Bill James website and "cursed" teams

The "Bill James Online" website is now up and running. It's got about 50 articles by Bill James, and a bunch of statistics. The statistics aren't the usual numbers, but, rather, more interesting Jamesian stuff. For instance, there's something called "Ghost Runs," which is when a runner is out on a fielder's choice or forceout, but the runner who takes his place later scores.

The organization of the website, unfortunately, makes it difficult to download large quantities of this data, but you might have some screen-scraping techniques that might help.

Anyway, my main interest isn't the stats, but the articles – many of them are full sabermetric studies, of the usual high Jamesian quality. It's $9 for a three-month subscription, which, in my opinion, is well worth the price: assuming another article a week, that's about seven for a dollar. I'd pay several times that price; there are few things in life I enjoy more than a Bill James study, and those cost a lot more than 15 cents.

I'll review a few of the articles as I read them. Here's one now.

In the study called "Curses" (subscription required), Bill figures out that the chance of a team winning the World Series is almost exactly proportional to the cube of the difference between its wins minus its losses. So if the Yankees go 100-62 (38 games over .500), while the Red Sox go 97-65 (32 games over .500), New York has almost a 50% higher chance to win the World Series than Boston does.

Of course, that doesn't mean *in the same season*. Obviously, if the Yankees and Red Sox had those records in the old two-division AL, the Red Sox would have *zero* chance. Rather, what it means is that if those teams had those records in two separate seasons, the Yankees would be 50% likelier to win their WS than the Red Sox would be to win theirs.

Bill nonetheless uses this method to figure the chances in the same season. He lists all the teams with winning records from 2003, and gives them "claim points" based on the cubes of their games over .500. The Yankees were 40 games over, so get 64,000 points. The Blue Jays were 10 games over, and get 1,000 points. Therefore, the Yankees have 64 times the chance the Blue Jays have.

The sum of all the teams' cubes is 272,767. So the Yankees chances of winning the WS are 64,000 divided by 272,767, which is 23%. The Blue Jays are 1/64 of this, at an 0.4% chance.

It seems weird, but it also seems to work. Bill writes,

-- Of the teams which were estimated to have a 40 to 50% chance to win the World Championship, 48% actually did.
-- Of the teams which were estimated to have a 30 to 40% chance to win the World Championship, 33% actually did.
-- Of the teams which were estimated to have a 20 to 30% chance to win the World Championship, 22% actually did.
-- Of the teams which were estimated to have a 10 to 20% chance to win the World Championship, 17% actually did.
-- Of the teams which were estimated to have a 5 to 10% chance to win the World Championship, 7% actually did.
-- Of the teams which were estimated to have a 1 to 5% chance to win the World Championship, 3% actually did.
-- Of the teams which were estimated to have less than a 1% chance to win the World Championship, one of 339 actually did.

So now we have the ability to estimate how many Series teams should have won based on their historical W-L records. Between 1920 and 2003, six teams were at least one championship below expectation:

Expected 2.69, Actual 0: Red Sox
Expected 2.21, Actual 0: Cubs
Expected 4.17, Actual 2: Braves
Expected 4.07, Actual 2: Indians
Expected 4.84, Actual 3: Browns/Orioles

Overall, the Yankees were the least-cursed team in history: expected to win 16.7 World Series, they actually won 26.

I'm not completely sure what I think of this method, but the fact that it's so accurate certainly impresses me. When you use it to measure how much a team is "cursed," you're limiting it to one specific sense: having a good year, but failing to win the World Series despite your good year. If there's a curse that keeps your team at 79-83, this method won't tell you.