The 2026 Sabermetric Season

Where we agree with the poll

23 of the poll's 25 teams also make ours, and both have Texas first. Average disagreement among shared teams is 2.4 places.

The independent read and the human consensus land in nearly the same place, which is reassuring for both. The disagreements are where it gets interesting: the poll is kinder to Mississippi St. Bulldogs (-6) than we are, and harsher on Florida St. Seminoles (+7).

Where we split

Ranked by the poll, not by us

• Saint Mary's (CA) Gaelsour #53
• Virginia Cavaliersour #26

Ranked by us, not by the poll

• South Carolina Gamecocks#23 for us, unranked by the poll
• Ole Miss Rebels#25 for us, unranked by the poll

Why Texas, and not Arkansas

On run margin alone Arkansas rated first. It won the season series in Austin and posted the better schedule-adjusted margin, but it did so by running up regional blowouts, then went 0-2 at the WCWS including an 0-11 loss to UCLA.

Once wins carry equal weight and postseason games count double, Texas takes the top line. Elo agrees emphatically: Texas gained 283 rating points after May 1, the largest climb in the top 40. They were not the best team all season. They were the best team by June.

Schedule is the great distorter

Grand Canyon went 51-9 and rates 24th. Missouri went 28-29 and rates 33rd. Twelve of our top 25 are SEC teams, and the five toughest schedules in the country all belong to the SEC.

Expected wins from each team's actual schedule track real records closely: Oklahoma outran its expectation by 1.9 wins, the largest gap in the top 60, and Texas underran by 0.4. There is far less luck in a 60-game season than the standings suggest.

What each event is worth

Run expectancy

Batting WAR

Minimum 100 plate appearances

#	Player	Team	PA	wOBA	wRC+	WAR
1	Jordan Woolery	UCLA	222	0.719	354	10.1
2	Megan Grant	UCLA	225	0.695	340	9.7
3	Katie Stewart	Texas	208	0.612	306	7.8
4	Taylor Shumaker	Florida	246	0.566	267	7.7
5	Rylee Slimp	UCLA	227	0.574	267	7.1
6	Jocelyn Erickson	Florida	237	0.551	258	7.1
7	Mya Perez	Texas A&M	190	0.608	296	6.8
8	Isa Torres	Florida St.	160	0.691	329	6.6
9	Aminah Vega	Duke	193	0.597	280	6.4
10	Taryn Kern	Stanford	188	0.608	278	6.2
11	Kaniya Bragg	UCLA	212	0.550	252	6.1
12	Jordyn Frahm	Nebraska	211	0.545	250	6.0

Pitching WAR

Minimum 100 batters faced

#	Pitcher	Team	BF	wOBA	WAR
1	Maya Johnson	Belmont	769	0.186	19.4
2	Ruby Meylan	Oklahoma St.	882	0.291	19.0
3	Jordyn Frahm	Nebraska	639	0.216	18.8
4	Keagan Rothrock	Florida	878	0.300	18.3
5	Teagan Kavan	Texas	771	0.295	17.8
6	Jocelyn Briski	Alabama	583	0.203	17.4
7	Alyssa Faircloth	Mississippi St.	682	0.265	16.8
8	Jori Heard	South Carolina	742	0.300	16.3
9	Peja Goold	Mississippi St.	620	0.254	16.0
10	Lyndsey Grein	Oregon	683	0.276	15.6
11	Madalyn Johnson	Georgia Tech	852	0.315	15.4
12	Maddy Azua	Texas St.	869	0.298	15.1

Best FIP, park-adjusted

Minimum 300 batters faced. FIP is refit for softball on 1,051 official pitcher seasons and in log form, so the sport's most extreme strikeout arms cannot be handed a negative run estimate.

