Who would have guessed that the number of major league regulars on a roster would be such a poor predictor of how well a team would do in the World Baseball Classic?
The graph (click the image for a larger version) shows the number of major leagers plotted against the winning percentage of each team, through the semifinals. The two teams that will play in the finals, Japan and Cuba, are clearly at the far low end in major league talent. The two teams with the most major leaguers, USA and Venezuela, were just .500 in the tournament. The Dominican Republic, the team with the next highest number of major leaguers, was in the final four, but lost to Cuba, which has no major leaguers, in the semifinal.
For all sixteen teams, the number of major league regulars (MLR) is positively related to winning percentage, at just over 90% confidence. Each additional MLR contributes .012 to the winning percentage (W-pct), with a standard error of .007, and a t-score of 1.84.
However, the positive relationship is thanks only to such hapless teams as South Africa, Australia and the two Chinas.
When only the eight teams that escaped the first round (those whose names are underlined in the graph) are included in the analysis, the sign on MLR actually turns negative: Each additional MLR reduces W-pct by -.0058. This result, however, would be significant only at 80% confidence (which ought to be good enough for WBC blogging: standard error is .0039, t=-1.49).
How is a “major-league regular” defined? Simply. I looked at the rosters and identified the players I knew were in a starting lineup or rotation, were regular relievers, or back-up players who spent most of 2005 on a major-league roster, and players expected to be major-league regulars in 2006. It is unlikely that others with identical or better knowledge of major-league rosters would come up with materially different numbers, although the interest reader is invited to try.