A couple days ago, Chris St. John posted an article looking at how various Top 100 lists saw certain players differently than the consensus in terms of liking, disliking, including and excluding them. This is a very useful way of comparing lists from a micro (individual player) perspective, but I am also interested in the macro perspective - that is, how do to the lists compare with each other in terms of generally agreeing or disagreeing on prospect rankings.
Chris included 14(!) lists in his article, but I am only comparing 8 of those lists here - Baseball America, Bullpen Banter, Kevin Goldstein, Keith Law, Jonathan Mayo, Frankie Piliere, Project Prospect and John Sickels. The primary reason is that I already had the data from doing my farm rankings, but also, I'm sticking with the lists that are done by people who I know see a good number of players. For fun, I've included the community ranking done here at Minor League Ball just for comparison's sake.
For starters, I compared each list against each of the other lists to see how many players were common to both lists (for lists that went beyond 100, I cutoff the lists at 100 to ensure a fair comparison):
One thing to note is that the table is a mirror image along the diagonal line - I did that so that it was possible to just read right across horizontally or vertically. The average number of common players between any two lists was 80, I highlighted both the maximum (88) and minimum (74). The community list, with an average 82.5 players in common with the other lists, had the highest average similarity, which makes sense since it represents an amalgamation of views. The Project Prospect list had the fewest average number of common players on thier Top 100, at 75.6. In general, however, there was a fairly high degree of commonality between the players included on the Top 100s.
Of course, merely having the same players on the lists does not tell the whole story, as lists can signifcantly vary on the placement of the players. Therefore, the next step is to look at the average number of spots by which players were seperated on lists. This is complicated by the fact that when comparing two lists, not all players are common to both lists. In these cases, I just assumed that the players not on one list would have been the next players ranked. So if 20 players on a top 100 list were left off the top 100 list with which it was compared, I just assumed all 20 players were ranked 110.5 (average of 101-120). Obviously, it's not the way things actually are, but I think it's the most reasonable assumption given the information available (and is superior to omitting those players, since the greatest differences occur in these players). I got the following average differences:
The list with the smallest average ranking difference vis-a-vis the others with the community list - again, not terribly surprising given the way the list is complied. Among the others, John's list and Baseball America's list had the lowest average difference in ranking, and Frankie Piliere and Project Prospect ended up with the highest average difference. The two lists with the lowest difference in ranking were John's and the community list; whereas Project Prospect and Frankie Piliere had the largest difference in rankings.
I conclude with a table showing, for each list, which other list (excluding the community list) is most similar (lowest average ranking difference) and most dissimilar (highest average ranking difference):