The rating softball never had

Every half-game of the season is one observation in a single ridge regression. A team’s offense, every team’s defense, and one home-field term are solved together, so the rating is what a team would do against an average D1 opponent on a neutral field.

Two more pieces sit inside that same solve. Each conference gets its own offense and defense term, so a league that only plays itself is still measured against the rest of D1. And each team starts from its final 2025 rating instead of from zero, which steadies the early weeks when a team has played a handful of games. Tested walk-forward on eight seasons, the two together lifted game-by-game accuracy from 69.6% to 71.9%, tournament games included.

runs_scored = 5.09 + O[offense] + D[defense] + 0.03 x home

home = +1 listed home, -1 listed away
O, D include the conference term and the 2025 prior

O > 0  scores more than average
D < 0  allows less than average
DiamondIQ = 50 + 47 x (O - D) / (best O - D)

Why not just iterate

The obvious approach — adjust offense by opponent defense, then defense by opponent offense, repeat — is a feedback loop. Good defenses inflate their opponents’ offensive ratings, which deflates those defenses further. It walked the best offense out to 12 runs a game before settling. One joint solve with a ridge penalty does not have that failure mode.

What the rating is not

It is predictive, not a résumé. It does not know who won a trophy and does not care. That is why the national champion sits fifth and a 32-28 team sits inside the top 25 — and it is the entire reason to put it next to a poll rather than in place of one.

Known limits

NCAA scoreboards do not flag neutral sites, so every neutral-site game — early-season tournaments and the WCWS included — carries a home-field term it should not, worth 0.03 runs either way. For the same reason the home and away splits here are the listed home and away teams, not true home and road records. Only games between the 304 teams in the NCAA’s D1 stat tables with a final score on the scoreboard count. The scoreboard lists some games — mostly conference tournaments — with no score, leaves a few out, and a handful are against programs outside those tables, which is why the records here run zero to 4 games light of the official NCAA totals. RPI on this page is computed from the same game log; it is not the NCAA’s official RPI.