Lyudmil Tsvetkov wrote:You gave Komodo a looot of time.
At 1"+50ms komodo was too strong so we lowered it to 1"+30ms.
The point is, I used contempt and should not have used contempt. Larry wrote me, "Contempt is not needed if yiou are playing normal chess with a large enough time handicap (or depth) to be about even."
It is not clear how much time the engine was getting. There were 2-4ms of network delay to connect to the Pentium III and more latency from the machine running the GUI, and more delay for the old machine to start.
Jesse Gersenson wrote:I ran an informal match between Komodo 32-bit on a Pentium III (1.13 GHz notebook, 16mb hash, no tables bases or opening book) against a talented junior, rated 2100 two weeks ago.
We played several handicap matches with the human getting 5'+3" and the engine getting either 5'+3 (with a handicap) or game in 1" (without handicap). Odds included rook odds, rook and move, depth 2, queen for knight.
In the even material games, 5'+3" vs 1"+30ms, the kid was able to score draws if contempt was set incorrectly -- it should have been set to contempt = 0 but I'd set it to contempt = 260.
You gave Komodo a looot of time.
What do you think the outcome would be if you give Nakamura 2' + 2'' for Komodo vs 5' + 3' out of 10 games?
Lyudmil Tsvetkov wrote:This afternoon I tried to play a couple of games against Komodo with only time handicap. Unfortunately, as I seemingly am not in a good form at all, I fared rather poor.
Conditions: latest Komodo 10.1, kindly provided by Komodo team, strictly keeping TC, engine output hidden, 4 cores with hyperthreading off.
At 2' + 2" for both I managed roughly 1-2 draws from 10 games, throwing away plentiful winning positions with seconds to go.
At 2' + 2'' for Komodo vs 5' + 3'' for me, roughly double TC, still no wins, 3 or 4 draws from 10 games.
At 2' + 2'' vs 10'+ 10'', 5-fold time handicap, I already scored a single win, and drew 3 or 4 out of 10 games.
Do not have time for trying longer TC, but obviously with more time humans start to fare better.
I think I can do much better than that, but today was a slump day for me. And looking constantly at the clock drives me crazy. That makes a big difference: I am missing moves, that otherwise, if not under pressure, would not have missed.
Below my single win, it is a pity that I threw away much more spectacular positions, where I am certain Komodo was already showing +500-600cps advantage.
The search depths you show in the opening seem pretty low for a quadcore machine playing 2' + 2"; they are a couple ply less than my laptop gets. How many nodes per second does it show in infinite mode in the opening position?
Lyudmil Tsvetkov wrote:This afternoon I tried to play a couple of games against Komodo with only time handicap. Unfortunately, as I seemingly am not in a good form at all, I fared rather poor.
Conditions: latest Komodo 10.1, kindly provided by Komodo team, strictly keeping TC, engine output hidden, 4 cores with hyperthreading off.
At 2' + 2" for both I managed roughly 1-2 draws from 10 games, throwing away plentiful winning positions with seconds to go.
At 2' + 2'' for Komodo vs 5' + 3'' for me, roughly double TC, still no wins, 3 or 4 draws from 10 games.
At 2' + 2'' vs 10'+ 10'', 5-fold time handicap, I already scored a single win, and drew 3 or 4 out of 10 games.
Do not have time for trying longer TC, but obviously with more time humans start to fare better.
I think I can do much better than that, but today was a slump day for me. And looking constantly at the clock drives me crazy. That makes a big difference: I am missing moves, that otherwise, if not under pressure, would not have missed.
Below my single win, it is a pity that I threw away much more spectacular positions, where I am certain Komodo was already showing +500-600cps advantage.
The search depths you show in the opening seem pretty low for a quadcore machine playing 2' + 2"; they are a couple ply less than my laptop gets. How many nodes per second does it show in infinite mode in the opening position?
It's a dell xps laptop from 5 years ago.
Between 2233 and 2244 knps in the first 15 sec., reaching 2250 at the 1 minute mark.
Ponder is set to off, as at one point it suddenly became clear to me that getting 5-fold advantage with ponder on is not a very wise way to proceed.
Lyudmil Tsvetkov wrote:This afternoon I tried to play a couple of games against Komodo with only time handicap. Unfortunately, as I seemingly am not in a good form at all, I fared rather poor.
