New Duel: Who will win: Rybka or Stockfish ?

[Sedat Canbaz]

Hello Chess Friends!

I am pleased to announce of organizing another exciting event:
-Very soon I plan to start a new duel: Rybka vs Stockfish

Note: Both participants will be played with different time controls

Rybka 4.1 x64's TC: 30 min + 1 sec
Stockfish 151014's TC: 1 min + 1 sec

Duel Conditions:
i5 3210M 2.50 GHz
Auto 232 Mode
Ponder OFF
1 core
512MB Hashtable size
Perfect 2015 book
Tablebases OFF

More Details:
-Minimum 20 games are planning for the current duel
-Stockfish will be played under Shredder Classic GUI
-Rybka will be played under Arena 2.0.1 GUI
*Note: Fritz GUI ponders always during Auto232 mode, that's why Arena GUI is planing...

According to CEGT (40/20):

Stockfish 5.0 x64 1CPU 3131 Elo
Deep Rybka 4.1 x64 1CPU 2972 Elo

Elo Difference: 159 Elo

---------------------------------

According to SCCT (3 min +2 sec):

Stockfish 151014 MP 3547 Elo
Rybka 4.1 MP 3350 Elo

Elo Difference: 197 Elo


I wonder a lot again...who will be the winner )?!
Rybka can gain from such TC advantage or not... ?!

Let's see...and all this unknown issue will be solved soon !!))

Many thanks in advance for your votes!


Best Regards,
Sedat Canbaz

[JJJ]

Gonna be closer this time.

[lkaufman]

[quote="Sedat Canbaz"]Hello Chess Friends!

I am pleased to announce of organizing another exciting event:
-Very soon I plan to start a new duel: Rybka vs Stockfish

Note: Both participants will be played with different time controls

Rybka 4.1 x64's TC: 30 min + 1 sec
Stockfish 151014's TC: 1 min + 1 sec

Duel Conditions:
i5 3210M 2.50 GHz
Auto 232 Mode
Ponder OFF
1 core

Rybka should win because I think double time is worth around 60 elo or so in the range in question, and the handicap is a bit below four doublings. I think it makes more sense when running these handicap matches to keep the ratio of base time to increment the same for both engines, so for example in this case instead of 30 min plus 1 second it could be 15 min plus 15 sec. First of all doing it this way makes the handicap perfectly clear. Also, the way you are doing it Rybka has a huge edge in the opening, while SF will have a huge edge in the endgame once the players have exhausted most of their base time. So the results will depend heavily on the nature of the openings chosen, whether they lead to quick fights or to long endgames. A uniform ratio avoids this problem.
Your matches are similar to the one reported on ChessBase where Komodo 8 beat Shredder (from 2006) by 5 to 1 effectively giving 50 to one time handicap by running Komodo on a cellphone with Shredder on an i7 with equal time. It's really amazing how much stronger both SF and Komodo are than the best engines of a few years ago. The claim made by some uninformed IM or GM that most progress in computer chess in recent years is from hardware is clearly refuted.

[Sedat Canbaz]

Current standings after 9 games: Rybka leads with + 2 = 6 - 1


.



Best,
Sedat

[Sedat Canbaz]

Rybka should win because I think double time is worth around 60 elo or so in the range in question, and the handicap is a bit below four doublings. I think it makes more sense when running these handicap matches to keep the ratio of base time to increment the same for both engines, so for example in this case instead of 30 min plus 1 second it could be 15 min plus 15 sec. First of all doing it this way makes the handicap perfectly clear. Also, the way you are doing it Rybka has a huge edge in the opening, while SF will have a huge edge in the endgame once the players have exhausted most of their base time. So the results will depend heavily on the nature of the openings chosen, whether they lead to quick fights or to long endgames. A uniform ratio avoids this problem.
Your matches are similar to the one reported on ChessBase where Komodo 8 beat Shredder (from 2006) by 5 to 1 effectively giving 50 to one time handicap by running Komodo on a cellphone with Shredder on an i7 with equal time. It's really amazing how much stronger both SF and Komodo are than the best engines of a few years ago. The claim made by some uninformed IM or GM that most progress in computer chess in recent years is from hardware is clearly refuted.

Thanks for your positive and useful comment dear Larry

Also I'd like to say thanks to Mr. Mark Lefler too,
he contacted several times to help me regarding K8 in SCCT

We need more Komodo teams !

About the Elo difference in doubling the time control,
I think it all it depends on used condition...before I was thinking approx. 80 Elo
but now I have different view...so my opinion is average 40 Elo
But to be sure we need more games...

Btw,
Later ...maybe after this event I can organize another interesting duel,
Komodo 2m+1s vs Stockfish 1m+1s, (100 games) what is your estimation, who will be the winner ?)


Best,
Sedat

[Sedat Canbaz]

Gonna be closer this time.

Yes...!

Both players are more equal each other under these conditions...

Best,
Sedat

[Sedat Canbaz]

New standings after 13 games, Rybka still leads: + 3, = 8, -2


Btw, here are a few testings more (played in Auto232 mode too)

1) Stockfish 1m+1s seems to be stronger than Rybka 7m+1s

.


2) Stockfish 1m+1s seems to be stronger than Rybka 15m+1s


.




