THE BOOK--Playing The Percentages In Baseball

A process that includes fielders might end up being like the responsibility of an individual play in Strat-O-Matic.  In Strat-O-Matic, the individual play result comes from the batter’s card, or the pitcher’s card, or the fielder’s portion of the pitcher’s card.  I’m not saying that’s the answer to create a process for WPA to include fielders, but it might be a starting point.  Or not.

You can only go so far as the data allows. 

If the data shows 6-3, as opposed to an in-the-dirt-throw-scooped out 6-3, then you may give all the fielding credit to the SS and none to the 1B.  Or, if the ball was hit deep into the hole that the SS had to range wide for, and make an off balance throw.  Or if the ball was hit directly to the SS who was positioned where any SS would have been positioned.

If all you see is: 6-3, then the story that you are going to describe with WPA may hide what really happened for that game, especially if the SS was involved in only 3-4 plays.

So, WPA takes an agnostic view here.  It doesn’t take any position here, because the data is too limiting.

An analogy would be that all you get from a hitter is whether he got on base or not.  You don’t know how many bases, if any, any of the runners moved.  You have no idea if a run scored.  You have no idea if there was a FC.  All you know is whether the batter reached base or not.

Given that, how helpful would WPA be?  You can just blanket infer that he got a single 70% of the time, a walk 10%, a double 5%, a HR 5%, and FC 10%.  You will estimate how many bases the runners moved, and how many scored.

Are you going to be told a good story? 

That’s what fielding WPA would be like, in terms of uncertainty of an accurate story.

A few years ago, before WPA was widely available, I recorded a series of Reds games using my WPA spreadsheet for a Reds bulletin board.  I tried my best to include the impact of fielding by apportioning percentages to pitchers and fielders, and also double-entering plays (one with an “expected” fielding impact and one with the actual fielding impact).

As Tango says, it’s a tough thing to do.  For instance, what if there are two great fielding plays on the same play?  How do you apportion that?

Still, I think it’s a worthy goal. I’m a WPA junkie and I’d like to see the impact of fielding included. Things I’d note:

- You have to watch and record the game.
- The best way to judge fielding is vs. average (as opposed to runs scored or allowed, which start at zero).  It’s unclear to me how to use that scale in a WPA structure.
- There will always be judgment involved.

Right, if someone else does all the hard work, then I’d love to see the results.

That said, I think the value would go beyond WPA as a story stat, and would actually have analytical value.

Human input to fielding plays is really what is missing.

Hi,

Studes - If you’d incorporate fielding in WPA, would you still maintain the principle that for every credit there’s a debit? The reason that I ask, is that in a bad fielding play there is no obvious hero, while in a great fielding play there’s no obvious goat.
I’d like to thank Tango for quoting my original mail and bringing it to this amazing blog.

Yehuda

The principle of WPA has established that you need to have a balance.  For every plus on one side, there’s a minus on the other.

So, even if no one “deserves” a plus, SOMEONE is going to get it.  That’s why I say WPA is about involvement, not responsibility.

Otherwise, you can construct a scenario when a team gets no pluses, and yet still wins the game.  (Imagine a perfect game destroyed by a 9th inning error.)

That’s why I keep saying you need to get out of who “deserves” the credit.  The guy was there when something good for his team happened.  He won the lottery.  Count that in his bank account.  That’s his story.

You could add “team fielding” without breaking to individual level. Homers, walks and strikeouts all accrue to the pitcher. Credit or debit for grounders, liners, flies and pop-ups would be split between the pitcher and defense.

So in the fabulous example above, the pitcher might get 810 WPA points, the batter might get 818 WPA points, and the defense would get -1,628.

When I first presented WPA eight years ago, I showed the results for pitchers based on: all PA, only the non-batted ball PA.

http://www.tangotiger.net/archives/stud0194.shtml

On a team level (i.e., the sum of all players), the WAA is exactly equal to the team’s actual
(Wins - Losses) / 2

I also present a WAA2 column that assumes that 100% of the BIP goes to the pitcher, just so that you can see what the difference is. Not much is the short answer.

So, in the chart, I show that in 1999-2002, Randy Johson was either +23 wins or +24 wins, depending how you handle the BIP.  Pedro was either +20 or +21.  Schilling was +13 or +14.

Basically, you can try it different ways, and it all depends what kind of story you want to tell, and how much effort you want to make in telling that story.

Yes, I had a “team” category for some fielding things, too.

Credits have to equal debits, but you can do a lot with debiting pitching and crediting fielding on the same play.  Consider: pitcher gives up a fly ball that is headed over the fence but Torii Hunter leaps over the fence to rob the batter.

In that case, you’d give a big negative to the pitcher for giving up the home run, then give that all back to Hunter PLUS the extra WPA for the out.  Big WPA hit goes to Hunter.

I have never figured out why no one else has ever thought of WPA above replacement. Or even just a baseline WPA Wins. I think it works out to be something like WPA+ + WPA-*.63 if I remember correctly.

No one else?

http://www.hardballtimes.com/main/statpages/wpab_form/

Nevermind then, thanks studes.

Bill/11: If you read the comment section of the link I provided, I talk about that… 8 years ago!

Thanks guys, read the link and like what I see. Has anyone ever attempted making Win Shares out of WPA and compare it to the original?

Win Shares and WPA are totally different things.  I’m not sure what you mean.

Use WPA+ and WPA- to create a stat that determines player wins above a threshold of zero.

I guess you could do that on the WPAB calculator.  Just set the replacement level to zero.  I’m not sure how legitimate that approach is, however.

Tango has explained this issue well in the past.  There is a lot of “churn” in WPA.  That is, even a .000 team will have a lot of positive WPA.  It’s just that they’ll have a lot more negative.

The amount of “churn” you want to assign to a player is not a straightforward matter.