Where the same bet is cheaper today

Latest bid/ask midpoint on each venue, ranked by the gap left after round-trip taker fees:

fee_adjusted_cents = |mid_k - mid_p| - 200 * (0.07 * Pk * (1 - Pk) + rp * Pp * (1 - Pp)), where Pk and Pp are the Kalshi and Polymarket mids in dollars and rp is the Polymarket market's own catalog rate (0.03-0.07 by category; fee-free markets pay 0). Kalshi's fee is its general schedule, 0.07*P*(1-P) per side; both legs are doubled for a round trip (entry + exit). Every match is gated and ranked at the conservative maximum rp=0.07 first - a catalog read for all ~10.9k matches would OOM the export's memory budget - then the tens of pairs that pass the gate are re-priced at each market's real rate, so the fee-adjusted number can only rise on re-pricing and no pair is ever admitted that its true rate would have excluded. A market with no readable schedule keeps the maximum and says so in fees.pm_basis ("assumed_max" vs "catalog" vs "fee_free"). Only gated pairs (pairs, inside_fee_line) are re-priced this way, and only they carry the fees block; suspect pairs are rejected before re-pricing runs, so their fee_adjusted_cents stays at the conservative maximum and carries no fees block at all - two rows at identical mids can therefore show different fee_adjusted_cents depending on which list they are in, and a bare fee_adjusted_cents with no sibling fees block is how to tell a suspect row's number from a gated one's. fork_net_cents instead prices both legs of a locked fork off the book (YES ask on the cheap venue, NO = 100 - YES bid on the other) and subtracts ONE-WAY taker fees per leg: Kalshi ceil(7 * P * (1 - P)) cents per contract, Polymarket rate * P * (1 - P) * 100 cents at the market's own catalog rate (0.03-0.07 by category; fee-free markets pay 0; a market with no readable schedule is charged the 0.07 maximum and says so in fees.pm_basis). P is the leg's executed price in dollars - a fork holds to resolution, so there is no exit leg.