What Can Advanced Metrics Tell Us About Big Ten Softball?
By Ari Levin
Baseball is famous for sabermetrics – advanced calculations that combine statistics and use multiple parts of the box score to analyze the value of players. They’re far more common in MLB than in college baseball, both because of the availability of data and because of the closed circle of competition between teams and aren’t mainstream in softball.
During the season Noah Coffman and I worked to apply these stats to college softball. This process required some important adjustments. Games are seven innings, singles are less likely to advance runners, and games are higher scoring. Moreover, there’s far more variance both with teams and in individual players. We may not have any Statcast or play-by-play data, but we did our best to use everything available.
Our analysis only included Big Ten games. We wanted to normalize the opponents and schedule, so we used what is essentially a closed system where theoretically everyone has roughly the same strength of opponents.
Additionally, our calculations use average performance as a baseline instead of replacement level. So, you’ll see Big Ten Softball Wins Above Average instead of Wins Above Replacement. Replacement level is a funny concept for college sports, and it is probably different by program. We did include a rough WAR calculation but aren’t using it as the topline statistic.
I won’t bore you with the messy calculation methods and processes. Contact me if you really want to know those details.
We’ve uploaded our entire spreadsheet containing every Big Ten player’s statistics, and brief explanations for the measurements. We’re hoping that posting these stats now will provide some insights on the sport that you can’t find anywhere else.
Here are four of my biggest takeaways from the analysis:
The top pitchers are worth more than the top hitters – Amber Fiser, then Danielle Williams lead the pack
No hitter was worth more than 2.4 wins above average, but three pitchers were. And the top two were way above any hitter. Amber Fiser of Minnesota and Danielle Williams of Northwestern were the top two pitchers in nearly every category, and usually in that order. Wins above average is no exception, as they are respectively worth 6.7 and 4.6 wins. (Or 4.9 and 3.4 depending on the calculation; we’re not sure which method is more accurate.)
This finding makes sense to us. The best softball pitchers dominate not just a game but an entire season. Both those pitchers started nearly 70% of their team’s games. But the difference between pitchers and hitters’ value is only at the very top. Six (or seven) pitchers contributed at least 1.0 WAA, compared to 12 hitters.
We also find something very interesting with luck. The traditional “luck” statistics for pitchers are batting average allowed on balls in play and left on base percentage. But these statistics in softball are highly correlated with performance. Either softball pitching is largely determined purely by luck, or, more likely, the pitchers are very good at controlling these statistics.
It’s interesting to analyze the difference between Williams and Fiser. In most places, Fiser comes out well ahead. Her FIP – fielding-independent pitching, estimating ERA – is 0.21 compared to 1.23 for Williams, and she’s nearly two wins ahead. But Williams is close in several categories, with a slightly better K% and a better BB%, and she’s equal in total (earned plus unearned) runs allowed – 21, in 98 innings for Fiser and 96 for Williams.
The difference is in home runs. One of the most impressive statistics we found is that the Big Ten hit 291 home runs in conference play, and the pitcher with the most innings allowed none of them. But Williams and Fiser both allowed 21 extra base hits. Fiser allowed more doubles while Williams let up eight home runs. There’s skill in that, but it’s likely some luck. If those extra base hits were distributed the same way, Williams comes out slightly ahead.
But Williams also may have gotten a little lucky. Her opponents’ batting average on balls in play of .243 leads all qualified pitchers. The numbers also suggest that she was able to limit runs on the clustering of hits, which may not continue.
Lucky for us, we get another year to watch both of them. We’ll find out if the regression hits harder on Fiser’s home runs or Williams’ balls in play.
Northwestern was “lucky” (or controlled their own luck)
If you add the total Wins Above Average for every player on a team, that should give roughly the number of wins above .500 for each team. It’s a fairly good estimate, but Northwestern outperformed statistical expectations significantly. (Rutgers actually outperformed by even more, but WAA is better at predicting their negative run differential.)
The sum suggests Northwestern should have won 18 conference games and been 4-6 wins behind Michigan and Minnesota instead of right in the middle of them at 21 wins. That wasn’t from game-by-game luck; Northwestern’s run differential was good enough to be near where they were and doesn’t account for the 4-6 game difference. The Wildcats scored more runs and allowed fewer than expected.
Part of this difference is luck, but part of it we think comes from what we can’t measure. The metrics love Danielle Williams (and for good reason) but not necessarily any one of Northwestern’s hitters. Which was true throughout the season by any measure; none of the hitters really stood out on their own, but a top-to-bottom consistent lineup made the difference.
On the offensive side, some of the extra production probably comes from timely hitting with runners on base. That may not be repeatable going forward. Success may have also come from base running, not just stolen bases. Northwestern had a lineup almost entirely full of great runners, and aggressive coaching likely added extra runs.
But the biggest difference is the defense. For team batting average on balls in play allowed, Northwestern is on top and it’s not even close. The Wildcats are 34 points below the next highest team and 85 points better than league average.
This could be a factor of good luck or pitchers successfully getting weak contact. But the fact that all of Northwestern’s pitchers perform so well (Danielle Williams – .243, Kenna Wilkey – .244, Kaley Winegarner – .176, Morgan Newport – .100) suggests that it’s more than just luck. Northwestern’s fielding was great at every position and routinely turned hits to outs. That may be the biggest difference between the team the metrics predict and the Super Regional team they turned out to be.
For hitters, Lilli Piper is a clear number one; Kayla Konwent is a clear number two.
Wisconsin’s Konwent won Big Ten Player of the Year and was deserving of it. But in just Big Ten games, Ohio State’s Piper was the better player after overcoming a slow-ish start to the season.
Both traditional and our new stats show the same thing. Piper hit .574 (!) with seven home runs and struck out one (!!) time. Konwent was at .500 with six home runs and added 10 doubles. These two players were incredible and far better than anyone else in the Big Ten. But Piper comes out slightly ahead.
Piper’s weighted runs created plus (wRC+) was 296. Average is 100. The scale of wRC+ means that Piper was 196% better than average in the Big Ten. That’s how she contributed nearly 2.4 wins in just 68 at bats. Piper’s weighted on base average (wOBA) was .687, slightly better than her OBP of .681; she was phenomenal at getting on base yet still raked extra base hits. Konwent, meanwhile, had a wOBA of .642 and a wRC+ of 268. She was “only” 168% better than average.
Minnesota’s lineup is stacked – and the rotation is, too
By wOBA or wRC+, after Piper and Konwent, Minnesota takes spots three, four and five. MaKenna Partain, Hope Brandner and Natalie DenHartog all mashed. The trio combined to be worth 4.7 wins above average. That’s a junior, a sophomore, and a freshman by the way. Minnesota’s team was so good that their total WAA suggests they would win 23.4 conference games – pretty impressive given that they played 22.
The pitching staff was even better. I’ve already talked about Fiser (also a junior) and her dominance. But Sydney Smith was Minnesota’s second pitcher and was one of the best in the conference. Smith had a 2.55 FIP and a BB% as good as Fiser’s. She was worth more than a win above average. Smith will be gone, as will Maddie Houlihan, but nearly everyone else likely returns to a team that looks to defend their spot in the Women’s College World Series and perhaps go even further.
Nothing the metrics found was shocking to us, which is good. We think we’re on the right track. Obviously we’d love to find ways to measure things like fielding and baserunning, or at least get fly ball rates (HR/FB rates for Williams and Fiser would be an interesting comparison) but we’ve done what we can for now. All of these numbers are easily updatable, so we should have the metrics again next year, if not more. Hopefully this article gives some enlightenment into the details behind Big Ten softball.