Best,
Sedat

[lkaufman]

Rybka should win because I think double time is worth around 60 elo or so in the range in question, and the handicap is a bit below four doublings. I think it makes more sense when running these handicap matches to keep the ratio of base time to increment the same for both engines, so for example in this case instead of 30 min plus 1 second it could be 15 min plus 15 sec. First of all doing it this way makes the handicap perfectly clear. Also, the way you are doing it Rybka has a huge edge in the opening, while SF will have a huge edge in the endgame once the players have exhausted most of their base time. So the results will depend heavily on the nature of the openings chosen, whether they lead to quick fights or to long endgames. A uniform ratio avoids this problem.
Your matches are similar to the one reported on ChessBase where Komodo 8 beat Shredder (from 2006) by 5 to 1 effectively giving 50 to one time handicap by running Komodo on a cellphone with Shredder on an i7 with equal time. It's really amazing how much stronger both SF and Komodo are than the best engines of a few years ago. The claim made by some uninformed IM or GM that most progress in computer chess in recent years is from hardware is clearly refuted.

Thanks for your positive and useful comment dear Larry

Also I'd like to say thanks to Mr. Mark Lefler too,
he contacted several times to help me regarding K8 in SCCT

We need more Komodo teams !

About the Elo difference in doubling the time control,
I think it all it depends on used condition...before I was thinking approx. 80 Elo
but now I have different view...so my opinion is average 40 Elo
But to be sure we need more games...

Btw,
Later ...maybe after this event I can organize another interesting duel,
Komodo 2m+1s vs Stockfish 1m+1s, (100 games) what is your estimation, who will be the winner ?)


Best,
Sedat

Stockfish 5 is only 4 elo above Komodo 8 on the 1' + 1" "UFRL" list, and is much lower rated on the intermediate level lists. I suppose if SF has gained 20+ elo points since SF5 it might be a close match, since your proposal is about half a doubling in time, but if you played at 10' +10" for each it should be about an even match with no time difference. One minute plus 1" is just about the optimum level for SF vs Komodo.

[beram]

Rybka should win because I think double time is worth around 60 elo or so in the range in question, and the handicap is a bit below four doublings. I think it makes more sense when running these handicap matches to keep the ratio of base time to increment the same for both engines, so for example in this case instead of 30 min plus 1 second it could be 15 min plus 15 sec. First of all doing it this way makes the handicap perfectly clear. Also, the way you are doing it Rybka has a huge edge in the opening, while SF will have a huge edge in the endgame once the players have exhausted most of their base time. So the results will depend heavily on the nature of the openings chosen, whether they lead to quick fights or to long endgames. A uniform ratio avoids this problem.
Your matches are similar to the one reported on ChessBase where Komodo 8 beat Shredder (from 2006) by 5 to 1 effectively giving 50 to one time handicap by running Komodo on a cellphone with Shredder on an i7 with equal time. It's really amazing how much stronger both SF and Komodo are than the best engines of a few years ago. The claim made by some uninformed IM or GM that most progress in computer chess in recent years is from hardware is clearly refuted.

Thanks for your positive and useful comment dear Larry

Also I'd like to say thanks to Mr. Mark Lefler too,
he contacted several times to help me regarding K8 in SCCT

We need more Komodo teams !

About the Elo difference in doubling the time control,
I think it all it depends on used condition...before I was thinking approx. 80 Elo
but now I have different view...so my opinion is average 40 Elo
But to be sure we need more games...

Btw,
Later ...maybe after this event I can organize another interesting duel,
Komodo 2m+1s vs Stockfish 1m+1s, (100 games) what is your estimation, who will be the winner ?)


Best,
Sedat

Stockfish 5 is only 4 elo above Komodo 8 on the 1' + 1" "UFRL" list, and is much lower rated on the intermediate level lists. I suppose if SF has gained 20+ elo points since SF5 it might be a close match, since your proposal is about half a doubling in time, but if you played at 10' +10" for each it should be about an even match with no time difference. One minute plus 1" is just about the optimum level for SF vs Komodo.

I think it is more about 55% for latest SF dev.
See for instance this report of Adam Hair: http://www.talkchess.com/forum/viewtopi ... 22&t=53844

I have tested Komodo 8 (Drawscore=0) and Stockfish 140919 at 40 moves/3 minutes (equivalent to CCRL 40/4) on a couple of different computers (Intel E8400 and Intel Xeon L5420), in Windows XP 64-bit and Linux Mint 15, using both the Linux and Windows compiles (including via Wine), and using various openings formats (Sedat's Perfect 2014t, Lyudmil Tsvetkov's 2 move pgn, and a subset of the 8 move pgn used to create the TCEC openings). The cumulative score was:

Code: Select all

Code:

   # PLAYER                     : RATING  ERROR   POINTS  PLAYED    (%) 
   1 Stockfish 140919 64-bit    :   41.5    7.9   1985.0    3555   55.8% 
   2 Komodo 8 64-bit            :    0.0   ----   1570.0    3555   44.2%

[Sedat Canbaz]

.
Rybka wins the duel... !!



Finely,
I managed to find a right time control, where Stockfish is weaker than Rybka ))!!

For a better conclusion we need more games, but however in my opinion,
Rybka 20m+1s or 25m+1s will be approx. equal in Elo to Stockfish 1m+1s

Breaking News,
Another Duel is started: Komodo 2m+1s vs Stockfish 1m+1s (100 games match)

According to SCCT (3m+2s),

Code: Select all

  # PLAYER                 : RATING    POINTS  PLAYED    (%)
  1 Stockfish 151014 MP    : 3547.4     954.0    1300   73.4%
  2 Komodo 8 MP            : 3518.1    1172.0    1678   69.8%

Stockfish is approx. 29 Elo stronger than Komodo
Let's see this time who will be the winner ?!)

And soon I plan to publish the first played results and all games too!

Have fun,
Sedat