=== Case In Pernt, Dept. ===
We keep saying that the Angels are beating Pythag by luck. Is this luck going to end anytime soon, pray tell? :- )
2009 -- 87 wins, 82 per Pythag (+5)
2008 -- 100 wins, 88 Pythag (+12!! )
2007 -- 94 wins, but only 90 Pythag (+4)
2006 -- 89 wins, only 84 Pythag (+4)
2005 -- 95 wins, only 93 Pythag (+2)
For the period 1998-2004 -- when the Mariners were good, and challenging them -- the Angels were even to Pythag. But the last five years, as Scioscia's boys have comfortably owned the division, they've gone out and routinely cashed in far more victories than the run differential predicted.
A bonus 5.4 victories per season is kind of big, when you consider that most divisions are won by 2-5 games.
Anybody want to bet that the 2010 Angels won't be over Pythag? I'd give 2:1 odds that they will be -- because the Mike Scioscia LA Angels know how to win.
They're going out there and grabbing those wins. You've been watching them do it. Don't tell me you haven't been watching. :- )
The Angels expect to win, they play to win, and they do win.
................
Now we're well aware that it's difficult to beat Pythag over (say) a 4-year period, and that not many teams do it. But the concept "no team ever beats Pythag intentionally" is not disprovable. It rests on assumptions.
..
=== Dr's Diagnosis ===
The 2009 Mariners just! now! finished! taking tough series from the Angels and Rangers. They didn't win those series because the Strat-O-Matic dice fell in again. They won because they played tough, because their pitchers hold games close at 2-1 or 3-1, and because they had Felix.
It says here that with the M's offense -- and scrub SP's -- the 2009 M's tended to win 4-2, and lose (occasionally) 9-2. I like math as well as the next dweeb, but I believe that Pythag applied less well to this roster than to the average roster.
... this roster was configured to win small when it did win (because it had no firepower). It was not capable of winning 12-3, but it was sure capable of taking beatings when a scrub SP imploded.
.................
As a completely separate factor, we've got the bullpen rocket ship out of baling wire act. In spring, D-O-V thought that the bullpen was doused in kerosene and Capt Jack was lighting up Marlboros with his elbow on it.
Somehow, some way, Jack and Wok not only found the only two relief aces available on 5 minutes' notice, they also identified them early ... and then coaxed great performances out of one scrub after another.
That sounds suspiciously like a repeatable skill to me, doesn't it you?
......................
The M's just won two more big ones the last two days. It's not luck. It's tough pitching, refusing to mail in a game when you're down 2-1 or 3-1, and then (sometimes) getting the late hit. The 2007 M's cashed in Thursday's White Sox game long before Jenks ever gets out there.
Wakamatsu has taught the Mariners how to win. Hopefully next year, he'll pick the right bullpen guys again, too. :- )
Cheers,
Dr D
...the 2009 Mariners are a 76-72 team by PythagenMatt...meaning the 77-71 actual record isn't very lucky at all for their distribution of RS/RA per game. Predicting that Seattle will find a way to have a P-Matt ahead of their P-Pat next year is risky...but their weakness is easy to identify and they have the capital to fix it...so...there's no reason to think that they can't go out, find two bats and be a legit .540 team next year by P-Pat.
Here's my problem with PythagenMatt's usefullness in cross-season projections ... the specific thing that P-Matt attempts to account for is *NON-STANDARD* run distributions. And it does a fine job of that. But, what it doesn't do AT ALL, is explain in any meaningful way *WHY* the non-standard run distribution is there. That's, of course, a job for the analyst.
That said - even if you correctly identify the "true" reason for the non-standard run distribution, when you change personnel, there is TONS of historical data to suggest that you are far more likely to remove non-standard run distribution than keep it. (Not impossible, but very, very unlikely). Of course, we don't have piles of data on P-Matt over multiple years, and perhaps when we do, we can discover sustainable pythagen skew methodologies.
My feeling is that there is a gigantic problem in the logic that a team can sustain 100% of it's positive regular pythagen skew while knowing full well that you're going to be changing personnel for the express purpose of changing the teams run generation, (and thereby, very likely, it's run distribution).
Say Seattle finishes -50 runs, but +9 to its raw pythag. If the club goes out and adds 50 runs to its run scoring, there is ZERO reason to believe they are inherently positioned to finish above .500.
And while P-Matt may have the club on-target, there is also a much, much larger can of data related to 1-run game performance from year to year. That data is NOT kind to teams that win lots of 1-run games. Seattle is 31-17 in 1-run games this season. If they had gone 24-24 in those 1-run games, the overall would be 70-78. Regular pythag says the team is 9 games "lucky". 1-run wins says the team is 7 games lucky.
When you have LOTS of different ways of examining "luck". There's a monumental danger for an analyst to choose to use the tools that make HIS team look really great, but use different tools for everyone else. There's also danger in looking at all of them, and then opting to believe the one that sells the best story for the analysts' favorite team.
At the very minimum, you'd need to run P-Matt on all the AL West teams for 2007, 2008, 2009, and see whether its any more (or less) predictive about which teams can sustain pythagen-skew.