nTCEC simulation

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: nTCEC simulation

Post by Don »

And after the Stockfish win in round 51 we see that Stockfish as a good chance of qualifying for the next round.

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         44.948      99.052
Houdini        43.799      99.201
Bouquet         6.348      89.544
Critter         2.011      72.319
Hiarcs          1.112      57.998
Stockfish       0.807      61.694
Gull            0.575      53.054
Rybka           0.304      48.722
Naum            0.096      18.296
Junior          0.000       0.119

Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: nTCEC simulation

Post by Don »

After round 52 where Komodo draws against Gull:

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         45.018      99.068
Houdini        43.729      99.213
Bouquet         6.383      89.547
Critter         1.993      72.201
Hiarcs          1.118      58.062
Stockfish       0.797      61.655
Gull            0.575      53.104
Rybka           0.291      48.654
Naum            0.094      18.377
Junior          0.000       0.119

Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
User avatar
Laskos
Posts: 10948
Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos

Re: nTCEC simulation

Post by Laskos »

Don wrote:After round 52 where Komodo draws against Gull:

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         45.018      99.068
Houdini        43.729      99.213
Bouquet         6.383      89.547
Critter         1.993      72.201
Hiarcs          1.118      58.062
Stockfish       0.797      61.655
Gull            0.575      53.104
Rybka           0.291      48.654
Naum            0.094      18.377
Junior          0.000       0.119

Maybe would be somewhat better to compress a bit the ratings. The draw rate in this stage is now at 67%, you assumed a peak of 60% for strongest equal opponents. Just a gut feeling, but I think 89% for K and H to win the stage is a bit too high.
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: nTCEC simulation

Post by Don »

Laskos wrote:
Don wrote:After round 52 where Komodo draws against Gull:

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         45.018      99.068
Houdini        43.729      99.213
Bouquet         6.383      89.547
Critter         1.993      72.201
Hiarcs          1.118      58.062
Stockfish       0.797      61.655
Gull            0.575      53.104
Rybka           0.291      48.654
Naum            0.094      18.377
Junior          0.000       0.119

Maybe would be somewhat better to compress a bit the ratings. The draw rate in this stage is now at 67%, you assumed a peak of 60% for strongest equal opponents. Just a gut feeling, but I think 89% for K and H to win the stage is a bit too high.
I already compress the ratings by 80% - but the draw rate peaks at 60% which may be too low. But with such a horribly low sample of games it's difficult to assume anything.

Maybe we should check the draw rate in last seasons late stages, say the last 2 stages. What was the draw rate in the super-final?

I will check later.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
User avatar
Laskos
Posts: 10948
Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos

Re: nTCEC simulation

Post by Laskos »

Don wrote:
Laskos wrote:
Don wrote:After round 52 where Komodo draws against Gull:

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         45.018      99.068
Houdini        43.729      99.213
Bouquet         6.383      89.547
Critter         1.993      72.201
Hiarcs          1.118      58.062
Stockfish       0.797      61.655
Gull            0.575      53.104
Rybka           0.291      48.654
Naum            0.094      18.377
Junior          0.000       0.119

Maybe would be somewhat better to compress a bit the ratings. The draw rate in this stage is now at 67%, you assumed a peak of 60% for strongest equal opponents. Just a gut feeling, but I think 89% for K and H to win the stage is a bit too high.
I already compress the ratings by 80% - but the draw rate peaks at 60% which may be too low. But with such a horribly low sample of games it's difficult to assume anything.

Maybe we should check the draw rate in last seasons late stages, say the last 2 stages. What was the draw rate in the super-final?

I will check later.
38/48 ~ 79% draw rate in the Superfinal (season 1). Maybe you should adjust a bit the draw rate. And the rating compression maybe to 70%?
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: nTCEC simulation

Post by Don »

Simulation results after game 54, the Bouquet vs Houdini draw:

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         31.747      93.098
Houdini        26.162      91.380
Stockfish      10.343      78.600
Bouquet         9.574      77.290
Critter         8.135      73.859
Rybka           6.489      69.499
Gull            6.442      69.546
Hiarcs          0.535      21.638
Naum            0.533      21.487
Junior          0.040       3.602

Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: nTCEC simulation

Post by Don »

Note how Stockfish has somehow lifted itself up from it's bootstraps to prove the naysayer to be wrong!
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: nTCEC simulation

Post by Don »

Don wrote:Simulation results after game 54, the Bouquet vs Houdini draw:

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         31.747      93.098
Houdini        26.162      91.380
Stockfish      10.343      78.600
Bouquet         9.574      77.290
Critter         8.135      73.859
Rybka           6.489      69.499
Gull            6.442      69.546
Hiarcs          0.535      21.638
Naum            0.533      21.487
Junior          0.040       3.602

At this point, even Junior has a slight chance to qualify, but it's not much. Naum and Hiarcs definitely have their work cut out for them.
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.
User avatar
Laskos
Posts: 10948
Joined: Wed Jul 26, 2006 10:21 pm
Full name: Kai Laskos

Re: nTCEC simulation

Post by Laskos »

Don wrote:
Don wrote:Simulation results after game 54, the Bouquet vs Houdini draw:

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         31.747      93.098
Houdini        26.162      91.380
Stockfish      10.343      78.600
Bouquet         9.574      77.290
Critter         8.135      73.859
Rybka           6.489      69.499
Gull            6.442      69.546
Hiarcs          0.535      21.638
Naum            0.533      21.487
Junior          0.040       3.602

At this point, even Junior has a slight chance to qualify, but it's not much. Naum and Hiarcs definitely have their work cut out for them.
Very interesting, keep us updated.
User avatar
Don
Posts: 5106
Joined: Tue Apr 29, 2008 4:27 pm

Re: nTCEC simulation

Post by Don »

Laskos wrote:
Don wrote:
Don wrote:Simulation results after game 54, the Bouquet vs Houdini draw:

Code: Select all


Name         Win Odds     Stage 4
---------  ----------  ----------
Komodo         31.747      93.098
Houdini        26.162      91.380
Stockfish      10.343      78.600
Bouquet         9.574      77.290
Critter         8.135      73.859
Rybka           6.489      69.499
Gull            6.442      69.546
Hiarcs          0.535      21.638
Naum            0.533      21.487
Junior          0.040       3.602

At this point, even Junior has a slight chance to qualify, but it's not much. Naum and Hiarcs definitely have their work cut out for them.
Very interesting, keep us updated.
I asked Adam to give me a new draw model - which I will incorporate if/when he does.

Don
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.