Rajai Davis, Stolen Bases, and WPA

I don’t think that the reader “is left to conclude that Rajai Davis is not merely an average player, but one that’s actually harmful to the A’s chances of winning baseball games”.

Regardless of the usefulness of WPA, the post says that it is only taking in consideration offense and concludes that using this specific metric:

Rajai Davis has a -1.47 WPA over this season, which means that he’s cost the A’s almost one and a half wins below an average player offensively

A’s suck on offense. They scored almost 50 runs less than an average AL offense this year – or about 5 wins worth of offense. That means that while replacing Rajai on offense with a league average player would increase the A’s chance to win, the same can be said for most any other player not named Cust or Barton.

A’s do not have league average offensive players in the lineup for most part and much less on the bench. Therefor “replacing Rajai with a league average offensive player” is not really an option. Being below average in the league and above average on the A’s is, unfortunately, not a contradiction.

are one thing that could be a significant flaw in WPA. We know that the coliseum suppresses scoring, but I don’t know exactly how that affects win probabilities. It might be very significant, since those probabilities are very non-linear with runs scored.

It looks at a game situation and the factors are:
1. Inning (1-9, extra innings are treated as 9th)
2. Runs differential (-11 to +11, scores above are treated as 11)
3. Outs (0-2)
4. Base runners (6 combinations)
that is over 8000 different scenarios.

It then looks in the database with over 9 mil. game situations, finds all that fit the mold and says how often a team won when faced with such a situation. That’s WP (Win Probability)
After a player has finished his PA, it looks att the new situation and calculates WP again. The difference is the individual WPA ( Win Probability Added). It also adjusts the situation for the context a bit (low scoring games from 1930s or playing in Colorado and so on).

So, it values a single with two outs in a tie game more than it does a grand slam in a blowout, for example, as the difference in win probability before and after the batter did his thing is greater.

What he doesn’t do — get on base very often — is less damaging because in a low-scoring game, hardly anyone gets on base. Run environment usually doesn’t depress SB as much as hitting.

We get people on base and they never make it home. SBs help the cause of making it home. But yeah, our lineup is a craphole of hitlessness so whatever.

It isn’t a way to figure out a player’s talent, but a way to describe (fairly accurately) what they’ve done. In EC’s example, if I hit with the bases loaded and zero outs in a tie game every at bat, I’m going to have a huge WPA since my hits will add a ton and my outs won’t subtract all that much. If I’m Jack Cust, I get screwed by WPA because my bases-empty two-out walk when my team is down by 2 doesn’t produce much WPA at all.

I use it in GOGs as a way to identify the stars and goats of a particular series, and I think it is quite effective at that. I also think that it’s a valuable way to look at base runners, as they can choose to steal or not steal therefore making the WPA (in my mind) more important than a context-neutral measure. If Rajai’s actual effect were to cost us wins by attempting to steal, he should stop even if his context-neutral ability were a positive. The same goes for sac bunts.

that offers more than a “what have you done for me lately” evaluation, is the point about Rajai being able to steal in more “clutch” situations than some other basestealers. If you pick the right spot (i.e. late innings of a tie game), then a stolen base can have a large impact. Of course, the other team knows this, so you can expect your success rate to go down.

But if Rajai is such a good basestealer that he can beat the pick-off moves and pitch outs and still get himself to second at a good rate, then that is a real advantage — and possibly the first ever use of the WPA clutch stat for useful player evaluation.

going to be as bad as it’s been this year, I’d probably look to bring in a real CF — either Coco or someone else.

That’s exactly the method I used, but yes only for steals of 2B and times caught stealing 2B.

But then I don’t know why wOBA would do that since each is equally valuable.

But wOBA does (a little), because that’s just how the numbers laid out. It’s possible that pitchers who hit batters are more likely to give up runs than pitchers who walk batters.

there are many movements that are not accounted for

everybody else is retired

We traded him for Robinson Cano.