Wins + Dingers
HR Board Results NFL SystemBox Office

Results

Who homered -- 2026-09-22

Board rankPlayerPark Model P(HR)Fair odds
#4 of 198De La Cruz, EllyATL15.1%+561
#5 of 198Alvarez, YordanSEA15.1%+562
#28 of 198Walker, ChristianSEA12.3%+716
#41 of 198Suárez, EugenioATL11.7%+755
#74 of 198Bellinger, CodyNYY10.9%+817
#60 of 198Murakami, MunetakaKC10.6%+847
#61 of 198Abrams, CJDET10.5%+850
#66 of 198Grichuk, RandalKC10.3%+866
#74 of 198Bellinger, CodyNYY10.1%+892
#75 of 198Stowers, KyleCHC10.1%+895
#104 of 198Paredes, IsaacSEA9.1%+999
#124 of 198Marsh, BrandonPHI8.2%+1124
#125 of 198Pasquantino, VinnieKC8.1%+1129
#143 of 198Doyle, BrentonKC7.4%+1242
#156 of 198Loftin, NickKC7.0%+1335
#159 of 198Myers, DaneATL6.7%+1395
#161 of 198Meidroth, ChaseKC6.6%+1408
Board rank is where the model placed the player among all 198 scored hitters that day -- low numbers were called; high numbers were surprises.

The day vs the model

Model expected
18.3
Actually homered
17
Predictions scored
198
BucketN PredictedRealized
0%-10%1227.5%5.7%
10%-15%7011.8%11.4%
15%-20%615.4%33.3%

Model scoreboard -- current version (v1.2-g075, since 2026-08-25)

Days tracked
29
Predictions
6,171
Predicted / realized HR rate
10.8% / 11.7%
Model bias vs reality
-7.2%
Top-10 picks: HRs vs expected
57 / 50.6
BucketN PredictedRealized
0%-10%26408.1%7.5%
10%-15%295112.1%14.3%
15%-20%55916.5%17.0%
20%-25%2121.3%19.0%

Previous version (v1.0 -- 2026-08-03 to 2026-08-24, retired after level recalibration)

Predictions
4,761
Predicted / realized
12.2% / 11.1%
Bias vs reality
+9.7%
Top-10: HRs vs expected
41 / 50.5
BucketNPredictedRealized
0%-10%17067.6%8.4%
10%-15%186912.3%11.6%
15%-20%88917.0%12.8%
20%-25%23022.0%15.7%
25%-100%6728.2%28.4%
Brier score 0.1016 vs constant-forecast baseline 0.1030 (beating it -- lower is better). "Model bias vs reality" is the running gap between what the model predicted and what happened -- positive means the model runs hot; it needs roughly two weeks of days before it stabilizes into a trustworthy number. Every prediction is frozen at publish time and scored against Statcast the next morning; scratched players are excluded. Small samples early -- calibration takes weeks.