Pittsburgh Pirates And Win Probability Added

The Pittsburgh Pirates 2018 season is now complete, with the club going 82-79 on the year.  With the time to review how the season shaped up, what better way to start off with how players contributed to wins.

This past weekend, I looked at how the Pirates hitters did with respect to their hits tying or taking the lead, a basic and broad way of how players were clutch, and with clutch being a term used looking at the past to describe a moment.  But there’s a better framework to use to describe how clutch a player was, and the premise comes from win probability added, or simply WPA.

First, to define the terms used throughout this post (all definitions will be from Fangraphs):

• Win Expectancy the percent chance a particular team will win based on the score, inning, outs, runners on base, and the run environment. These percentages are calculated using historical data
• Win Probability Added (WPA)- captures the change in Win Expectancy from one plate appearance to the next and credits or debits the player based on how much their action increased their team’s odds of winning
• Leverage Index (LI)- an attempt to quantify this pressure so we can determine if a player has been used primarily in high-leverage or low-leverage situations

WPA is a counting stat that measures what happened, so if a player has a plate appearance that increases the team’s chances of winning by 10 percent, he receives 0.1 in WPA.  Below are the Pirates non pitchers with 50 plate appearances this season:

Gregory Polanco tops the charts, accumulating in 2.77 wins added through his plate appearances.  That’s above average (2.0) but not necessarily a level of a player producing impact moments to swing the game in their plate appearances.  The trailer is Jordy Mercer, who is a free agent this offseason, costing the Pirates -1.59 wins in his plate appearances, which is poor.  The good news for him is that WPA is backwards looking and not predictive, but in 2018 his plate appearances swung the pendulum the other way.

But what about adjusting for the leverage in which a batter finds himself in?  A one run deficit with runners on-base in the bottom of the eighth would have a bigger change of win expectancy compared to a tie game with nobody on and two outs in the bottom of the second.  Here’s the same list of players once adjusted for the leverage:

Accounting for the leverage still leaves Gregory Polanco at the top of the list, adding about 1.5 wins, and Jordy Mercer is still at the bottom of the list subtracting off about 1.3 wins.  A lot of that is how the two hit in high leverage situations, with Polanco posting a .453 on-base and .581 slugging compared to Mercer’s .267 on-base and .308 slugging.  The two had 53 and 45 plate appearances in high leverage situations respectively.

In terms of relief pitching, this is how the Pirates bullpen (minimum ten innings) stacked up:

Felipe Vazquez led the bullpen in win probability added, no surprise since he came into the game in the ninth inning to shut the game down.  At the bottom of the list was Clay Holmes, whose WPA was -0.84 when he was used as a reliever.  The good news with Holmes was that he was used in the low leverage situations.  The average leverage index when Holmes entered games in relief was 0.38 (the third column and that of which this list is ordered).

The pitcher who entered games with the second highest leverage index was George Kontos, who lasted 19.7 innings and 21 games.  He was dreadful and cost the team 0.35 of a win in his short time.  One thing that the last column does look at is, on average, who was used based on the leverage situation.

The first eye-popping one Keone Kela entering the game with an average leverage index of 1.12.  This past season, Kela pitched to a 3.29 ERA, 2.97 FIP, and 31.1 percent strikeout rate.  For his career, the reliever possesses a 30.4 percent strikeout rate.  With the club insisting to have a closer and Vazquez being that, using Kela in situations of higher leverage would be a nice change in 2019.  He can help swing the win expectancy with his strikeout ability.

Kyle Crick had an average leverage index of 1.14, but after he moved to the eighth inning role in that jumped to 1.25, so he was seeing more high leverage chances.  Michael Feliz started the 2018 season with a backend role compared to Richard Rodriguez, but Rodriguez posted a 31.5 percent strikeout rate, so maybe seeing more high leverage situations next year would be nice (LI of 1.00 is average and above 2.00 is high leverage).

Going forward in 2019, to maximize the win expectancy the Pirates should use the relievers in a way to maximize it based on changes in win expectancy of this reliever compared to another on the matchups.  But even with Vazquez in the backend, having a trio of Kela, Crick, and Rodriguez could contribute to that given the strikeout rates.

Overall, the Pirate bats didn’t do much to swing the win expectancy in their favor.  Only Polanco, Starling Marte, and Josh Bell were at least average.  On the pitching side, Vazquez was distributed in good spots on average, mainly because of his use as the closer, and Kela was low, but, hopefully, that’ll change in 2019.