DFS Strategy: Luck and Weighted on Base Average

Weighted on base average, or wOBA as it's commonly abbreviated has become the catch-all metric for DFS analysis. Whether the subject is a hitter or pitcher, wOBA has become a go-to stat and is beginning to seep over into seasonal fantasy evaluation as well.

For those unaware, wOBA is a souped up version of standard on base percentage where the components are weighted in accordance to their run production potential. Developed by Tom Tango, it was first introduced in, you guessed it, The Book: Playing the Percentages in Baseball. As an aside, it's remarkably prescient how integral a seven-year old publication has become to DFS. The theory is due to the points-based nature of DFS scoring, ranking by wOBA better matches DFS potential than the other rate stats.

Using wOBA is a great down-and-dirty means of looking at players for DFS but it isn't without shortcomings. For instance, it doesn't account for the stolen base potential of a player so when evaluating the likes of Dee Gordon or Billy Hamilton, some consideration has to be given to their added points scoring potential. The other issue is wOBA is not park corrected, As such, the same wOBA from a San Francisco Giants' hitter is more impressive than if it came from a member of the Colorado Rockies, Similarly, the same wOBA against from a pitcher on the Boston Red Sox is more impressive than if it belonged to a Pittsburgh Pirates hurler. The means to account for this is to instead use wRC+ as your measuring stick. This is another Tango machination and is essentially a park-corrected wOBA.

Like most other DFS analysts, wOBA is often referenced in my writing and as a player, it is part of my toolbox. I'm aware of the pitfalls of the stat and can account for it when making player recommendations and choosing my own lineup.

While thinking about a couple of improvements to my daily player projection engine, I had a rather disturbing thought. Using luck, specifically a lucky or unlucky batting average on balls in play (BABIP) has become commonplace in all areas of fantasy analysis. How come we don't ever say a wOBA is lucky or unlucky and due to regress? We treat it as a concrete snapshot of the player, with no deference heeded to the fact a crucial aspect of wOBA -- hits -- has a substantial element of luck associated with it.

We start a pitcher because the wOBA of the opposition's lineup is below average. What if it is being torpedoed by low BABIP?

We look at a hitter's splits and note so far this season, a hitter's wOBA versus southpaws is exceptional. Aside from the sample size flaw of that observation, does anyone glance at the BABIP in that scenario to see if the wOBA is buoyed by a fortunate BABIP?

In order for this concern to be valid there needs to be some correlation between wOBA and BABIP. It may seem intuitive that there is but remember, home runs are absent from the BABIP calculation whereas they're very much a part of wOBA. What follows is data from 2012-2014 showing the correlation of BABIP and wOBA in terms of teams' hitting and pitching statistics.

Sure enough, wOBA and BABIP have a decent level of correlation. Further, Assuming 2015 isn't an aberration, the fact the correlation is less so far the season for both hitting and pitching suggests instability and as such, using wOBA blindly may mean we're not basing the decision strictly on skills and are instead counting some happenstance as skill.

There's a chance the lowered correlation emanates more from variance in BABIP than from wOBA. To investigate this, I looked at the standard deviation for each around their respective league averages. The notion is if the BABIP and wOBA were still unsettled, the lucky teams would have a high BABIP and wOBA while they'd be low for the snake bitten squads. This would be reflected in a higher standard deviation than season-long numbers, where each had more time to regress towards the mean, reducing the league standard deviation.

 The current standard deviations are indeed bigger, suggesting some regression is still in the offing. However, there is less work to be done with wOBA than BABIP, especially with respect to pitchers. In fact, pitcher wOBA may already be stable though if we average the standard deviations from 2012 -2014 we're still a smidgen above.

Based on this, my conclusions are as follows:

- When using a hitter's wOBA, some consideration has to be paid to the luck element involved, as demonstrated by a lucky or unlucky BABIP

- It's safer to look at the wOBA a pitcher has given up, though the BABIP against should be noted to make sure he's not an outlier

We're basically one-quarter of the way through the season. A fair question to ask is at what point do the season-to-date standard deviations match the year-ending marks. Sounds like a great topic for next time, don't ya think?