Tuesday, September 23, 2014

Revisiting FanGraphs WAR

Back in June, I posted something about the problems that I saw with team WAR about a third of the way through the baseball season. In that post, I said I'd revisit the WAR issue, seeing as how there was a relatively small sample size. Now that the season is really almost over, let's revisit that.

At the time of the last post, the line of best fit of the WAR-adjusted wins and real wins had a slop of about 0.4, quite far from the "ideal" 1. This is what that graph looks like now:


As we can see, the line of best fit has a slop of over 0.85, which is essentially as close to perfect as one can expect. This makes me feel better. However, there are some major differences between the team WAR calculations in June and those today.

This time, the highest FanGraphs team WAR belongs to the Nationals at 2.0, and the lowest is the Cubs at 0.6. The essential definition, or perhaps goal, of WAR is to measure a player's value above a hypothetical "replacement" player. That replacement-level is defined as the level of players that, if an entire team consisted of such players, its winning percentage would be .294. So if we are to take that definition to heart and apply it to team WAR as it stands now on FanGraphs, no team would have more than 50 wins at this point by WAR. Obviously, team WAR is not measuring this, but I'm not sure what it's measuring. It's been adjusted somehow, but I'm not sure how. I've adjusted the WAR win totals to more closely reflect real wins so that they could be compared more easily apples-to-apples.

So as we can see, there are very few serious outliers at this point in the season, and the line of best fit more or less accurately reflects a pretty good approximation of wins as seen through the lens of WAR. But that team WAR certainly is not just a stat that adds up all the individual players' WAR throughout the season, so I'm not sure what it is exactly. Player WAR is still not clarified by this examination.

Next season, I'll be keeping an eye on this and digging deeper into these issues. For now, let's throw all these out the window and enjoy the randomness of 1, 5, and 7-game series.

No comments:

Post a Comment