How good are they? I honestly and truly hate to be "that guy", but we need to have a serious talk. Regrettably, we might not have much reason to expect them to much different than a .500 team going forward, perhaps a nudge better depending on their strength in close games (high-leverage situations) and how sequencing has affected them (see my previous post about linear weights). For now, let's just focus on Pythagorean win expectations.
Currently the Mariners are 36-22, and Baseball Reference's handy Pythag field tells us they "would be" 31-27, but for a "luck" factor of +5 games. That +5 "luck" factor is #1 in MLB, with MIL at +4 at #2.
We can improve upon this "classic" style Pythagorean estimator (exponent = 2) by fitting the exponent to the 2018 MLB season data (exponent = 1.598). This decreases our RMSE from .0467 to .0425. (For more details see this post from a couple years ago and this github repo for the code.) Here is a plot of our residuals:
The punch line is the same: the Mariners are the luckiest team in MLB and HOU is the unluckiest. The Mariners' winning percentage has been boosted by ~ +0.086 due to their lucky distribution of runs, while HOU's winning percentage has been hurt by ~ -0.090. This is why Fangraphs has SEA with only a 2.6% chance to win AL West, even though they are in 1st place.
To be clear, here are some things I am not saying:
- The Mariners are bad. (They appear to be slightly better than average.)
- The Mariners will miss playoffs. (It appears to be a 50/50 bet right now due to +3.5 game lead for 2nd Wild Card spot.)
- The Mariners' luck so far in 2018 is a huge outlier. (It's only about 2 st dev's above mean.)
- The Pythagorean model is perfect. (No model is perfect. In fact a linear model has a lower RMSE so far in 2018: 0.0399. But that also tells the same story: SEA luckiest; HOU unluckiest.)
I genuinely hope I am wrong.
No comments:
Post a Comment