ECO Classification of EPD records
Moderators: hgm, Rebel, chrisw
-
- Posts: 12541
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
ECO Classification of EPD records
Is there any utility that will give me the correct ECO classification for a collection of EPD records?
-
- Posts: 364
- Joined: Sun Oct 04, 2009 1:27 pm
- Location: Italy
Re: ECO Classification of EPD records
If I understand correctly, it should be a byproduct of Swami question on sts-x importances.
At the moment, the regression method has been applied to about thirty engines, with prediicted cegt elo error +- 50, more or less, but there is some instability. I'm waiting for other engines results, and a more well founded regression analysis .
Of course, you should run your own regression on your epd set.
At the moment, the regression method has been applied to about thirty engines, with prediicted cegt elo error +- 50, more or less, but there is some instability. I'm waiting for other engines results, and a more well founded regression analysis .
Of course, you should run your own regression on your epd set.
-
- Posts: 6401
- Joined: Thu Mar 09, 2006 8:30 pm
- Location: Chicago, Illinois, USA
Re: ECO Classification of EPD records
I wonder what happens if you transform it to pgn, load it with scid, and give the opening classification. I have not tried.Dann Corbit wrote:Is there any utility that will give me the correct ECO classification for a collection of EPD records?
Miguel
-
- Posts: 364
- Joined: Sun Oct 04, 2009 1:27 pm
- Location: Italy
Re: ECO Classification of EPD records
Sorry, I'm too focused on evaluating ELO performance by STS.
Beg your pardon for this misreading ... (I grabbed a cow by the balls ... )
Beg your pardon for this misreading ... (I grabbed a cow by the balls ... )
-
- Posts: 6640
- Joined: Thu Mar 09, 2006 4:21 am
Re: ECO Classification of EPD records
Hi Enrico,noctiferus wrote:If I understand correctly, it should be a byproduct of Swami question on sts-x importances.
At the moment, the regression method has been applied to about thirty engines, with prediicted cegt elo error +- 50, more or less, but there is some instability. I'm waiting for other engines results, and a more well founded regression analysis .
Of course, you should run your own regression on your epd set.
I have searched the ratings list and added these engines in the list of to be tested. I'd use 8 STS suites and append the results to the excel file. In future, I'll test them again for the partial credit scores once when the new version STSStat is released which I hope would come with mathematical calculations of Total/Average scores from all the test suites.
Filtered out engines that have established ratings in CEGT/CCRL:
Code: Select all
Frenzee
ETChess
Chiron 0.8.7
Greko
Arion 1.7
Little Thought 1.04
Gosu 0.16
Leila 0.53h
Homer 2.01
Crazy Bishop 0052
Queen 3.09
Chezzz 1.0.3
Bruja 1.9
Phalanx XXII
Amateur 2.8.2
Popochin 3.2
Waxman 2009
KnightX 1.92
EXchess 5.01
Thors Hammer 2.28
Neurosis 2.5
Averno 0.81
Comet B68
Feuerstein 0.4.6.1
Dragon 4.6
SOS
Chronos
Bright
Thinker
ChessMaster
Twisted Logic
Spike
StockFish
Fruit
Toga
Will run the test tonight, hopefully will be ready by monday.
-
- Posts: 364
- Joined: Sun Oct 04, 2009 1:27 pm
- Location: Italy
Re: ECO Classification of EPD records
Hi, Swami
Fine. May I suggest to specify which versions are you testing, in order to, if the case, countercheck the correspondence to ratings?
BTW - Which is the test time you are using: CEGT 40\4 for all? and for CCRL?
THX.
E.
Fine. May I suggest to specify which versions are you testing, in order to, if the case, countercheck the correspondence to ratings?
BTW - Which is the test time you are using: CEGT 40\4 for all? and for CCRL?
THX.
E.
-
- Posts: 6640
- Joined: Thu Mar 09, 2006 4:21 am
Re: ECO Classification of EPD records
Sure I'll specify in excel file which engine versions have been tested.noctiferus wrote:Hi, Swami
Fine. May I suggest to specify which versions are you testing, in order to, if the case, countercheck the correspondence to ratings?
BTW - Which is the test time you are using: CEGT 40\4 for all? and for CCRL?
THX.
E.
I used CEGT and CCRL 40 moves in 4 minutes list.
-
- Posts: 364
- Joined: Sun Oct 04, 2009 1:27 pm
- Location: Italy
Re: ECO Classification of EPD records
I asked it because at least in one case (leaving apart Bison), my results on Cerebro were definitely anomalous, performing on your test 100 elos more than CEGT. May be bad time management in games, or poor model, due to a short Elo range.
Anyway, when we'll see it , looking at a more spreaded elo range testsets, we should get a more stable model, I hope, and outlier detection could also help authors. May be I'm too optimistic..
Anyway, when we'll see it , looking at a more spreaded elo range testsets, we should get a more stable model, I hope, and outlier detection could also help authors. May be I'm too optimistic..
-
- Posts: 778
- Joined: Sat Jul 01, 2006 7:11 am
Re: ECO Classification of EPD records
I have been doing something similar, and have trouble with epd analysis in Arena with these enginesswami wrote:Hi Enrico,noctiferus wrote:If I understand correctly, it should be a byproduct of Swami question on sts-x importances.
At the moment, the regression method has been applied to about thirty engines, with prediicted cegt elo error +- 50, more or less, but there is some instability. I'm waiting for other engines results, and a more well founded regression analysis .
Of course, you should run your own regression on your epd set.
I have searched the ratings list and added these engines in the list of to be tested. I'd use 8 STS suites and append the results to the excel file. In future, I'll test them again for the partial credit scores once when the new version STSStat is released which I hope would come with mathematical calculations of Total/Average scores from all the test suites.
Filtered out engines that have established ratings in CEGT/CCRL:
Test datas of these engines would give better predictions of elo contributions of all the test suites.Code: Select all
Frenzee ETChess Chiron 0.8.7 Greko Arion 1.7 Little Thought 1.04 Gosu 0.16 Leila 0.53h Homer 2.01 Crazy Bishop 0052 Queen 3.09 Chezzz 1.0.3 Bruja 1.9 Phalanx XXII Amateur 2.8.2 Popochin 3.2 Waxman 2009 KnightX 1.92 EXchess 5.01 Thors Hammer 2.28 Neurosis 2.5 Averno 0.81 Comet B68 Feuerstein 0.4.6.1 Dragon 4.6 SOS Chronos Bright Thinker ChessMaster Twisted Logic Spike StockFish Fruit Toga
Will run the test tonight, hopefully will be ready by monday.
Code: Select all
Amateur 2.8.2
Chronos
Crazy Bishop 0052
Gosu 0.16
-
- Posts: 6640
- Joined: Thu Mar 09, 2006 4:21 am
Re: ECO Classification of EPD records
Yes, I just installed them and only some of them actually support analysis mode in Arena. A few of them don't.jwes wrote:I have been doing something similar, and have trouble with epd analysis in Arena with these enginesswami wrote:Hi Enrico,noctiferus wrote:If I understand correctly, it should be a byproduct of Swami question on sts-x importances.
At the moment, the regression method has been applied to about thirty engines, with prediicted cegt elo error +- 50, more or less, but there is some instability. I'm waiting for other engines results, and a more well founded regression analysis .
Of course, you should run your own regression on your epd set.
I have searched the ratings list and added these engines in the list of to be tested. I'd use 8 STS suites and append the results to the excel file. In future, I'll test them again for the partial credit scores once when the new version STSStat is released which I hope would come with mathematical calculations of Total/Average scores from all the test suites.
Filtered out engines that have established ratings in CEGT/CCRL:
Test datas of these engines would give better predictions of elo contributions of all the test suites.Code: Select all
Frenzee ETChess Chiron 0.8.7 Greko Arion 1.7 Little Thought 1.04 Gosu 0.16 Leila 0.53h Homer 2.01 Crazy Bishop 0052 Queen 3.09 Chezzz 1.0.3 Bruja 1.9 Phalanx XXII Amateur 2.8.2 Popochin 3.2 Waxman 2009 KnightX 1.92 EXchess 5.01 Thors Hammer 2.28 Neurosis 2.5 Averno 0.81 Comet B68 Feuerstein 0.4.6.1 Dragon 4.6 SOS Chronos Bright Thinker ChessMaster Twisted Logic Spike StockFish Fruit Toga
Will run the test tonight, hopefully will be ready by monday.
What time control are you running at?Code: Select all
Amateur 2.8.2 Chronos Crazy Bishop 0052 Gosu 0.16
These engines do work in Arena EPD Testing:
Code: Select all
Arion
Averno
Chezz
Comet
Exchess
Feuerstein
Homer
KnightX
Leila
Little Thought
TCB
Thorshammer
SOS
Bright
Thinker
ChessMaster
Twisted Logic
Spike
StockFish
Fruit
Toga