Conditions: latest Komodo 10.1, kindly provided by Komodo team, strictly keeping TC, engine output hidden, 4 cores with hyperthreading off.
At 2' + 2" for both I managed roughly 1-2 draws from 10 games, throwing away plentiful winning positions with seconds to go.
At 2' + 2'' for Komodo vs 5' + 3'' for me, roughly double TC, still no wins, 3 or 4 draws from 10 games.
At 2' + 2'' vs 10'+ 10'', 5-fold time handicap, I already scored a single win, and drew 3 or 4 out of 10 games.
Do not have time for trying longer TC, but obviously with more time humans start to fare better.
I think I can do much better than that, but today was a slump day for me. And looking constantly at the clock drives me crazy. That makes a big difference: I am missing moves, that otherwise, if not under pressure, would not have missed.
Below my single win, it is a pity that I threw away much more spectacular positions, where I am certain Komodo was already showing +500-600cps advantage.
The search depths you show in the opening seem pretty low for a quadcore machine playing 2' + 2"; they are a couple ply less than my laptop gets. How many nodes per second does it show in infinite mode in the opening position?
It's a dell xps laptop from 5 years ago.
Between 2233 and 2244 knps in the first 15 sec., reaching 2250 at the 1 minute mark.
Ponder is set to off, as at one point it suddenly became clear to me that getting 5-fold advantage with ponder on is not a very wise way to proceed.
Still considering 1.e4 e5 as best line of play.
Ok, my I7 laptop which is nearly 2 years old gets about 4100 knps, so that's another handicap. Still, even with five to one time odds, no ponder, White every game, no opening book, and slow hardware you did well to get some points with equal material. Maybe it's not too hard to make draws with White under these conditions if you pick the right openings, I haven't really researched this question.
Ok, my I7 laptop which is nearly 2 years old gets about 4100 knps, so that's another handicap. Still, even with five to one time odds, no ponder, White every game, no opening book, and slow hardware you did well to get some points with equal material. Maybe it's not too hard to make draws with White under these conditions if you pick the right openings, I haven't really researched this question.
Larry, come on.
At 2' + 2'' vs 10'+ 10'', 5-fold time handicap, I already scored a single win, and drew 3 or 4 out of 10 games.
In those conditions, it's a 2800-2900 human performance.
Lyudmil Tsvetkov wrote:This afternoon I tried to play a couple of games against Komodo with only time handicap. Unfortunately, as I seemingly am not in a good form at all, I fared rather poor.
Conditions: latest Komodo 10.1, kindly provided by Komodo team, strictly keeping TC, engine output hidden, 4 cores with hyperthreading off.
At 2' + 2" for both I managed roughly 1-2 draws from 10 games, throwing away plentiful winning positions with seconds to go.
At 2' + 2'' for Komodo vs 5' + 3'' for me, roughly double TC, still no wins, 3 or 4 draws from 10 games.
At 2' + 2'' vs 10'+ 10'', 5-fold time handicap, I already scored a single win, and drew 3 or 4 out of 10 games.
Do not have time for trying longer TC, but obviously with more time humans start to fare better.
I think I can do much better than that, but today was a slump day for me. And looking constantly at the clock drives me crazy. That makes a big difference: I am missing moves, that otherwise, if not under pressure, would not have missed.
Below my single win, it is a pity that I threw away much more spectacular positions, where I am certain Komodo was already showing +500-600cps advantage.
I don't believe you and I don't understand why some take you serious.
Ok, my I7 laptop which is nearly 2 years old gets about 4100 knps, so that's another handicap. Still, even with five to one time odds, no ponder, White every game, no opening book, and slow hardware you did well to get some points with equal material. Maybe it's not too hard to make draws with White under these conditions if you pick the right openings, I haven't really researched this question.
Larry, come on.
At 2' + 2'' vs 10'+ 10'', 5-fold time handicap, I already scored a single win, and drew 3 or 4 out of 10 games.
In those conditions, it's a 2800-2900 human performance.
Well, just by math, you are right, and this should be impossible. However we already know from the games posted by the player using the handle "father" that with enough trial and error, you can find openings that will result in a sufficiently locked position to get draws sometimes. In the present case, I'm inclined to believe that he has tried out enough openings to find the ones that Komodo (without a book and little time) will permit to reach closed positions, where it is easy enough to draw. Maybe I'm gullible, but I tend to believe people until I have a strong enough reason not to do so. If I ever have the time, I may try to find out myself whether I can find openings that allow me to close the position enough to make draws against a bookless Komodo playing Black.
Lyudmil Tsvetkov wrote:This afternoon I tried to play a couple of games against Komodo with only time handicap. Unfortunately, as I seemingly am not in a good form at all, I fared rather poor.
Conditions: latest Komodo 10.1, kindly provided by Komodo team, strictly keeping TC, engine output hidden, 4 cores with hyperthreading off.
At 2' + 2" for both I managed roughly 1-2 draws from 10 games, throwing away plentiful winning positions with seconds to go.
At 2' + 2'' for Komodo vs 5' + 3'' for me, roughly double TC, still no wins, 3 or 4 draws from 10 games.
At 2' + 2'' vs 10'+ 10'', 5-fold time handicap, I already scored a single win, and drew 3 or 4 out of 10 games.
Do not have time for trying longer TC, but obviously with more time humans start to fare better.
I think I can do much better than that, but today was a slump day for me. And looking constantly at the clock drives me crazy. That makes a big difference: I am missing moves, that otherwise, if not under pressure, would not have missed.
Below my single win, it is a pity that I threw away much more spectacular positions, where I am certain Komodo was already showing +500-600cps advantage.
The search depths you show in the opening seem pretty low for a quadcore machine playing 2' + 2"; they are a couple ply less than my laptop gets. How many nodes per second does it show in infinite mode in the opening position?
It's a dell xps laptop from 5 years ago.
Between 2233 and 2244 knps in the first 15 sec., reaching 2250 at the 1 minute mark.
Ponder is set to off, as at one point it suddenly became clear to me that getting 5-fold advantage with ponder on is not a very wise way to proceed.
Still considering 1.e4 e5 as best line of play.
Komodo would spank you around like a little boy getting disciplined. Quit making stuff up. I don't care if you are a renowned computer chess specialist; are you sure that you are not ARB in disguise? Your claims certainly aren't that different from his. Why don't you take on Komodo 10.1 on my hardware? It's only 36 cores and 72 threads. Haha. I'll give you 60 minutes versus Komodo with one minute. I would be willing to bet that the only result that you MIGHT get here and there is a draw. Every claim men like these make always CONTROL their conditions. Let's take that control away.
There is very simple logic to be understood here. If super grandmasters cannot beat today's engines on cell phones (or even draw), literally, then you certainly aren't going to do anything remotely great here either. Openings mean nothing. Closed positions will lead to a draw AT BEST if you are lucky enough - and by lucky I mean LUCKY. Claiming that you can win is a joke. The days of 'Man versus Machine' are long dead - so clearly there is something that you are doing that is not adding up.
Ok, my I7 laptop which is nearly 2 years old gets about 4100 knps, so that's another handicap. Still, even with five to one time odds, no ponder, White every game, no opening book, and slow hardware you did well to get some points with equal material. Maybe it's not too hard to make draws with White under these conditions if you pick the right openings, I haven't really researched this question.
Larry, come on.
At 2' + 2'' vs 10'+ 10'', 5-fold time handicap, I already scored a single win, and drew 3 or 4 out of 10 games.
In those conditions, it's a 2800-2900 human performance.
Well, just by math, you are right, and this should be impossible. However we already know from the games posted by the player using the handle "father" that with enough trial and error, you can find openings that will result in a sufficiently locked position to get draws sometimes. In the present case, I'm inclined to believe that he has tried out enough openings to find the ones that Komodo (without a book and little time) will permit to reach closed positions, where it is easy enough to draw. Maybe I'm gullible, but I tend to believe people until I have a strong enough reason not to do so. If I ever have the time, I may try to find out myself whether I can find openings that allow me to close the position enough to make draws against a bookless Komodo playing Black.
"Father"'s draw rate is below 5% IIRC, and he often posts won games on time (lag) from lost, but pretty locked positions, which resolve with 50 move rule or Contempt. I don't remember any single clear win of him in all these years. He still overperforms his rating, but by 200-300 ELO points, not 800. And the win claim... I would be surprised to see even Carlsen or Nakamura in these conditions getting a single win in a short match.
