wOBA, wRC+ and luck: Improving Fantasy Alarm Projection and Pricing Tool
Todd takes a look at luck with wOBA and wRC+ in an effort to improve the Projection and Pricing Tool
If you’re into next level analysis with DFS baseball, you’ve no doubt read up on wOBA and wRC+. If you haven’t, you should. Both are detailed on Fangraphs.
The Cliff Notes version is both are a catch-all metric that correlates very well with DFS scoring. That is, if you’re looking for one number to best estimate a player’s DFS potential, wOBA and wRC+ do the best job of capturing all the necessary components and generating a single number.
Weighted on base average, or wOBA is a souped up version of on base percentage where each component is weighted according to its contribution to run scoring. With respect to DFS, the deficiency with wOBA is it isn’t park corrected. Some of the components are inflated in hitter’s parks which will in turn result in a higher wOBA. To the layman, this isn’t an issue. After all, the stats you see on the back of a baseball card are not park adjusted. But, if a San Diego Padre and a Colorado Rockie generate the same wOBA, the one with half their games in PETCO Park is more impressive.
The Fantasy Alarm Projection and Pricing Tool doesn’t use a pitcher’s wOBA to adjust the hitter’s expectations, instead focusing on each skill individually. That is, a hitter has a certain strikeout rate and a pitcher fans batters at a certain rate. I put the two together to adjust the hitter’s expected whiff rate based on that pitcher. The same is done for hits, homers etc. Neither wOBA nor wRC+ are incorporated.
However, the tool does employ a team’s hitting wOBA (using the appropriate handedness) to adjust the expectation of the opposing pitcher. I turn the wOBA into a park neutral measure then adjust based on the venue for that game to account for the fact wOBA is not park corrected. A universal home/away factor is applied since parsing the data into four sectors (home v RHP, away v RHP, home v LHP and away v LHP) can introduce too much variance into the sample, especially versus southpaws.
As DFS analysis has evolved, there has been a shift to using weighed runs created, or wRC+, in lieu of wOBA since there is a park correction. In short, if wOBA is a souped up version of OBP, then wRC+ is wOBA on an undetectable PED.
Since the Projection and Pricing Tool has a park adjustment in the algorithm, I decided not to switch the code to wRC+ – at least not yet. I’m the first to admit the engine is a work-in-progress. I have no doubt it’s honing the expectation of hitters and pitchers based on their matchups but that doesn’t mean I can’t keep plugging away to make it better.
Instigated by a g-chat with a friend that is very good DFS player (hi Paul), I had a thought which was worth investigating. Next level analysis with traditional fantasy baseball revolves mostly around fleshing out the luck from the surface stats. Be it BABIP, HR per FB, LOB percentage or your preferred expected ERA, after using fortune, serendipity, happenstance, snake-bit and victimized, my fellow analysts and I run out of different ways to describe good and bad luck.
But there’s not as much citing of luck in DFS analysis. Both wOBA and wRC+ for hitters have components that dovetail on BABIP and HR/FB. I thought it would be interesting to see if either wOBA or wRC+ did a better job of accounting for the happenstance of a round bat meeting a round ball. The metric that did the better job of neutralizing luck should be the one used in the Projection and Pricing Tool.
The easiest way to measure luck is to correlate BABIP to wOBA and wRC+, using BABIP as a proxy for luck (I know, it isn’t that easy, more in a moment). The better the correlation, the more the metric carries over some element of chance. Ooh, there’s another synonym.
Keep in mind the closer to 1, the higher the correlation. The closer to zero, the more the relationship between BABIP and the metric is random. Below is a table correlating BABIP to wOBA and wRC+, broken down by handedness for the last three seasons plus the present one through the games of August 20
|wOBA vs. LHP||0.62||0.59||0.79||0.65|
|RC+ vs. LHP||0.57||0.51||0.77||0.69|
|wOBA vs. RHP||0.69||0.70||0.50||0.46|
|RC+ vs. RHP||0.50||0.64||0.30||0.42|
On the surface it appears wRC+ may do a better job of mitigating luck. In almost all instances, the wRC+ correlation is less than the corresponding wOBA correlation. But remember, wRC+ has a park correction and BABIP (as well as HR/FB) are affected by the venue so there’s a very good chance the lower correlation is a reflection of the park factor and the luck is equally present in both.
I apologize that I don’t have an earth-shattering conclusion based on these cursory studies. This is an area that requires some deeper thought and a lot of back-testing. My goal is for the Fantasy Alarm Projection and Pricing tool to be recognized as the industry standard. By looking at this and other means to improve the numbers at the fringes, I’m going to do just that.
Though, I’m having a serious issue accounting for busted pipes and inept ground crews.