A trap in the Sicilian, O'Kelly Variation

The O'Kelly Variation of the Sicilian Defense, 1.e4 c5 2.Nf3 a6, is an offbeat line that sets a bit of a trap at move 2! White's natural response, and easily the most popular move in the databases, is 3.d4?! That move, however, gives Black an easy game, and a 58% score, after 3...cxd4 4.Nxd4 Nf6 5.Nc3 e5! Then, 6.Nf5?! is met by 6...d5!; 6.Nf3 and 6.Nb3 by 6...Bb4!; and 6.Nde2 by 6...Bc5, in each case with equality or better for Black.

It can get considerably worse than that, as seen in the game below. After 6.Nf3 Bb4 7.Bd3, Black played the thematic break 7...d5! White responded with 8.exd5?, presumably expecting 8...Nxd5. Instead, he was shocked by 8...e4!, winning a piece.

White's best response to the O'Kelly is 3.c3! Then Black usually plays 3...d5, transposing to an Alapin Sicilian where the move ...a6 is not very useful. In my opinion, this makes the O'Kelly not worth playing against players rated, say, 2000 and above. But some players stronger than I disagree. IM Irina Krush has played the O'Kelly many times against high-level opposition - though she's struggled when her opponents have found 3.c3! (she scored 1-2 against that move in the three games I found). Incidentally, 3.c3 d5 4.exd5 Nf6!? 5.c4 e6 is an interesting gambit that scores about 50% for Black. But White can decline the gambit with 5.d4, when Black has nothing better than 5...Qxd5, reaching the same inferior Alapin Sicilian line.

I should mention that 3.c3 e6 4.d4 d5 is also possible (and perhaps objectively best), reaching a French-type position where Black scores over 47% (as opposed to less than 42% with the more common 3...d5). But most Sicilian players aren't happy playing the French.