🏁 DOC Race Lab v1 · static-grounded strength & best-trip estimator (244 CPU horses)

Model card. Parts of the DOC race formula are recovered (SH-4 static reads plus a live GDB trace, see _sh4/decode/*.md), but this page is a model built on them, and several of its inputs were corrected after it was built (2026-09-30): (1) its 4-band curve (≤115: 0.0030, 116-120: 0.0035, 121-127: 0.0042, ≥128: 0.0052, rising with the band) is indexed in the game by the internals sum (speed + stamina + sharp), not by the dirt rating; this page still applies it to the dirt rating as a stand-in surface term, and its turf branch is invented; (2) the 12-key distance table @0x10F204 is a pace multiplier (1.0323 = 1600m), but distance × multiplier is not constant: it tends to rise with distance; (3) the 220-300 ladder the sim uses as a stamina drain is really the game's training damper; (4) the whip has no grades: the game reads one whip input; (5) the race does not run in six fixed phases: the sim's phase split is the old player model, while World Edition's race code (Rev C, spot-checked in Rev D, read twice independently; effect sizes not measured at a cabinet) has each external work in its own situation (Start at the gate and in the opening seconds, Corner in the turns, Out of the Box whenever you whip with three or more horses close, Competing with two or more close and never while leading, and Tenacious while the gauge is nearly full, both in the last 500 to 1300 metres, and Spurt only in the final 280 metres (430 from 2000 metres up) while you whip); (6) the 92-99.5 "condition" breakpoints are windows on an in-race gauge, not on the horse's condition; (7) the sim's internals are the old reading (speed a ceiling, stamina a drain, sharp acceleration): in the game stamina sets how fast the gauge fills while you whip (urging is the whip), speed adds a little to the top speed while you whip, and sharp has no effect of its own in World Edition because a programming slip gives every horse the same value; (8) the sim leaves out several verified race-day terms: hearts (a small bonus in the last 500 to 1300 metres for player horses), PERFECT (a whip in the first 14 ticks after the gates open doubles early acceleration), the top-two streak cost, and Rev D's handicap on strong young horses; (9) the sim has no catch-up term: until 2026-09-30 it sped up the horses still running once the leader had finished, to bunch their finishing times, but World Edition Rev C's race code has no such catch-up, so the term was removed and the gaps behind the winner are now the model's own. What the game does have is a late speed-up in the day's 6th race (6R) for 44 computer horses whenever they are not near the front (37 get a small one; seven get a bigger one that also works closer to the front: El Condor Pasa, Hit Maker, Big Shuttle, Helissio, Sunday Silence, French Deputy and Bubble Trouble) and, on a cabinet set to a Hong Kong race mode, a small speed bonus for every computer horse in every race (a little more in 3R and 6R). The sim has no race number and no race-mode setting, so it models neither. How much externals weigh against internals also rests on one tuning constant, and that balance flips under another calibration. This page includes BOTH the transparent strength & best-trip estimator (aggregate, normalized 0 to 100) AND a tick-by-tick simulator (the ▶ Run race / Odds buttons) ported from _sh4/sim, which integrates per-tick velocity, a stand-in stamina drain and the [10,160] clamp across the race from each horse's roster stats. The sim runs at real-racing pace: average speed is held near-constant across distances (~9% sprint→stayer fade, matching real thoroughbreds), anchored to the measured 1600m turf time (winner 48.5s). The Realistic field mode uses DOC's grade + surface groupings in a competitive class band; the gaps behind the winner are the model's own (they grow with distance), not the game's.