#	Pitcher	Team	BF	FIP	Adj	K%
1	Maya Johnson	Belmont	769	1.44	1.45	49.3%
2	Jordyn Frahm	Nebraska	639	2.26	2.33	34.9%
3	Jocelyn Briski	Alabama	583	2.41	2.36	33.6%
4	Alyssa Faircloth	Mississippi St.	682	2.37	2.52	38.9%
5	Sage Mardjetko	Tennessee	463	2.56	2.60	35.2%
6	Izzy Kemp	Dayton	655	2.43	2.60	40.9%
7	Faith Aragon	New Mexico St.	719	2.98	2.65	30.7%
8	Alexis Jensen	Nebraska	580	2.60	2.68	35.0%
9	NiJaree Canady	Texas Tech	538	2.85	2.81	34.0%
10	Maddia Groff	Omaha	558	2.79	2.81	29.9%

Win Probability Added

Minimum 100 plate appearances. WPA leaders skew away from the top of the rankings, which is the metric working as designed: elite teams win by too much for late heroics to accumulate.

#	Player	Team	PA	WPA	Clutch
1	Hannah Di Genova	Nevada	156	5.07	+1.46
2	Sydney Stewart	Arizona	157	4.90	+0.95
3	Courtney Poulich	Robert Morris	154	4.47	+0.88
4	Emily LeGette	North Carolina	169	4.47	+0.02
5	Jocelyn Erickson	Florida	213	4.45	+0.94
6	Lauren Holt	Cornell	111	4.42	+0.78
7	Brooke Klosowicz	Penn St.	183	3.77	-0.07
8	Aminah Vega	Duke	180	3.74	+0.34
9	Megan Grant	UCLA	193	3.63	-0.66
10	Shelby Barbee	North Carolina	159	3.63	+0.67

What the simulation got right

Texas, the actual champion, comes out as the single most likely winner at 23%, and seven of the eight teams the model ranks highest by title odds did reach the WCWS. That is a good result for a model that never saw a bracket, and a reminder of how open this tournament is: the favorite still loses more than three quarters of the time.

Format matters more than seeding. Alabama and Texas separate from the field not because they were most likely to win any single game, but because double elimination gives the best teams a second life at exactly the moment a single upset would otherwise end them.

*The first two trips hold exactly nine batters each for any pitcher who reached a third, so those counts match by construction rather than by coincidence.

A steady decline, not a cliff

Each trip costs the pitcher. The second time is worth +0.013 wOBA to the batting side and the third +0.039 wOBA, roughly a third of a run per nine batters faced. Strikeout rate falls from 17.0% to 12.0% while home run rate climbs 40%. The effect is highly significant (t = 9.7) but gradual, so there is no single inning where the wheels come off.

Usage varies enormously. Penn State let a pitcher reach a third trip in 10% of outings; Arizona State in 86%.

Where the ball goes

Direction parsed from the text on 95% of the league's 354,000 balls in play. D1 average is 41% pull, 37% opposite field. Balls up the middle produce a .431 average, far above pulled (.333) or opposite-field (.266) contact.

#	Team	Pull	Oppo	GB/FB
1	Texas Longhorns	43.4%	33.1%	1.26
2	Alabama Crimson Tide	41.5%	36.3%	1.26
3	Texas Tech Red Raiders	44.7%	34.7%	1.17
4	Nebraska Huskers	42.1%	35.7%	1.23
5	Arkansas Razorbacks	40.8%	37.4%	1.63
6	UCLA Bruins	45.5%	31.8%	0.97
7	Oklahoma Sooners	44.3%	35.8%	0.94
8	Tennessee Volunteers	41.7%	39.2%	1.48
9	Florida Gators	40.5%	36.9%	1.35
10	Florida St. Seminoles	45.3%	36.4%	1.30
11	Georgia Bulldogs	41.3%	36.1%	1.50
12	LSU Tigers	46.7%	33.1%	1.38

Park factors

Fit by ridge regression with offense and defense controlled, so a park is never credited for hosting weak opponents, then regressed toward neutral by sample size. New Mexico State plays at 3,900 feet. Applying these moves individual pitcher FIP by up to 0.25 runs.

