Chess Opening Theory/1. e4/1...e5/2. Nf3/2...Nc6/3. Bb5/3...a6/4. Ba4/4...Nf6/5. O-O/5...Be7/6. Re1/6...b5/7. Bb3/7...O-O/8. c3/8...d5/9. exd5/9...Nxd5/10. Nxe5/10...Nxe5/11. Rxe5/11...c6/12. d4/12...Bd6/13. Re1/13...Qh4/14. g3/14...Qh3
Ruy Lopez, Marshall Gambit
We've arrived at a critical point in the Marshall Gambit. Black has a simple threat of ...Bg4, after which the White queen will have to guard against the checkmate threat with ...Bf3 and ...Qg2. This would draw the queen away from the defence of the e1-rook, and the unsupported rook would then have difficulty resisting a Black takeover of the e-file starting with ...Rae8.
The old main line is 15. Be3. White pre-emptively blocks the e-file against Black's potential takeover. 15. Be3 makes sense in the context of White's development; the b1-knight will have to come to d2 if it hopes to influence matters on the kingside, but playing it immediately would box the c1-bishop in, so get that bishop out of the way first.
But what does the bishop do on e3? Well, nothing. Black has no threats against either f2 or d4, and a few moves down the line White may well be forced to meet a Black pawn push to f5 with the blocking move f4 (to stop Black playing ...f4 herself), at which point the bishop itself will be a target of the d5-knight and a rook on e8.
If White wants to take decisive action against the invading Black queen, the move 15. Re4 threatening Rh4 and preventing ...Bg4 is an excellent alternative to the main line. How excellent? Excellent to the tune of giving White 34% wins instead of 26%, and restricting Black to 23% wins instead of 32% (yes, Black wins considerably more games than White in the 15. Be3 line!) There is a caveat: White players of 15. Re4 are about 50 points higher rated to begin with. But there is a caveat to the caveat: after normalising for the White player's grade, 15. Re4 still outperforms 15. Be3 by about 80 points. In fact, as of 2018 White's last 13 wins against GM opposition have come in lines other than 15. Be3.
The general feeling on ChessPublishing.com is that Re4 requires more work - more memorisation of lines - than Be3, which may explain why anyone at all still plays Be3.
15. Qe2 is starting to get some attention too. White's queen hustles over to f1, the square it needs to be on to cut out any light-squared mating ideas. Since it could have reached f1 in a single move after Re4, White is essentially arguing that e4 is a worse square for the rook than e1.