Type a BGG username in each box to calculate the correlation between the ratings of the two users. Type a BGG username in the first box and leave the second box empty to match that user against everybody in the cache. That'll take a while, so please be patient. Because the process is fairly heavy, the minimum number of common games is 75 (also, the data is fairly meaningless with small sample sets).

For further information on correlation, check Correlation at Wikipedia. Basically, correlation of +1 or -1 means one set of ratings can be predicted from the other; positive number means the ratings are in perfect agreement, negative means the ratings are in perfect disagreement. Values near zero mean the ratings aren't useful in predicting each other. With small samples, these values have very little significance; with larger samples, they start to make sense.

If you match against all users and make sure "Refresh cache" is on, you'll find "BoardGameGeek average rating" on the list so you can see how well your ratings correlate with the Geek average rating.

This correlation calculation tool written by Mikko Saari (mikko at mikkosaari.fi). Lautapeliopas is the leading Finnish board game resource.