Next Komodo vs. GM handicap match - "Standard chess&quo

[Ozymandias]

Why there is no 3+ or 4 points? I voted 3 only..

Oops! The last entry was supposed to be 3.5 or more, but somehow it didn't make it, and I can't change it now. We'll just have to treat 3 as meaning 3 or more.

Well, that makes voting much easier
I'm actually going to watch these games for a change.

I think Komodo on a tablet or smartphone vs. a GM with no handicaps at all would be interesting.

Larry already considered using a cellphone.

[Modern Times]

1. Komodo will run on a single core of a $750 laptop computer, instead of the usual 24 core monster.

This is a fantastic idea. It places more emphasis on the intelligence of the program, the software itself, rather than brute force of huge hardware. I've always wanted to see one of these competitions run on more average hardware, although I expected that to possibly be an older quad. Kudos to you for giving the human a fighting chance with just one core

[Vinvin]

Nice setup, Larry !
BTW, will you set a contempt value ?

[Laskos]

Our next Komodo vs. GM handicap match is set for June 25 and 26 on chess.com, rounds at noon and 5 pm Eastern Daylight Time. The opponent is GM Sergey Erenburg, FIDE 2591, USCF 2676, born 1983. This will be the first match that starts from the standard position with no prescribed openings. But there are six handicaps, listed below:

1. Komodo will run on a single core of a $750 laptop computer, instead of the usual 24 core monster. On this laptop, Komodo 10 looks at about 1.36 million nps in the opening position.

Compared to 24 core fully enabled Komodo, rough estimates:

Basically modern day smartphone speed.
150 ELO points here.

2. Sergey will get the standard FIDE 90' + 30" increment (plus three 3 min breaks if requested), Komodo will get 3' + 1" increment. So standard chess vs. blitz, or 30 to 1 time odds.

300 ELO points

3. Ponder off, so Komodo can't think on his time. This makes it similar to Komodo playing a 30 board simul.
4. Sergey gets White in every game.

50 ELO points

5. No endgame tablebases for Komodo.

30 ELO points

6. No opening book for Komodo past move 3 (to provide variety). It's hard to say how much of a handicap this is. If Komodo plays a lot of book moves on its own, it may find itself in a bad position before Sergey even has to think, since many book lines are bad for Black.

60 ELO points - very speculative, hard to say.

Four game match, two games per day. Prize money of $300 per win and $100 per draw on top of base match fee.

I suppose Kai or someone else will do the math and conclude that Komodo is still the favorite, but the book handicap is of unknown magnitude.
Games will be played on a real chessboard, with the computer clock governing Komodo and the physical clock governing Erenburg. Commentary expected from chess.com.
A second match is planned for July 16 reverting to material handicaps, in this case two pawns. Details to be announced later.

All in all almost 600 ELO (notice, computer ELO) handicap compared to full Komodo. Means probably 400 human ELO points handicap. Assuming 3250 human ELO for 24 core full Komodo, it means around 2850 human ELO level for Komodo. Human is expected to score a bit less than 1 point , so I would bet on one draw, two draws would be an over-performance by human. A lot might depend on how the openings are handled.

[Ozymandias]

2. Sergey will get the standard FIDE 90' + 30" increment (plus three 3 min breaks if requested), Komodo will get 3' + 1" increment. So standard chess vs. blitz, or 30 to 1 time odds.

300 ELO points

3. Ponder off, so Komodo can't think on his time. This makes it similar to Komodo playing a 30 board simul.
4. Sergey gets White in every game.

50 ELO points

There's something conceptually wrong here, an engine playing at a TC of 3"+1' may be around 300 ELO weaker (than one using the standard FIDE TC), but if your opponent has 30 times more time, and you don't turn ponder off, the ELO hit is MUCH lower. Ponder off is therefore MUCH more relevant, than those 50 ELO you mention. The way you lay down your numbers, you're giving ponder influence over performance, but for players using symmetrical clocks. In fact, under the match conditions with asymmetrical clocks, the ELO loss for these two handicaps, should be equalised or even reversed.

[Laskos]

2. Sergey will get the standard FIDE 90' + 30" increment (plus three 3 min breaks if requested), Komodo will get 3' + 1" increment. So standard chess vs. blitz, or 30 to 1 time odds.

300 ELO points

3. Ponder off, so Komodo can't think on his time. This makes it similar to Komodo playing a 30 board simul.
4. Sergey gets White in every game.

50 ELO points

There's something conceptually wrong here, an engine playing at a TC of 3"+1' may be around 300 ELO weaker (than one using the standard FIDE TC), but if your opponent has 30 times more time, and you don't turn ponder off, the ELO hit is MUCH lower. Ponder off is therefore MUCH more relevant, than those 50 ELO you mention. The way you lay down your numbers, you're giving ponder influence over performance, but for players using symmetrical clocks. In fact, under the match conditions with asymmetrical clocks, the ELO loss for these two handicaps, should be reversed.

I just neglected ponder issue, as it is hardly a factor in normal games if used symmetrically in engine matches. Also, I always neglected ponder in these handicap matches, assuming them always as ponder off. As you put here the problem, sure "ponder on" would give Komodo a significant advantage compared to "ponder off". I just roughly estimated the strength of this Komodo in "normal for me" ponder off games against humans.

[lkaufman]

Can some of you who have Komodo 9 on your cellphones report here what phone and how mand NPS it gets in the opening position after 1 min? Some cellphones may not be that much inferior to single core of laptop, we can compare. I'm sure they are not 30 times slower, so your proposal would be even harder for GM than the actual terms.

[lkaufman]

Mark says that his cellphone gets about 80% of the NPS of my laptop set to single thread, and his cellphone is by no means the fastest on the market. So I think it's fair to say that this match is equivalent to running Komodo on a very good cellphone while still giving 30 to one time odds with no ponder.

[lkaufman]

Our next Komodo vs. GM handicap match is set for June 25 and 26 on chess.com, rounds at noon and 5 pm Eastern Daylight Time. The opponent is GM Sergey Erenburg, FIDE 2591, USCF 2676, born 1983. This will be the first match that starts from the standard position with no prescribed openings. But there are six handicaps, listed below:

1. Komodo will run on a single core of a $750 laptop computer, instead of the usual 24 core monster. On this laptop, Komodo 10 looks at about 1.36 million nps in the opening position.

Compared to 24 core fully enabled Komodo, rough estimates:

Basically modern day smartphone speed.
150 ELO points here.

2. Sergey will get the standard FIDE 90' + 30" increment (plus three 3 min breaks if requested), Komodo will get 3' + 1" increment. So standard chess vs. blitz, or 30 to 1 time odds.

300 ELO points

3. Ponder off, so Komodo can't think on his time. This makes it similar to Komodo playing a 30 board simul.
4. Sergey gets White in every game.

50 ELO points

5. No endgame tablebases for Komodo.

30 ELO points

6. No opening book for Komodo past move 3 (to provide variety). It's hard to say how much of a handicap this is. If Komodo plays a lot of book moves on its own, it may find itself in a bad position before Sergey even has to think, since many book lines are bad for Black.

60 ELO points - very speculative, hard to say.

Four game match, two games per day. Prize money of $300 per win and $100 per draw on top of base match fee.

I suppose Kai or someone else will do the math and conclude that Komodo is still the favorite, but the book handicap is of unknown magnitude.
Games will be played on a real chessboard, with the computer clock governing Komodo and the physical clock governing Erenburg. Commentary expected from chess.com.
A second match is planned for July 16 reverting to material handicaps, in this case two pawns. Details to be announced later.

All in all almost 600 ELO (notice, computer ELO) handicap compared to full Komodo. Means probably 400 human ELO points handicap. Assuming 3250 human ELO for 24 core full Komodo, it means around 2850 human ELO level for Komodo. Human is expected to score a bit less than 1 point , so I would bet on one draw, two draws would be an over-performance by human. A lot might depend on how the openings are handled.

I agree with your analysis except for the three move book part, which I believe is worth quite a bit more than 60 elo. If we make it 100 elo, this would make the conditions fair for World number 2 or 3 (Kramnik and Caruana) based on your analysis. So I'll tell Sergey that I think his chances are similar to what his chances would be against Caruana.
Another way to approach the problem is to start with the single-core ratings of Komodo 10 on the three best-known blitz lists (CCRL,CEGT, IPON) which are supposed to be rough estimates of what top engines would get in FIDE competition. Then there is no need to estimate the 24 core to 1 core dropoff. But I think the answer will be similar.

[Ozymandias]

I'll tell Sergey that I think his chances are similar to what his chances would be against Capuana.

He should certainly play accordingly.