominated” only counts the futures where a candidate already broke through, which flatters longshots. And some races trade on thin volume, so those numbers are noisier than they look.