Advertisements
Leave a comment

Digging Deep – Buy Low hitters the rest of the way

Digging DeepWith about four months of the season in the books and the trade deadline in the rearview mirror, we’re hitting the final home stretch in fantasy leagues.  Let’s try to help find some ‘hot rods’ for you (and also hopefully identify some ‘lemons’ for you to avoid or trade).

To get a robust sample size, I took the StatCast data for hitters since June 1st (about 2 months) and downloaded their xwOBA and wOBA. As a reminder, wOBA is essentially the hitter’s overall production (and correlates highly with their dollars earned…but that’s a whole other article). xwOBA is what their StatCast data (exit velocity, launch angle) would have expected their production to be. Because wOBA is what players have done (and, as mentioned, is essentially “how good they’ve been”), most fantasy owners have a good idea of who these guys are. The inefficiency we can potentially exploit is (a) who actually should have been good and (b) who has been recently good (which might be lost in the overall season numbers). The reason we go back 60 days is that someone whose recent production is either better or worse than their overall production may not be generally known to the fantasy owner community at large. This is the edge we want.

So let’s do some quick back-testing first to attempt to show that wOBA (and/or xwOBA) is predictive, or at the very least sticky. In other words, let’s check if someone’s wOBA (or xwOBA) over a two month period correlates to (and/or has any ability to predict) their wOBA (or xwOBA) over the next two month period. After all, we are trying to predict wOBA over the final two months of the season.

So I took the wOBA and xwOBA data for all players with 100 PA from the start of the season to May 31 (aka “Apri-May”), and the wOBA and xwOBA data for all players with 100 PA from June 1 to July 30 (aka “June-July”).

When comparing the expected wOBA for April-May, it weakly correlates with expected wOBA for June-July (R2 of 0.30). What this means is that xwOBA can be generally considered to be (a) a representation of a player’s true offensive ability because (b) it is (reasonably) sticky. In other words, a .400 xwOBA player for 60 days will (generally) continue to be a .400ish xwOBA player over the next 60 days.

Dylan July 31

Interestingly, a player’s two-month (April-May) expected wOBA is a better predictor (R2 of 0.18) than their actual April-May wOBA (R2 of 0.12) of his next two-month (Jun-Jul) wOBA. In other words, it seems that in this case, expected stats are actually somewhat more predictive than actual stats. But as I said, only weakly predictive. But, remember, this is in aggregate. If a player has truly changed (broken out, changed swing path, a new approach, become injured, etc.) expected statistics would signal this to us much more quickly than using projections (which are slower to react to real changes – due to their reliance on historical – that is, old – performance. And, we as fantasy owners are looking for any incremental edge that we can use

So what have we concluded? Well, two-month xwOBA seems to be a so-so proxy for a player’s “true” ability. And also that actual performance (wOBA) has quite a lot of variance. Well, how can we use this?

Let’s look at whose xwOBA for the last 60 days suggests that they are good bets (and bad bets) as hitters for the final two months. Even better, let’s find the hitters whose xwOBA is significantly better than their wOBA…as these are the players who owners may have soured on – but which we think are only randomly underperforming due to noisy variance. (Also, whose performance may be more smoke and mirrors than anything?)

Top xwOBA, last 60 days:

Mike Trout 0.464
Nelson Cruz 0.438
Josh Donaldson 0.421
Christian Yelich 0.421
Jorge Soler 0.414
Evan Longoria 0.410
Anthony Rendon 0.409
Mookie Betts 0.407
Pete Alonso 0.406
Max Muncy 0.400

Wow, Mike Trout. But also, look at Josh Donaldson, Jorge Soler, and…Max Muncy.

Lowest xwOBA, last 60 days (of only the players who are likely rostered in leagues)

Adalberto Mondesi 0.223
Andrelton Simmons 0.224
Billy Hamilton 0.245
Niko Goodrum 0.247
Dee Gordon 0.247
Michael Chavis 0.256
Cesar Hernandez 0.264
Adam Jones 0.274
Elvis Andrus 0.275
Jean Segura 0.278

Ouch, Adalberto Mondesi…and I’m not just talking about his sub-luxated shoulder. We expected regression, but this is ridiculous. A bunch of other shortstops surprisingly show up on this list too: Andrelton Simmons, Elvis Andrus, and Jean Segura.

And poor Niko Goodrum…he was identified earlier in the season as a potential sleeper because of his impressive hard-hit rate and his xwOBA of 0.348. But he has since fallen on hard times. You can’t hit on all of your lottery tickets.

OK, so that’s out of the way. Let’s look at the hitters with the biggest differences between their expected statistics and their actual statistics.

The biggest positive differences between xwOBA and wOBA (of players with positive xwOBA):

i.e. Players to target:

Name xwOBA wOBA Diff
C.J. Cron 0.363 0.308 +0.055
Lorenzo Cain 0.342 0.293 +0.049
Wil Myers 0.346 0.298 +0.048
Dan Vogelbach 0.386 0.352 +0.034
Jason Kipnis 0.341 0.308 +0.033
Brandon Belt 0.340 0.312 +0.028

A couple of interesting names. Lorenzo Cain and Wil Myers have both had well-documented struggles this year. The underlying statistics tell a slightly more optimistic story as they are both hitting the ball in an above-average way – unfortunately, the results aren’t coming in for them. In Myers’ case, San Diego might be looking at the same numbers and expect a return to form – hence, informing their decision to ship Franmil Reyes to Cleveland and open up an outfield spot.

Dan Vogelbach has quietly been putting up big offensive numbers all season. The underlying profile suggests it’s real. Always bet on the athlete 😉

 The biggest negative differences between xwOBA and wOBA (of players with positive xwOBA):

i.e. Players to avoid:

Name xwOBA wOBA Diff
Danny Santana 0.338 0.429 -0.091
Daniel Murphy 0.294 0.376 -0.082
Eric Thames 0.305 0.387 -0.082
Fernando Tatis Jr. 0.347 0.429 -0.082
Yuli Gurriel 0.343 0.416 -0.073
Kris Bryant 0.322 0.389 -0.067
Andrew Benintendi 0.310 0.376 -0.066

Danny Santana and Yuli Gurriel have been some of the hottest names right now. As you would expect, regression is coming and they “should” be slightly above league-average hitters. This is how most of us value them – and the numbers support our suspicions. They have likely not become all-stars (but they have it in them).

Fernando Tatis Jr. is an interesting name to see here as he has been out-performing expectations all season. In fact, in the April-to-May interval, his expected wOBA was 0.326 while actually posting a wOBA of 0.382 (difference of -0.056). I’m at a loss as to what this means – but what I do know that even though he is “over-performing”, his expected performance is still of an above-average hitter. Don’t sell, just temper expectations.

Daniel Murphy, Kris Bryant, and Andrew Benintendi are all players who earlier in the season had some struggles that frustrated fantasy owners. So far in June and July, it may seem like they have turned it around (with wOBA’s all around 0.380 area or ~30% better than average). Unfortunately, all three of them actually have expected wOBA’s below average. I hate to say it – but don’t be surprised if they stumble in the final months.

 

%d bloggers like this: