On the heels of evaluating players based on how many wins they can provide to their team, let’s look at how clutch players are. Do they shine in the big moments, or just generally perform well yet fail to make an impact when it truly matters?1This post will conclude the first 10 days of Blogmas, which I decided from the beginning would be dedicated to hitting. The next 10 days will be dedicated to pitching statistics.
Win Expectancy measures how likely it is for a team to win given the exact situation of the game at any point. I’ll leave it to FanGraphs to say something about its calculation (emphasis is mine).
Calculating win expectancy from scratch is one of those things that’s extremely easy conceptually and very challenging in practice. All you would have to do to find the win expectancy of a situation would be to identify all similar situations in the last ten years or so (the sample you choose depends on the run environment) and then find the winning percentage of teams who found themselves in those situations.
So, let’s leave the precise steps for calculating Win Expectancy to those who have an interest in doing so. Today, we’ll focus on using the fruits of their labor to learn something interesting.
First: Not all home runs are created equal. A solo home run in the 7th inning of a 12-0 blowout is not nearly as important as a solo home run in the top of the 9th of a tied game. Those two results change the win expectancy in very different ways, and it is this change in which we are interested.
Similar to the run-scoring matrix we learned about for calculating wOBA, Win Probability Added (WPA) is as simple as taking your huge chart of Win Expectancy, and comparing the value from the current state to the state after a batter completes their plate appearance. WPA is a counting statistic, so a player’s WPA changes every time they hit, going up and down depending on their effect on the game.
Numerically, WPA is a change in percentage. So, a +1 WPA is worth a 100% increase in win percentage, or 1 additional win.2This isn’t quite the same as 1 WAR. We’re solely considering performance within a situational frame of reference. We don’t even care what kind of result it was, only its impact.
Intuitively, WPA changes the most when a batter appears in a key moment — their team is down by several runs late in the game, so the chances of winning are slim — and provides a hit or other play that swings Win Expectancy into their team’s favor. In this sense, WPA is a piece of flavor text around a player’s season and career. We certainly don’t assume that WPA is likely to predict their ability to hit in clutch moments in the future, and people generally agree that overall performance metrics are better to use. So, we use WPA to tell good stories about a player, and as a way to look deeper into the situations that led high (or low) WPA players to have their particular value.
Again, I’ll point you toward Foolish Baseball who made a great video exploring clutch hitting, where he discussed Win Expectancy and WPA.
We’ll finish off our first ten days of hitting with an interesting player I discovered while trying to find cool statistics to share.
Troy O’Leary
In 1996, Troy O’Leary played outfield for the Boston Red Sox. He appeared in 149 games3A reminder there are 162 games, so he was a regular. and had some pretty underwhelming numbers. Armed with what we’ve learned thus far, we can make a good profile of him as a hitter.
- Slash line of .260/.327/.753, and a below-average 88 OPS+.
- Had an 87 wRC+ and “accumulated” -3.2 wRAA.
- Ended up with 0.6 oWAR, but a total -0.9 bWAR.4As a reminder, that’s WAR according to Baseball Reference. FanGraphs has him at -1.3 fWAR.
A pretty poor season in what was a slightly below-average career. Yet, something odd happened in 1996: Troy had a WPA of 3.1. Somehow, he hit just well enough in exactly the right situations to add just over 3 win’s worth of probability for his team.
I dove into the game logs for Troy’s 1996 season, and started looking for his biggest impact moments. Sorting by WPA, I saw:
- Three games with a WPA between +0.3 and +0.4.
- Four games with a WPA between +0.4 and +0.5.
- One game with a WPA of +0.823.
All but one of these games was decided by 1 run, and naturally, Boston won all of them.5There was an 8-6 win over Toronto where his WPA was +0.378. These are exactly the sort of tense moments where WPA gains are made. Let’s look at that game where he contributed 82% of the Win Expectancy.
It was Friday August 2nd at Fenway Park in Boston. The game began at 7:08PM, so the sun wouldn’t be properly gone for a at least an hour and a half. Minnesota Twins put Rick Aguilera6A key to their 1991 World Series win. on the mound against Tom Gordon.
In the bottom of the 1st, the Red Sox came out swinging and went up 4-0 on a lead-off home run, followed by a two-out rally. In the top of the 3rd, the Twins put a couple on the board. Back and forth the teams went, until in the top of the 6th the Twins tied it 5-5.
After giving up a home run and two singles in the bottom of the 6th, Aguilera was replaced by Mike Trombley, who retired the next three hitters, limiting the damage to a single run on the inning.
In the bottom of the 7th, the Red Sox added an insurance run, going up 7-5. But, in the top of the 8th the Twins faced reliever Stan Belinda for his second inning of work. After an error and a strikeout, he gave up a single and was replaced by Heathcliff Slocumb.7Baseball names are just *the best.
Slocumb let things fall apart. By the time he was replaced by Mark Brandenburg following several more singles, a walk, and a wild pitch, the Twins were ahead 9-7. Mark Brandenburg got a groundout that allowed another run to score, then induced another groundout to end the inning with the Twins up 10-7.
The Red Sox went down quietly in the bottom of the 8th, including Troy himself grounding out to first base. Brandenburg continued a solid outing in the top of the 9th, hitting the first batter but retiring the next three.
So, to the bottom of the 9th with the Red Sox down by 3 runs. The closer for the Twins, Dan Naulty, got two quick outs: a strikeout looking, and a groundout to first base. Naulty walked the next batter. As they say, walks will haunt.
Reggie Jefferson hit a home run on a 1-0 count, narrowing the game to 1 run.8This is a great example of WPA dynamics. With 2 outs in the bottom of the 9th, getting to 1 run away only changed the win expectancy by 4%. That home run broke the seal; it was followed by a single and a walk to get runners on 1st and 2nd base.
Up comes Troy O’Leary. The chance that the Red Sox win with 2 outs in the bottom of the 9th, down by 1 run, with runners on 1st and 2nd, is only 19%. On a 1-0 count he hits a triple between right and center field, driving in the two runners and winning the game. From a win expectancy of 19% to 100%, he gets 81% WPA and a very small role in this unimportant blog.
Continue to Day 11 – Earned Run Average
- 1This post will conclude the first 10 days of Blogmas, which I decided from the beginning would be dedicated to hitting. The next 10 days will be dedicated to pitching statistics.
- 2This isn’t quite the same as 1 WAR. We’re solely considering performance within a situational frame of reference. We don’t even care what kind of result it was, only its impact.
- 3A reminder there are 162 games, so he was a regular.
- 4As a reminder, that’s WAR according to Baseball Reference. FanGraphs has him at -1.3 fWAR.
- 5There was an 8-6 win over Toronto where his WPA was +0.378.
- 6A key to their 1991 World Series win.
- 7Baseball names are just *the best.
- 8This is a great example of WPA dynamics. With 2 outs in the bottom of the 9th, getting to 1 run away only changed the win expectancy by 4%.