Two weeks ago we promised paying customers that GameIQ would start showing predictions on the Slate and that the Report Card would track every single one of them. The numbers are now in for the first stretch of the season, and they are honest because we cannot edit our own homework after the game has ended.
Here is what GameIQ has done so far across the three sports we currently grade. 798 games called. 661 right. That is 82.8 percent overall, which is a number we are happy with for early May.
By the sport
Softball is leading the building. 284 games graded. 250 right. 88.0 percent accuracy with a Brier score of 0.134. The full softball Report Card lives here. Softball is the cleanest of the three this spring because it is the most predictable. The favored team usually wins. The upset rate sits at just under 12 percent.
Lacrosse is right behind. 238 games graded across boys and girls combined. 197 right. 82.8 percent accuracy. The lacrosse Report Card aggregates both genders so you can see one headline number for the sport. Boys came in at 82.0 percent on 111 games. Girls posted 83.5 percent on 127 games. Brier score on each side: 0.158.
Baseball is doing what baseball does. 276 games graded. 214 right. 77.5 percent accuracy with a 22.5 percent upset rate. The baseball Report Card sits here. Baseball is chaos in the best way. A starting pitcher having a bad inning can swing a 9 run game. We are within 10 runs of the final score on 89 percent of games, but the swing in any individual matchup is wider than the other two sports.
What the numbers actually say
Two things are true at the same time. Overall accuracy is in the low 80s. That is solid for the first two weeks of public predictions.
The biggest single miss so far happened in baseball, where we predicted a 7 run win and watched a team lose by 6. The Report Card has the full miss list sorted by how wrong we were on the margin. We do not hide them. The whole point of the page is to be accountable for the swings, not just the hits.
What you do with this
If you are a coach, the Report Card tells you when GameIQ trusts a number and when it does not. The calibration tab on each sport breaks every prediction into confidence buckets. When we say a team has a 70 percent chance to win, you can check whether the cohort of 70 percent calls actually came in around 70 percent. Right now most buckets are landing inside two points of where they should be.
If you are a fan, the practical takeaway is simpler. The favorites are mostly the favorites for a reason. The Cutline uses the same underlying engine to project tournament fields, and the seedings it spits out for May are already pretty close to where the rankings will end up.
The honest part
We did not build GameIQ to be 100 percent right. Nobody is going to be 100 percent right at predicting Maine high school baseball. We built it because before this season nobody was publishing what they thought was going to happen BEFORE the games happened, and then showing the receipts after. Every prediction GameIQ has ever made on the public site is in the database with the date it was made and the result it got compared to. That is what makes the number trustworthy. Not that it is high. That it is real.
Two weeks down. Plenty of season left for us to look smart. Plenty of season left for us to look stupid. The Report Card will keep posting either way.