About queen for a bishop handicap by nakamura.

[lkaufman]

There are many blitz games when nakamura beat botez when he sacrificed his queen in the opening for a bishop and won.

I wonder what is your opinion of this handicap and if it is more than rook handicap or less than rook handicap.
Based on the value of pieces it is bigger than rook handicap but I think that practically it is not and I wonder what is your opinion.

The problem is that it is more easy to play when you have more pieces and with queen for a bishop advantage it is not the case.

I am sure that queen for bishop is a much bigger handicap than rook. It's quite easy for a much stronger player to develop an attack when missing a rook, but without his queen (even for a bishop) it is almost impossible to attack. I would agree than queen for bishop is a smaller handicap than two knights (both nominally six points by normal count) for the reason you mention, but two knights is WAY more than rook odds.
Regarding the Nakamura-Botez games, two questions: 1. Roughly what percentage of points were scored by Nakamura? (You say he won many games, but maybe he lost even more?) 2. Did he just give away queen for bishop more or less at random, or was there often some other compensation involved?
Also, are there other pairings where Nakamura has played a lot of games giving some normal handicap to a specific reasonably strong player? At 3 minute chess with no inc, I would guess he could give rook odds to an average FM and knight odds to an average IM; he just has to last long enough to win on time. With even a one second increment it is much harder.
I think it would be an exciting and reasonably close match for Komodo Dragon to give Nakamura knight odds (g1) at 3 min no inc. chess. I say this because Dragon is about even with GM Alex Lenderman at those odds at 5' + 3", and it's not obvious whether Nakamura with 3' only plays better or worse than Lenderman with 5' + 3".

At 3+0 I would bet on Lenderman (if I had to), but at 3+1 I would bet on Nakamura. I don't have a rationale for this, it is just my feeling. Maybe an explanation would be that 3+1 is about 4 minutes total per game, and 5+3 is about 8 minutes per game, that is, one doubling, so probably not much more than 100 Elo difference (even at those fast time controls), while rating difference between Nakamura and Lenderman is 200 points or so.

I have a theory, totally unsuported by any data, just by mu feeling and limited personal experience, that, for example, 8 vs. 4 minutes is a much bigger rating difference than 60 vs. 30. I believe that linear increase per time doubling, which is probably correct for computers, can not be appiled to humans without respect to particular time controls.

For that reason, I believe that 3+2 to 15+10 is about 300 Elo difference, and 15+10 to 90+30 is about 200.

I agree regarding humans, but I thinks this is also true for computers. It is pretty well established that doubling the time or speed or threads produces diminishing elo gains as the depth reached gets higher and higher, probably primarily due to increasing draw percentage. If you used unbalanced opening books like S. Pohl's UHO22, then this wouldn't be the case, each doubling would probably show a roughly similar elo gain. Maybe also for humans, but they never play with such books so we have no data.

[lkaufman]

It is possible that I misinterpreted what was going on.

In any case I think that the final number may be too low considering the small number of losses.

I checked it out a bit, and here is a more complete version of what happened. Hikaru started out the series of 40 sessions as a new player with a 1500 rating, and for the first few dozen games he was mostly played with players around or below 1000. He gave away the queen for just a pawn in those games, saying that he would switch to giving it away for a piece when the opponents were mostly above 1000, and perhaps to giving it for a rook when they were much higher, but of course this is very hard to do in the opening, so he actually aimed for piece and pawn against good players. Against the very weak players he won forty games to none giving up just one draw at queen for pawn. Once he got to 2300 the rating seemed to roughly stablize, though I agree it would probably climb a bit higher with enough games. Curiously at the start he estimated that he would reach 2300 at the end, amazing prediction! I think the only reason he doesn't do the same sort of thing with standard handicaps like knight or rook odds is that chess.com isn't set up for rated games with handicaps using the normal rating system, so he wouldn't have an easy way to determine what rating he might achieve at say knight handicap (he could play unrated games and someone could calculate the rating, but then they wouldn't be rated for the humans so they might not be fully motivated, though I imagine most would do their best anyway). Anyway at least we have some idea what chess.com blitz rating Hikaru gets at 3' + 0" giving queen for minor and pawn, and based on this I think he could earn about 2600 giving knight odds.

Of course the situation is dramatically different as the time control gets longer. Even 10 minute chess would be hugely different.