Norway Armageddon

[lkaufman]

Some amazing results in the Armageddon playoffs in Norway: After every classical draw, a second game is played with the same colors with White getting 10 minutes vs. 7 minutes for Black (each with one second increment after move 40), with Black winning if this game was also drawn (Armageddon scoring). This was expected to be about fair, based on prior experience. But incredibly, after 16 Armageddon games, the score stands at 14 to 2; Black only held the draw twice in 16 games!! This despite the fact that on average the players with the White pieces tend to be the weaker players, since when the stronger player has White he is more likely to win the initial game than when he has Black! Most people probably thought that Armageddon only works in blitz, that the game is too drawish for Armageddon to be reasonable at longer time controls, but this seems to be very far from the truth. This suggests that Armageddon chess with time odds might even be quite practical for Classical Chess, for example White gets 90 minutes plus 30 seconds vs. 60 minutes plus 20 seconds, or something like that. It seems that White's advantage in chess is more than was generally believed.
So what does this mean for computer chess? Based on tests a few years ago, it seems that engines need quite large time odds for White to be able to overcome draw odds against an equal opponent, especially with longer time limits and/or more threads. But perhaps it's worth some tests to see what time odds White needs if Black has 2' + 1" and draw odds, with no opening books and say four threads, with top engines on each side. Probably it would be a pretty big ratio, perhaps even 10' + 5" for White wouldn't be enough, but anyway worth finding out.

[jp]

Some amazing results in the Armageddon playoffs in Norway: After every classical draw, a second game is played with the same colors with White getting 10 minutes vs. 7 minutes for Black (each with one second increment after move 40), with Black winning if this game was also drawn (Armageddon scoring). This was expected to be about fair, based on prior experience. But incredibly, after 16 Armageddon games, the score stands at 14 to 2; Black only held the draw twice in 16 games!!

It's a very interesting result, isn't it? But Black has struck back in the last round.

Perhaps we should look in more detail at what has gotten Black into trouble. e.g. I happened to be watching when Topalov, who was doing fine against MVL, blundered a pawn (the f7 pawn was overworked defending both e6 and g6), and that cost the game.

[Ovyron]

In the age of NNUE and big hardware I'd expect time odds need to be HUGE for this to work for engine. Like, white gets 30 minutes +1 sec inc, black gets 1 minute +1 sec inc. But first people need to code this into Stockfish, naive changes to the code making it think a draw is a win don't work.