WPA Thoughts on the M's through May 27

by Two Rs and Two Ls on May 28, 2009 2:13 PM PDT Comment 8 comments

I find the Mariners' bullpen confusing. The TV keeps telling me that the bullpen is a team strength, but we've seen Morrow blow three games, we've seen a lot more Denny Stark than any "good bullpen" should be showing, and Mark Lowe has been awesome at times and rather hittable at others.

I wanted to disprove the TV a little bit, so I thought about the bullpen's job. It seems to me that Priority One for any bullpen is, when given the lead, maintain the lead so that your team wins. Priority Two would be something along the lines of, well, when you're not in the lead (or tied), give up as few runs as possible so that your offense has a chance to climb back in it.

I thought what I'd do is take a look at the WPA charts for each game so far and see if the M's have been reasonably successful in games where their WPA reached above 90, 80, and 70. Unfortunately, I think this is some pretty noisy data, because you have games like May 18, when Washburn blew the game and it had nothing to do with the bullpen, or games like the 15-inning affair on May 3 where we never had the lead when we came up to bat (and the bullpen surrendered 3 runs in extras... it's hard to give the 'pen credit for a good job when they do things like that).

What I end up with isn't exactly a good proxy measure for team bullpen usefulness, but I still find it interesting.

The Mariners have won 22 games so far. By definition, their WPA reached above 70, 80, and 90 in these games.

The M's are 22-3 in games where WPA was at one point 90 or above: April 7, May 14, and May 26.

The M's are 22-9 in games where WPA reached 80 or above, including the games listed above and also April 29, May 2, May 5, May 13, May 18, and May 23.

Finally, the M's are 22-11 in games where WPA reached 70 or above, including the games already mentioned and April 16 and May 9.

Now for some math. The M's played 25 games where WPA > 90.We'd expect them to win 90% of those games. 90% of 25 is 22.5, so we're only .5 games under that. That strikes me as okay, but nothing special.

31 games where WPA > 80, and 80% of 31 is 24.8, which is 2.8 games more than actual. Looks like we've been having a little trouble here.

33 games where WPA > 70, and 70% of 33 is 23.1, just 1.1 games more than actual, so again, looks like we're okay in this area, but we don't have it on lockdown.

I'm not sure where to take this information from here, especially because the data seems really noisy, but I thought I'd at least put it out there for public consumption.

1 recs | Comment 8 comments

Story-email Email Printer Print

Comments

An interesting way to look at it.

If you’re simply interested in measuring our bullpen using WPA, you can look it up on fangraphs here. Turns our the Mariners bullpen WPA for the season is 0.57, which is good for 12th in MLB. Also interesting to note, the Mariners achieved their mark of 0.57 WPA with an accumulation of ~~14.56 WPA~~ (5th worst in MLB) and 15.13 WPA+ (2nd best in baseball). I guess I would interpret our high totals for WPA- and WPA+ relative to other teams to our bullpen’s volatility, our bullpen’s increased number of innings pitched relative to other teams (154 IP, 10th in MLB), or likely some combination of both with perhaps other factors.

by FlaskInSafeco on May 29, 2009 6:12 AM PDT reply actions 0 recs

That's supposed to read a WPA(minus) of -14.56. Damn SB Nation formatting.

by FlaskInSafeco on May 29, 2009 6:13 AM PDT up reply actions 0 recs

Win probabilities are tricky

Saying that if Win expectancy (what I think you mean by WPA) is above 70 at some point during the game, we should win 70% of those games is a bit dicey. A lot depends on when we reach 70%. If you score 5 runs in the top of the first to get there it means something very different than a 1 run lead after the top of the 9th in terms of bull pen effectiveness. I would expect that a sample of games in which a team had a 70% or greater Win Expectancy (WE) at some point in the game would not results in a .700 winning result. I think it would be less due to variations in WE.

WE will vary a lot during a game. We start every game with 50% WE and our record is under .500 but you can’t draw any conclusions from that.

What Flask says makes sense to me. Our bullpen has moments of brilliance and moments of spectacular ineptitude. This seems a natural result when you have a bunch of guys with great stuff but poor command. I’d expect this trend to continue unless we switch out some bullpen arms.

by PDXTai on May 29, 2009 10:27 AM PDT reply actions 0 recs

To your first point

that’s why there’s a lot of noise in the data. Like I said, how do you credit the bullpen with a great job in the 15-inning game when they gave up 3 runs in the 12th to almost certainly lose the game? That’s why it’s really hard to say anything at all about the bullpen based on this research.

I think I’d expect WE% to be more accurate the more extreme its value. Clearly, a poor team doesn’t actually have a 50/50 shot of winning every game, but I don’t see a reason why that poor team should have significant problems holding onto games when they have a 90% WE. On the other hand, I might expect a good team to have more games in which they reached 70 or 80%, but they may not necessarily win those games at a higher percentage. I’d probably have to do more work on that to justify the claims though.

As an extra note, I noticed that Mariners relievers have the second highest LI when they enter the game when compared to all other bullpens in baseball, a 1.34 gmLI. They’re tied with the Dodgers and are second to the Nationals. That might explain the spikes in WPA+ and WPA-.

It is a truth universally acknowledged that a zombie in possession of brains must be in want of more brains. -- P&P&Z

by Two Rs and Two Ls on May 29, 2009 1:21 PM PDT up reply actions 0 recs

That's a good point about the high average LI.

I agree that likely has something to do with the Mariners higher levels of WPA+ and WPA-.

by FlaskInSafeco on May 29, 2009 4:41 PM PDT up reply actions 0 recs

PDXTai makes a good point.

I dont know if you can actually say that a team should win 70% of games in which their highest WE meets or exceeds 70%. When the WE is exactly 70% the team should with the game 70% of the time going forward. I’m not sure if that’s the same thing as retroactively looking back at each game where it reaches 70% at any point and saying you should win 70% of those games. I suspect it’s actually not the same. If we repeated your process above for a much larger sample of games we could probably get to the bottom of this.

by FlaskInSafeco on May 29, 2009 4:39 PM PDT reply actions 0 recs