September 11, 2002
Best Hitter Outside Cooperstown
Just the mention of Keith Hernandez should have clued you into questions like these you are bound to receive--Hernandez #7 and no mention of Don Mattingly? Especially when considering peak value--he certainly had a higher peak than Hernandez.
There are two main things that distinguish Hernandez from Mattingly offensively:
For what it's worth, my method does not rate Mattingly's peak as superior to Hernandez's. That's because it doesn't define peak as some arbitrary small number of seasons. While Mattingly's best four seasons are better than Hernandez's best four, Hernandez had ten seasons worth more than five wins over replacement offensively (by my numbers); Mattingly had only four. The Pennants Added method recognizes that each of those ten seasons (and all of the players' other seasons) contributed something toward their teams' chances of winning the pennant. It doesn't ignore them just because it they fall outside of the player's best N seasons, for some arbitrary N.
That said, the extra "peak" value doesn't make much of a difference one way or the other in the comparison between these two players. The main differences are the walks and career length.
I enjoyed reading your last two pieces on the 4-man rotation. In today's article, you show numbers for 160 starters from 1978 to 2001 who pitched at least eight times on both 3 and 4 days rest. I have a couple of questions regarding that chart.
You make a very good point, and I was remiss not to account for this bias in my original study. (I did point out that the majority of pitchers in the study had lower ERAs on three days' rest than on four, and since that comparison had the same pitcher-seasons in each group, there was no potential for bias.)
The reason I did not try to account for the increased offensive numbers of the last 10 years was that so few pitchers in the study were affected by it.
Of the nearly 30,000 innings in the study, less than 900 of them--just 2.8% of the total--were thrown since 1993, when the modern hitters' era began.
Still, if we look at the innings in each group--those thrown on 3 days rest and those thrown on 4 days rest--there is a small difference in the distribution of those innings. Of all the innings in the study thrown through 1985, 47% were made on 3 days' rest. Since 1986, only 42% were made on 3 days rest.
Is this enough to change the conclusion of the study? I tried to alter the study to find out. Whereas, in the original study, I simply summed all the innings for all the pitchers in both groups, this time I calculated each pitcher's "average start" for the season in question.
For example, Phil Niekro made 32 of 40 starts on just three days' rest in 1978, while Denny Neagle made just 8 of 25 starts on three days' rest in 1995. The original study would end up with 40 starts in the three days category, and 25 starts in the four days category--but 32 of the 40 starts on three days rest came in the low-offense 1978, while 17 of the 25 starts on four days rest were thrown in high-octane 1995.
For the new study, Niekro's performance in 1978 would be whittled down to a single start in each category--that is, his numbers on three days' rest would be divided by 32, while his numbers on four days' rest would be divided by 8. The same adjustment was made for every pitcher in the study, so in the end we should end up with 160 matched pairs, with exactly the same number of adjusted starts in each group.
After making that adjustment, here are the new numbers:
Rest IP H ER BB K HR ERA H/9 BB/9 K/9 HR/9 3 days 1097 1046 442 343 603 87 3.62 8.59 2.81 4.95 0.71 4 days 1076 1055 467 344 602 93 3.91 8.83 2.88 5.03 0.78
As you can see, the rate numbers barely change at all. Pitchers still show a nearly 30-point improvement in their ERA on less rest, and they still show increased command (fewer walks per 9 innings) and increased sink on their pitches (fewer homers per 9 innings).
So to answer your question: yes, there is some bias in the original study. But accounting for it doesn't change the conclusion at all.
I was looking at the list for the teams, and I noticed that there were eight instances where a team on the plus side was in the same league/season as a team on the minus side. I was curious if there was any way to determine how much of the variance for those two teams was contributed just by the season series. For example, look at the 2002 season series between Minnesota & KC. Minnesota has won 12 of 16, yet I believe we have been outscored by KC by 3 runs. This should influence the Pythagoreans for both teams by 4 wins.
Parts of these are easy to calculate, thanks to the good folks at Retrosheet who have put the score for every game played since 1900 online.
Here are the matches:
I think the support neutral stats you post are terrific and I consult them frequently. However, I had a question regarding the updated stats for September 1st. I noticed that a lot of the numbers changed considerably. For example, Derek Lowe was at the top of the list yesterday with an SNWAR of 6.7. Now, his SNWAR has dropped to 6.4, and he has dropped to third place despite the fact that he didn't pitch yesterday. Pedro Martinez also dropped a bit, and Randy Johnson went up quite a bit. Neither of them pitched yesterday either. So my question basically is, what happened?
The answer is that I finally got around to updating the park effects. Fenway had been listed as a hitters' park previously, but it's actually been playing as a pitchers' park (neutral after the DH "ground rule" is taken into account) the past two years. So Lowe and the other Boston starters dropped a little. The BOB has been playing as a pretty extreme hitters' park the past two years, so Randy Johnson's and the other Diamondbacks' numbers were bumped up a bit. The changes to the numbers were fairly small in all cases, but they were enough to make the slight changes in the standings you noted.
Isn't the elimination of draft pick compensation for free agent signings going to be a disaster for small market clubs? Last offseason, the A's lost Jason Giambi, Jason Isringhausen, and Johnny Damon to free agency, but that allowed them to restock their organization with 7 top 40 draft picks.
I don't know that it's going to be a disaster, but it certainly isn't a welcome development. In the specific case of the A's, yes, it's a good thing that the A's received picks for the loss of Damon, Isringhausen, and Giambi, and if Ray Durham leaves without draft choice compensation, that's a bad thing.
I think you've clearly identified an effect of the new CBA, in terms of reduced leverage for "free agent sellers" around deal time. I don't know whether or not the revenue sharing money that will presumably flow to clubs in smaller markets (work with me...) will counteract the effect of not receiving draft pick compensation. Until we see it play out for a few years, we won't know for sure. Given that the new CBA is 4/5 years, we may not know until after the next CBA is in place.
I love your analysis of the game, but I wish you guys would stop shilling for the players in the labor debate. It's costing me money.