In the first of two articles on translations for winter and fall leagues, Clay Davenport crunches the numbers to find out just how good these leagues really are. You can read more material by Clay on park effects and league translations in Baseball Prospectus 2004.
The regular seasons for the various winter leagues are over now, although the various league playoffs leading to the Caribbean World Series are still ongoing–a process that runs almost as long as their regular season, kind of like the NBA.
The Caribbean World Series is a contest between the winners of the four regional leagues: the Dominican, Puerto Rican, and Venezuelan winter leagues, plus the Mexican Pacific League. What I’m going to do here is to give an overview of those four leagues, plus the Arizona Fall League, with an emphasis on how to make sense of winter league statistics.
The process for working out the talent level of a league depends on, number one, having a large number of players in the circuit who have played in other leagues; and number two, knowledge of how good those other leagues are. Every player who has played here and elsewhere becomes a data point: You rate the player’s hitting (or pitching) level, relative to league average, with park adjustments when you have them, in both leagues. If his relative offensive level gets worse, that is a (slight) argument that the new league is tougher than the old. If it gets better, that’s an argument that the new league was easier.
The weight of any individual’s argument is determined by his common plate appearances (CPA) between the two leagues–simply the lesser of his PA in the two leagues. (Other means of equalizing the PA between the two leagues are available, and really wouldn’t create a meaningful difference in the analysis. It is essential, though, that each individual player’s performances be given the same weight in each of the two data sets.) One, two, even 10 players is a very weak argument, but with the winter leagues we are almost always able to use every single player in the league, getting us into the realm of 100 players and 10,000 CPA. With these numbers, we can be fairly confident–if not quite certain–that individual variations have balanced out.
It also helps if you minimize the time differences in your comparisons: The more time you cover, the more likely it is that the player himself will change, upsetting the fundamental assumptions of the procedure. For these tests, I have only used the season immediately before and after the winter league, except for the 2003 winter leagues, which are only rated against the 2003 regular season.
So you end up with a lot of lines that look something like
41. DIAZ, JUAN 2001 AFL 0.306 2001 INT 0.305 78. HUDSON, ORLANDO 2001 AFL 0.419 2001 INT 0.311
This shows the CPA, player name, and EqA in each of two leagues. Repeating that for every common player between the leagues gives you something like this:
980 2001 AFL .294 2001 INT .278 .947
The 2001 AFL rates as being 5.3% easier than the International League, in terms of EqA. The IL, in turn, is rated at 13% below the AL (again, in EqA terms), making the net assessment of the AFL to be 18% below the AL.
That’s based solely on the International League. Repeat the process for the Pacific Coast League, the Eastern League, and all the other known leagues to get the final difficulty rating.
The Arizona Fall League, played in Phoenix and its suburbs during October and November, is supposed to be a sort of finishing school for top prospects to face each other, to test themselves against stiffer competition than they’ve ever faced before. It is a very high-offense league, with a .276/.349/.421 line over the past three seasons. During that time, 80% of the CPA in the AFL have come from Double-A and below (even the year after playing in the AFL, more than half of the CPA are still Double-A and under). The results (2003 AL = 1.000):
2001 2002 2003 Avg Value .848 .835 .832 .838 %Mj 4 2 3 3 %AAA 17 17 15 16 %AA 57 54 56 56 %A 22 27 27 25
(Percentages reflect the common plate appearances between the AFL and the preceding regular season).
Given that the Triple-A leagues rate between .850 and .870, and the Double-A leagues come in between .790-.800, the AFL does in fact give the majority of players a taste of competition that is more advanced than what they have faced before.