Home park	Run factor
New Mexico St. Aggies	1.24
Akron Zips	1.23
Western Ill. Leathernecks	1.22
UNI Panthers	1.22
Morehead St. Eagles	1.22
UNCW Seahawks	0.86
Sam Houston Bearkats	0.84
FGCU Eagles	0.84
California Baptist Lancers	0.82
UT Martin Skyhawks	0.76

Run expectancy is the wrong yardstick here

Average runs is not what a manager is buying with a sacrifice. In a tie game the objective is the chance of scoring at all, which is a different number and can move the other way. Both, measured on this season:

On the scoring-odds yardstick the sacrifice stops being a blanket mistake. A successful bunt from first and second with nobody out actually raises the chance of scoring, even while it lowers expected runs. Two of the seven sacrifice situations come out positive.

But the attempt is not the same as the success

That +0.9 point gain assumes the bunt works. The offer on the table is the attempt. From first and second with nobody out, across 312 sacrifice attempts, only 17% actually produced second and third with one out. The rest popped up, got the lead runner, or found a hit.

So the honest verdict survives the better yardstick. Giving away an out is defensible in theory in a couple of spots, but execution eats the whole edge and then some. And the fear driving the bunt is overstated: swinging away from that same spot (8,630 times this season) produced a double play just 2.0% of the time.

The bunt base hit is the real opening

Bunting for a hit and bunting to give up an out are different strategies that share a word. Separate them and the picture inverts.

That is the shape of a real inefficiency: defenses positioned for a league that swings for extra bases concede the short game. The honest caveat is that this is a conditional edge. Hitters bunt for a hit when they are fast and the corners are deep, so the .597 reflects good spot-picking as much as the tactic. It would not survive being used indiscriminately. But at 7,822 attempts league-wide against 6,449 sacrifices, the tactic is being under-used by teams that already know how to do it.

What actually produces runs

Each coefficient is opponent-adjusted runs per plate appearance per standard deviation of that axis, with the other five held constant. 307 teams, R² 0.75.

Power outweighs everything else by a factor of four, so the crowd is not wrong to chase it. The interesting rows are the last two: holding everything else equal, teams that bunt more score more, and teams that run more score less. Both are small, both are significant, and they point in opposite directions from how the two tactics usually get bundled together as "small ball."

And power has not saturated

Opponent-adjusted runs per PA by power quintile. If the strategy were crowded to exhaustion, the last step would flatten.

The gain from the fourth fifth to the top fifth is the largest of the four steps. There is no diminishing return visible yet, which is the clearest evidence against the market-correction thesis: the standard is standard because it still works.

Steals need more than teams think

Break-even success rate implied by this season's run expectancy, by situation.

Taking second with nobody out needs 71%. Taking third needs up to 86%, because a runner on second is already in scoring position and there is much less to gain. Most of the league runs at rates that lose runs on net, which is why the running coefficient above is negative.

Where the contrarian thesis fails

Two tests that could have supported it, and did not.

Being unusual does not pay by itself. Correlation between a team's distance from the league-average identity and the runs it scores above what its component rates predict: -0.16. If anything, unusual teams slightly underperform.

Small ball does not travel better against good defense. Facing top-60 defenses, small-ball offenses lose 0.083 of wOBA while power offenses lose 0.076. The gap runs the wrong way for the theory: elite defense suppresses the short game slightly more, not less.

39 of the top 50 teams sit in the top quartile of power; only 13 sit in the top quartile of small ball. The field is genuinely crowded. It is just not crowded into a mistake.

Gradient boosting beat the lookup table, barely

A boosted classifier on inning, half, margin, base state and outs scores 0.4425 log loss against 0.4498 for the lookup table, a 1.6% improvement. Two corrections were needed to get an honest answer: the first attempt trained on features a broken join had silently dropped, and the second graded the stored table in-sample against an out-of-sample model. The gain is real but modest, and it concentrates late.

The clustering found no clean archetypes

Replacing hand-set style thresholds with a Gaussian mixture was supposed to discover natural offensive types. It did not. By BIC the best fit is two clusters, one holding 90% of teams. Offensive style in D1 softball is a continuum, not a set of camps, so the identity labels above describe where a team sits on those axes rather than membership in a school of thought.