Post by ericmvan on Apr 6, 2018 14:11:59 GMT -5
(Hasn't been a Hanley thread since he was moved to 1B!)
First, a table of Hanley's wRC+ each year through the game that came closest to his current total of 30 PA, together with his final wRC+:
(No entry for 2013, when he was hurt early and was eased into action later.)
The first thing you notice: no average starts! Seven great ones, one very good one (this year) which might be revealed as a great one if I adjusted each number for what his teammates had done (I'm not that crazy), four terrible ones.
The next thing you notice: the four terrible starts have three of his four worst seasons.
We know what happened in 2015: he wrecked his shoulder playing an unfamiliar position and never hit again. He had a 150 wRC+ in 103 PA when he got hurt and 71 the rest of the way.
That leaves 2014, when he started lousy and ended up great, as a seeming outlier and puzzle. But it turns out not to be. The Dodgers opened the season in Australia that year, on March 22 and 23, and resumed more than a week later (on the 30th) in San Diego. He went 1/9 down under and 0/8 in his first 2 games against the Padres, then starting raking. That 71 wRC+ in his first 7 games of the year was -20 in his first 4 and 190 (4/11, 2 BB, 2 2B) in the next 3.
Even with the two outliers you have p = .13 for the correlation between start and the entire season. That's a strong enough trend to warrant eliminating them, although it's pretty obvious that you'll get a statistically significant correlation when you do. And you do, p = .01.
But that's not the best test for the difference. If you look at just the remaining good starts or bad starts separately, there's no correlation at all between the start and the season (p = .86, .95 respectively). It's simply that when he starts the season on a normal schedule, the first 30 PA are either terrible or excellent, and if he doesn't play with a ruined body, that tells you what the whole season will be like, roughly.
He has an average 135 wRC+ (raw, unweighted by season PA) in the six good-start seasons and a 97 in the three bad ones. How strong is this relationship? If you assume equal year-to-year variance regardless of start, the odds against the pattern being this strong in a random simulation are 1314 to 1. But if you think he has more genuine variance in a good season than a bad one (which may well all be years he fought nagging injuries from ST on), it's 2733 to 1.
So I'm very confident in saying that, barring a really severe injury that he plays through (which isn't going to happen with Moreland as backup), he's going to have a strong year. Whereas even after the strong 7 games, his Rest-of-Season projections are 104 ZiPS and 114 Steamer.
Now, the last time I had a similar idea and discovered it seemed to work, it was David Ortiz's start to 2016. That put a seemingly crazy number on his projected full season (and after his first couple of games!) -- which turned out to be low! This isn't as quick or precise, but I'm equally sure it's real.
This is the real Hanley, and the real Hanley can really hit.
First, a table of Hanley's wRC+ each year through the game that came closest to his current total of 30 PA, together with his final wRC+:
Year PA Start All
2008 30 240 144
2006 32 184 116
2010 28 171 127
2015 32 166 90
2009 31 162 149
2016 32 154 128
2007 32 151 144
2018 30 128 ?
2012 29 76 107
2011 29 72 97
2014 30 71 136
2017 32 67 93
(No entry for 2013, when he was hurt early and was eased into action later.)
The first thing you notice: no average starts! Seven great ones, one very good one (this year) which might be revealed as a great one if I adjusted each number for what his teammates had done (I'm not that crazy), four terrible ones.
The next thing you notice: the four terrible starts have three of his four worst seasons.
We know what happened in 2015: he wrecked his shoulder playing an unfamiliar position and never hit again. He had a 150 wRC+ in 103 PA when he got hurt and 71 the rest of the way.
That leaves 2014, when he started lousy and ended up great, as a seeming outlier and puzzle. But it turns out not to be. The Dodgers opened the season in Australia that year, on March 22 and 23, and resumed more than a week later (on the 30th) in San Diego. He went 1/9 down under and 0/8 in his first 2 games against the Padres, then starting raking. That 71 wRC+ in his first 7 games of the year was -20 in his first 4 and 190 (4/11, 2 BB, 2 2B) in the next 3.
Even with the two outliers you have p = .13 for the correlation between start and the entire season. That's a strong enough trend to warrant eliminating them, although it's pretty obvious that you'll get a statistically significant correlation when you do. And you do, p = .01.
But that's not the best test for the difference. If you look at just the remaining good starts or bad starts separately, there's no correlation at all between the start and the season (p = .86, .95 respectively). It's simply that when he starts the season on a normal schedule, the first 30 PA are either terrible or excellent, and if he doesn't play with a ruined body, that tells you what the whole season will be like, roughly.
He has an average 135 wRC+ (raw, unweighted by season PA) in the six good-start seasons and a 97 in the three bad ones. How strong is this relationship? If you assume equal year-to-year variance regardless of start, the odds against the pattern being this strong in a random simulation are 1314 to 1. But if you think he has more genuine variance in a good season than a bad one (which may well all be years he fought nagging injuries from ST on), it's 2733 to 1.
So I'm very confident in saying that, barring a really severe injury that he plays through (which isn't going to happen with Moreland as backup), he's going to have a strong year. Whereas even after the strong 7 games, his Rest-of-Season projections are 104 ZiPS and 114 Steamer.
Now, the last time I had a similar idea and discovered it seemed to work, it was David Ortiz's start to 2016. That put a seemingly crazy number on his projected full season (and after his first couple of games!) -- which turned out to be low! This isn't as quick or precise, but I'm equally sure it's real.
This is the real Hanley, and the real Hanley can really hit.