Thursday, May 21, 2009

Stats (Huh?)


Readers might come away from examining the Jerome Gambit (1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.Bxf7+) database statistics – see "Stats (1)", "Stats (2)" and "Stats (3)" – with a small sense of confusion, as the Opening Reports seem to both show that White scores "better than average" in a number of lines; yet there are Critical Lines where White scores poorly.

(Readers might also want to compare the analysis with the "A Closer Look" series, which took a look at the Ninja Knights T3 Jerome Gambit thematic tournament: 1, 2, 3, 4, 5 ; as well as look at the Comments following "Jerome Gambit: Transylvanian Terror!")

This result reflects both the nature of sampling bias (that is, because of how the games were collected – that Jerome Gambit wins are more likely to be published and available to the researcher than Jerome Gambit losses – there are more opening successes than failures, despite the opening's "objective" strength); and the nature of the opening itself (defenders are often taken by surprise, become intimidated, and do not take advantage of the winning opportunities that are presented to them).

A comparable case would be a database that collected all available games in which Queen-odds were given. Although being a Queen up is usually considered enough of an advantage for all but the rankest of amateurs to win with (see Geoff Chandler's "Blunder Table"), it is highly likely that the results of the database would highly favor the odds-giver.

No comments:

Post a Comment