82 ( +1 | -1 ) Black's responses to 1.e4 - revisedIn my last set of stats, I compared victories for White with victories for Black, without considering draws; this set of stats takes wins and draws into account. I am looking solely at the database for Gameknot's +1900 games, which, of course, is not necessarily any indication of the effectiveness of these openings at higher levels of play. I do not mean to suggest that the reader should adopt any of these openings on the basis of these stats. They appear in descending order of Black's successes: * 1) The database includes 4,555 games with 1…c5, Black earning 46% of the points. 2) The database includes 1,372 games with 1…e6, Black earning 45% of the points. 3) The database includes 3,667 games with 1…e5, Black earning 40% of the points. 4) The database includes 307 games with 1…d5, Black earning 35% of the points. * I have not looked at the stats for 1...c6, 1...d6, or 1...Nf6.
61 ( +1 | -1 ) I think you'd want at least 20,000 games with each Black response to have an idea about approximate winrates. If you want accuracy, try a sample size of 50,000 or better. It really requires some experience with statistics (like bonsai) or even <gasp> gambling to have an appreciation of why 5,000 examples is a small drop in the bucket, and shouldn't be taken seriously. The GK database also does not keep track of the relative ratings of the players who are contained in it. Suppose that Black averages a marginal 40 more ELO points with 1...c5. The winrate for Black with c5 could be inflated by as much as 3-5%!!!