Happy Thanksgiving! Regularly Scheduled Articles Will Resume Monday, December 1
February 5, 2004
Winter Leagues, Redux
What Can We Learn About Players' Futures?In the first part of our winter leagues study, we looked at what the player performances said about each league's level of quality. In Part II, we'll look at what player performance in these leagues say about a player's future.
The first thing to remember is that winter league statistics are forged in short-season leagues, so any stats represent a small sample relative to regular-season play. The highest AB+BB totals in these leagues over the last few years have been:
In order to test changes in player performances, I created a test set of players who met the following criteria:
This gave me 391 players, or rather 391 sets of three player-seasons. All of the statistics were translated, using the difficulty ratings in the previous half of the article. From the summary statistics, it is apparent that they performed similarly as a group in each of the subsets:
You would expect that the correlation between player EqAs in the regular seasons would be stronger than the correlation between either regular season or winter ball, simply because the smaller number of PA from winter ball increases the variability in player performance. That is, in fact, exactly what you find. The correlations are:
Next, I sorted the players based on the difference between their equivalent runs per out in the prior regular season and in winter ball. I am using equivalent runs per out, not equivalent average; EqA is nice for display purposes, but it isn't linear, and that is important for this study. The question at hand is, if a player improves between the regular season and winter ball, does that spill over to the following season?
The answer to that is a qualified yes--it does, if the difference is big enough.
To give people an idea of how big the differences are, a player with a .260 EqA (league average, by definition) will generate .172 equivalent runs per out (EqA, raised to 2.5 power, times 5 = .172). If he improved his offensive level by .025 EqR/out, his EqA would increase to .274. Improve by .050 EqR/out, and EqA is up to .288; improve by .075, and our formerly average hitter is up to .300. Declines of .025, .050, and .075 would reduce our average hitter to EqAs of .244, .227, and .207.
"Change in EqR/out" is winter league EqR/out minus regular season EqR/out. N is the number of player-seasons within the range. Nsame is the number of players whose following season's EqR/out changed in the same direction as their winter league EqR/out, although not necessarily by the same amount. Percent is 100*Nsame/N. "Chance" is the cumulative binomial probability distribution for Nsame and N, assuming a 50/50 chance for Nsame; the closer chance is to 1, the more likely this is a real effect, and not just random dumb luck.
Players whose winter league EqR/out scores improved by at least .05 EqR/run over the prior regular season had a 71% chance of having a better EqR/out (and hence a better EqA) in the following regular season than in the previous one. Players whose winter EqR/out dropped by .05 or more had a 78% chance of having a worse EqR/out (worse EqA) than in the previous year. Taken in total, players whose EqR/out changed by .05 or more are three times as likely to see their regular season EqA change in the same direction as their winter league performance. The average change was about one-fourth the winter league change.
Forty players improved their EqR/out by at least .050 points in 2003. In the list that followed, I've given the player's organization (as of 2003), the winter league he played in, his change in EqR/out, his combined 2003 EqA, and his "expected 2004 EqA," defined as 2003 EqA plus one-quarter the change in EqR/out.
By the same token, there were 39 players who lost .050 EqR/out or more.
For many of the players on these lists, their winter league performance really indicates nothing more than a return to form following an unusual regular season. Robert Machado, for instance, hit for a .227 EqA in 2001 and 2002 before busting out to .275 in 2003. His failure to hit in the Venezuelan league is hardly surprising, any more than the suggestion that he'll drop in 2004. Virtually every player on the down list (all but four) had a 2003 EqA that was higher than their 2001-03 average, itself a compelling reason to expect declines in 2004; the winter league results just add icing to the sweet rolls.
Similarly, 30 of the 40 players on the up list were coming off sub-par 2003 seasons. Some of the interesting names that were NOT coming off sub-par 2003s were:
Luis Rivas, who hit for his normal .242 EqA last year, but stuck a .287 in Venezuela.
D'Angelo Jimenez, whose .261 last year was well above the .249 he hit in 2001-02, went out and hit for a .292 EqA in the Dominican. BP has been high on him for years; could this be the year he returns the favor?
Adam LaRoche hit for a .218 EqA in 2001-02, as a first baseman in the Braves farm system. Last year he hit .247 at Double-A, raised that to .256 after being promoted to Triple-A, and took that to .282 in Puerto Rico. The winter league numbers suggest that last year's improvement was real, and that he could be a viable first baseman in the majors.