AylerKupp: <<jindraz> I think Shirov really refers to much higher level than just master games.>That makes sense. So, since the ChessTempo database allows you to filter games in rating point increments of 100, I calculated a similar set of statistics for those games where both players were rated 2600+ and for those games where both players were rated 2700+. Here are the results.
<A. Both players rated 2600+>
Games starting with 1.e4: 26,056 games. White wins 30.4%, Black wins 18.5%, 51.1% draws. White winning percentage 56.0%
Games starting with 1.e4 e5: 10,314 games, 39.6% of games starting with 1.e4. White wins 28.1%, Black wins 16.9%, 55.0% draws. White winning percentage 55.6%
Games starting with 1.e4 e5 2.Nf3 Nc5 3.Bb5: 6,575 games, 25.2% of games starting with 1.e4. White wins 28.1%, Black wins 15.8%, 56.1% draws. White winning percentage 56.2%
Games starting with 1.e4 e5 2.Nf3 Nc5 3.Bb5 Nf6: 1,989 games, 7.6% of games starting with 1.e4. White wins 24.4%, Black wins 15.4%, 60.2% draws. White winning percentage 54.5%
<B. Both players rated 2700+>
Games starting with 1.e4: 5,614 games. White wins 28.7%, Black wins 18.2%, 53.1% draws. White winning percentage 55.3%
Games starting with 1.e4 e5: 2,610 games, 46.5% of games starting with 1.e4. White wins 25.9%, Black wins 15.4%, 58.7% draws. White winning percentage 55.3%
Games starting with 1.e4 e5 2.Nf3 Nc5 3.Bb5: 1,732 games, 30.9% of games starting with 1.e4. White wins 25.2%, Black wins 15.6%, 59.2% draws. White winning percentage 54.8%
Games starting with 1.e4 e5 2.Nf3 Nc5 3.Bb5 Nf6: 636 games, 11.3% of games starting with 1.e4. White wins 22.5%, Black wins 14.6%, 62.9% draws. White winning percentage 54.0%
So in cases where both players were rated 2600+, White's winning percentage after 1.e4 e5 2.Nf3 Nc5 3.Bb5 Nf6 (54.5%) was only 1.5% less than all the games started with 1.e4 (56.0%). And in cases where both players were rated 2700+, White's winning percentage after 1.e4 e5 2.Nf3 Nc5 3.Bb5 Nf6 (54.0%) was only 1.3% less than all the games started with 1.e4 (55.3%). I'll let everyone decide whether they think that a 1.3% to 1.5% decrease in White's winning % constitutes "an end to nothing less than the move 1.e4". And since in only 30.2% of Ruy Lopez games where both players were rated 2600+ and in only 36.7% of games where both players were rated 2700+ did the Black player choose the Berlin defense when they had the opportunity, I don’t think that Shirov's opinion was shared by these players.