It was widely reported last fall that the pitching quality of the AFL was down substantially from prior years. As more teams make minimizing pitchers’ workloads a priority, there are fewer pitchers available for the extra work in Arizona. The tests above would not detect a change in pitcher quality–so long as the pitchers were still good enough to separate the better hitters from the lesser–so I reran the study from the pitching side:
2001 2002 2003 Value .820 .847 .836
From these numbers, it is 2001 which stands out as the anomaly, where the pitching quality was substantially worse-rated than the hitters. In 2002 the pitchers were somewhat better than the hitters. In 2003, (keeping in mind that we only have half as many comparisons to work with), the two were essentially even.
Dominican Winter League
The DWL consists of six teams, and has been covered by Carlos Lugo in several articles for BP. The offensive level is moderate, averaging .257/.323/.384 since 2000. From this season only, a sample of only 25 home and away games per team, it would appear that the Aguilas and Azucareros represent extremes of park factors, rating at 1.073 and .899 respectively. The Gigantes and Estrellas were apparently a little above average at 1.031 and 1.022, while the combined scores for the Leones and Tigres (who share the same stadium) was .961. To give you an idea of how variable the park factor ratings can be over 50 games, the Leones alone had a score of .877, while the Tigres were at 1.045–even though they played in the same stadium.
We have four years of data for the Dominican League, and its ratings are:
2000 2001 2002 2003 Avg Value .834 .836 .840 .838 .837 %Maj 27 24 23 26 25 %AAA 48 45 48 36 44 %AA 17 17 17 16 17 %A 9 10 4 10 8 %Mex -- 6 9 11 6
I don’t have the 2000 Mexican League statistics in the database, and yes I know it throws the average off. Relax. The DWL has the highest percentage of major league and Triple-A players among the Caribbean leagues, but the difference in quality is actually quite small.
Puerto Rican League
The Puerto Rican league is similar to the Dominican league: six teams, one island, emphasis on major league and Triple-A players. The PRL has a slightly higher offensive level (.261/.341/.395 since 2000), with the biggest difference between them being that the Puerto Rican league draws 15% more walks. Its table:
2000 2001 2002 2003 Avg Value .841 .822 .833 .842 .835 %Maj 24 23 26 19 23 %AAA 37 41 36 40 39 %AA 26 16 20 22 21 %A 13 12 10 11 12 %Mex -- 9 9 9 7
The difficulty ratings are essentially identical to those from the DWL, although there is more interannual variability. I don’t think the variability is significant.
Venezuelan Winter League
The Venezuelan league is larger than the two island leagues, with eight teams. Its offensive level is slightly higher (.269/.339/.401 since 2000).
2000 2001 2002 2003 Avg Value .831 .838 .812 .833 .829 %Maj 16 13 12 15 14 %AAA 36 29 36 36 34 %AA 27 28 27 32 29 %A 21 21 17 8 17 %Mex -- 9 10 9 7
The drop in quality in 2002 was probably the result of the political turmoil which disrupted the league and forced it to shut down early–because of the crisis, many non-Venezuelan players either didn’t come or left early. The percentage of CPA from A-ball players was way down this past year; I don’t know if that was a deliberate design or perhaps an artifact of last year’s political problems which held down the number of Venezuelan players who signed with American teams.
Mexican Pacific League
The MPL is a distinct league from the summertime Mexican League. The MPL consists of eight teams, playing in northwestern Mexico (most of the cities in the league are on the mainland coast along the Gulf of California/Sea of Cortez). Since none of them are at any significant altitude, the MPL has a decidedly normal offensive level (.257/.341/.396 since 2001), unlike the extremely hitter-friendly summer league. And even though half of the CPA in the league come from the summer league, the influx of Anglo players in the wintertime makes this the stronger league by a clear margin.
2001 2002 2003 Avg Value .834 836 .840 .837 %Maj 6 9 7 7 %AAA 29 40 28 32 %AA 13 15 4 11 %A 4 5 1 3 %Mex 48 31 62 47
The 2002 Mexican league stats I have are incomplete; I suspect that the actual percentage of CPA in the league is on the order of 50%. The summer league rating is .780.
Coming in part 2: High and low performers from the winter leagues, and what it means for them in 2004.