Market calibration

What the devigged DK close implied, against what actually happened — split by side and by price rung. This grades the market, not the model. Built 3:48 AM · 2026-08-05.

Both sides

50.0%

vs 50.0% implied (+0.0 ± 1.8) · n=806

Home

51.4%

vs 53.1% implied (-1.8 ± 2.5) · n=403

Away

48.6%

vs 46.9% implied (+1.8 ± 2.5) · n=403

Implied is the DK closing moneyline devigged to a fair probability, so it already excludes the hold; the gap between it and the raw price is what you pay to bet. Each graded game contributes two observations — the home side at its close and the away side at its close — because a price is what is being graded, not a game. ± is one standard error on the realised rate. A rung whose gap is smaller than about twice its ± is indistinguishable from a fairly priced market; at these sample sizes most of them are, and resolving a real favourite-longshot bias would take thousands of games rather than hundreds.

Price rung	Home side	Away side	Both
≤ -250	66.7%vs 71.2% implied (-4.5)n=12 · ±13.6	—	66.7%vs 71.2% implied (-4.5)n=12 · ±13.6
-249 to -175	60.9%vs 63.8% implied (-2.9)n=46 · ±7.2	68.8%vs 63.3% implied (+5.4)n=16 · ±11.6	62.9%vs 63.7% implied (-0.8)n=62 · ±6.1
-174 to -130	58.3%vs 56.7% implied (+1.6)n=127 · ±4.4	60.0%vs 57.0% implied (+3.0)n=45 · ±7.3	58.7%vs 56.8% implied (+1.9)n=172 · ±3.8
-129 to -100	51.6%vs 51.2% implied (+0.4)n=124 · ±4.5	52.9%vs 50.6% implied (+2.4)n=119 · ±4.6	52.3%vs 50.9% implied (+1.4)n=243 · ±3.2
+100 to +129	35.9%vs 45.6% implied (-9.6)n=64 · ±6.0	47.8%vs 44.8% implied (+3.0)n=138 · ±4.3	44.1%vs 45.0% implied (-1.0)n=202 · ±3.5
+130 to +174	29.6%vs 39.1% implied (-9.4)n=27 · ±8.8	36.8%vs 39.0% implied (-2.2)n=57 · ±6.4	34.5%vs 39.0% implied (-4.5)n=84 · ±5.2
+175 to +249	66.7%vs 34.0% implied (+32.7)n=3 · ±27.2	25.0%vs 32.5% implied (-7.5)n=24 · ±8.8	29.6%vs 32.7% implied (-3.0)n=27 · ±8.8
≥ +250	—	50.0%vs 26.0% implied (+24.0)n=4 · ±25.0	50.0%vs 26.0% implied (+24.0)n=4 · ±25.0