Best and Worst Playoff Teams Since 1995, and how they did

Thanks, it’s fixed now.

What alarms me here is that only 30% of the 10 best teams in the last 13 years have won the World Series. Eesh. Look at 1998. 3 of the best 10 teams come from the same YEAR. Must have been a lot of weak teams back then.

It’s tough to design a methodology for comparison within one year. I can’t think of anything more meaningful that just listing the data.

One design I considered, which I thought would be much better (and then didn’t do because it was too much work) was to go series by series, and see how well the difference between two teams’ RD’s predicted the winner of the series. So, the predictive variable would be x = RD(team A) – RD(team B)
In cases where x is large, one would hope team A wins most of the time. Then you could categorize P(series win) for 0<x<10 (toss-up), 10<x<20, 20<x<30 and so on. I think the best part of this method is that it really leverages the sample size, since we’d have 114 series, and probably enough in each bin of run differential.

I don’t know how clear that was…it’s hard to type this kind of stuff.

and it’s rare indeed that the series is going to come down to home field advantage. Generally speaking being on your home field is worth around half a run for a single game (home teams win about 55% of the time). So having one extra home game in a series has a pretty minute impact on the win probabilities in the series.

The rest of the difference in performance is presumably caused by the home teams being, on average, better clubs. Not much of an edge there. There are no “1 vs. 16” matchups in the baseball playoffs— every team has a substantial chance of losing every series. A 105-win juggernaut will go down to an 83-win mediocrity that squeaked in from a weak division a non-negligible fraction of the time.

His analysis basically looks logical— there’s a clear gradient downward from the best teams to the worst ones in terms of playoff performance. It’s just not a very steep gradient. Certainly not steep enough to justify trading long-term assets for short-term upgrades (eg this year’s Mark Teixeira trade).

A few flaws with just using run differential as a proxy for team quality:
1) It ignores homefield advantage, which I think is a significant factor in the postseason (and a reason to want to maximize wins above PT’s 95 number). This is particularly important when looking at the World Series because A) homefield advantage isn’t determined by team quality, and B) a run differential in one league is not necessarily equal to a run differential in another league due to disparities in league quality.
2) Schedule differences. The AL West, for example, had the best record of any division in the AL every year from 2000-2007, which made it harder for a an AL West team to rack up run differential than it would be for a an AL Central team.

It would be impossible to design this study to adjust for home-field. I’d have to come up with an adjustment factor, which would require a much better study than this one.
Same for #2…

3) Raw run differential isn’t as good of an indicator as Pythag—a team that outscores its opponents 800-600 should be expected to have a better record than a team that outscores its opponents 1000-780, even though the latter has a better run differential.

I wish I had done pythag. However, data entry was hard enough here. RD values were availible directly since 2001. Pythag would require twice the data entry, since I’d have to enter RS and RA separately. I still haven’t figured out how to get CSV data sets yet. Baseball prospectus’ custom statistics reports never seem to work.

4) Teams change in quality over the season with trades, promotions, and individual improvements/declines.

Probably not too substantially. I guess a superstar would mean 2-3 Wins, which over a series, may not be too big. Also, impossible to account for…tracking acquisitions = too much.

5) Player usage is different in the playoffs than in the regular season. In the playoffs, you don’t generally use your 5th starter or the back end of your bullpen, for example. You also don’t use reserves to rest starting position players.

THIS IS A GOOD ARGUMENT. I like it. I made a similar one when arguing the playoffs are not a crapshoot. I might try to find a way to test this…but didn’t think of one last night. plenty of regular season RD can be explained by horrendous #5 starters…This would deflate (or inflate, in the case of a great #5 starter) Regular Season RD, without impacting Playoff win odds whatsoever.
Definitely one reason why we didn’t see much predictive value in the above data.

makes sense. It was cruel watching Detroit losing that series. OTOH, our 2006 team was probably not ALCS quality, so can’t complain about that playoffs.

First of all, I don’t know if you read the fanpost, but I actually just did all this work to disprove my own original position. So your acting as though I’m being repetitive (“once again”) is cruelly ironic…you’re the one repeating without listening.

On the actual substance of your (argument?) claim…
The outcome of a series IS binary (W/L).
The Outcome of playoffs, (IE WS Champ) IS binary…you are champ, or you are not.

You defined crapshoot as a normal distribution…
1. You are the only person I have ever seen define it that way. Urbandictionary has it as ‘toss-up, roll of the dice’, while random house (via dictionary.com) has it as “anything random or risky”.
2. That’s not the definition people use when they deploy the argument regarding the playoffs. What they’re saying is that you can’t build a team to win the playoffs, just make it and hope. If crapshoot meant normal distribution, that would mean teams SHOULD go for upgrades even if they are going to make the playoffs either way, since raising their mean talent level would raise the probability of winning it all. Your definition is the EXACT opposite of what they mean in this context. They mean that improving your talent (thereby increasing the “mean” of the normal distribution) is pointless, since the playoffs are random.