Source Lichess
White wins at least 5.1 more of every 100 games than it loses
how far wins run ahead of losses, 90% confident
At All Ratings, Italian Game: Evans Gambit, Tartakower Attack keeps White's wins running +5.1% ahead of its losses at worst, 90% confident, from 25K Lichess games. This gambit is effective. Win rate, draw rate and loss rate come from games played December 2023 to November 2025.
| Rating | Edge at worst | White Win | Draw | White Loss | Games |
|---|---|---|---|---|---|
| 1000-1199 | +5.7% | 57.5% | 3.2% | 39.2% | 168 |
| 1200-1399 | +7.7% | 56.5% | 2.2% | 41.3% | 481 |
| 1400-1599 | +6.8% | 54.7% | 2.1% | 43.2% | 1.2K |
| 1600-1799 | +7.6% | 54.1% | 2.2% | 43.6% | 3.0K |
| 1800-1999 | +5.1% | 52.3% | 2.7% | 44.9% | 5.2K |
| 2000-2199 | +0.5% | 49.5% | 3.5% | 47.0% | 6.7K |
| 2200-2499 | +1.6% | 49.7% | 4.6% | 45.7% | 4.6K |
| 2500+ | +4.3% | 17.0% | 72.1% | 10.9% | 3.6K |
The games never stop proving it. White comes out ahead at every rating we can see.
White wins 46.5% of Italian Game: Evans Gambit, Tartakower Attack games, draws 13.1% and loses 40.4%. That rests on 25K games from Lichess.
1200-1399. There White wins at least 7.7 more of every 100 games than it loses (90% confident), its highest Edge at worst.
Blitz. For White, the Edge at worst — how far wins run ahead of losses, 90% confident — is +4.6% in Blitz (14K games) and +2.9% in Rapid (3.5K games).
Yes. The first rating where the games do not show White coming out ahead is 1000-1199 in Bullet, 1000-1199 in Blitz, 1200-1399 in Rapid and 1800-1999 in Classical.
The Italian Game: Evans Gambit, Tartakower Attack is sharp: 13.1% of its games are drawn, so 86.9% end with a winner.
Play it. White comes out ahead at every rating we can see.
This one is hard to meet at every level. As Black, develop fast, castle early, trade pieces, and do not get greedy with the extra material. If sharp positions are not your game, decline it.
This rests on 25K games from Lichess. Every figure carries the range it is in, 90% confident, and the table above shows how many games sit behind each rating and each clock — the fewer there are, the wider that range.