"One more fastball arrived, high and over the plate, and Carbo smashed it in a gigantic, flattened parabola into the center-field bleachers, tying the game."
    -Roger Angell Agincourt and After 1975

Almanac

Table of Contents show

384 Is the Division an Object of Predictable Scope?

Posted by Joseph Tue, 17 Mar 2015 23:37:35 -0400
Filed Under: art expected winning percent predictions
permalink

As always when the off-season draws to a close we present the Southpaw Season Prediction system, our annual effort to divine each team's expected winning percentage (EWP) based on our calculations of past performances, aging adjustments, transactions, position changes, retirement, and so forth. In my anecdotal memory each season's predictions have closely resembled the final standings of the previous season. Which is boring but possibly eminently reasonable... good teams are often good from year to year, right? Note to self: “See if Charles has already done the work on this, or study it yourself. Or just do it really crudely in a hurry”.

Here's a table of our rate of success each season since we started this process in 2011, showing for each season how many division winners we predicted correctly, how many times we picked the defending division champ, how many times we got that right, and how many division winners actually repeated from the previous year:

YearCorrect Picks Repeat PicksCorrect RepeatersRepeat Winners
20113212
20122522
20132212
20143633
20155 note



Same old same old, really. Score one for my anecdotal memory. No conclusion to draw on whether good teams stay good... the metric's bad and the sample's too small. I don't know what the success rates of other prediction systems are, but I'd like to see us getting one more division winner right each season, and predicting a couple less defending champs.

Anyway, in addition to the classic prediction page linked to above, this year we have a new chart that shows all the team predictions in one table. Check out the Mets up at #3! FYI, just as on the original page the RVL and EWP prediction numbers are based on the unbalanced schedule.

Comments (1) hide

Comment by Charles Sat, 21 Mar 2015 09:10:04 -0400

I think we're doomed to pick too many repeats. Because we can't make precise predictions we remain conservative which makes us heavy on repeats. One trouble stems from the big changes of fortune that are invisible to Southpaw, like the Texas Rangers' massive spate of injuries in 2014.

Add a Comment:

Enclose player and team names in pipes (i.e. |John Grubb|) to auto-generate hyperlinks.
CSV Table Generator

Username: Password: