Rating close to fide elo

[pedrox]

I edited my post.

[Ferdy]

Hi Ferdinand,

Could you play a few games against DanaSah with limit of strength?

I think you already have danaz04.zip (https://sites.google.com/site/danasah/s ... -danasah-z)

You should configure the danasah.txt file and put:

elo 2000
elo_type human

The engine should play as 2000 FIDE Elo points.

I used as reference the (Elo Aktiv + 70) as Elo FIDE.
http://www.schach-computer.info/wiki/in ... -Elo-Liste

Mephisto Amsterdam 68000 12 MHz 1932 = 2002 ELO FIDE.

This is the opinion that fans have to play against these machines of the years 80-90.

I'm not sure of the result.

I got this result in blitz TC 3m+2s.

Code: Select all

Score of CDrill 1800 vs DanaSahZ 0.4b 2000: 90 - 5 - 5 [0.925]
Elo difference: 436.43 +/- 135.38

100 of 100 games finished.

I use the WB exe with the following engine setting. The match is run under Cute Chess GUI.

Code: Select all

limitstrength true
elo 2000
elo_type human
style 3
diff_engine 0
diff_computer_engine 466
diff_human_computer 70
easylevel 0
ownbook false
tourbook ./book/none.bin
mainbook ./book/super.bin
hash 128
egbb_path 
egbb_men 4
egbb_cache_size 4
egbb_load_type 1
noise 0
castling KQkq

[pedrox]

Thanks for the test.

The difference is great and it surprises me although I could expect it.

I can modify a parameter of the engine to change the ELO and this way I do not need to modify the code source.

diff_computer_engine 466

I do not know the exact difference of Elo between the two versions that have played, but if I need for example about 400 more points of Elo could configure the parameter as:

diff_computer_engine 66

With this the 2 engines should play more or less with the same force.

[pedrox]

I will try to explain why I used the values ​​I had.

I made a list of Elo (with Hispanic engines) and more or less calibrated the list with CCRL values (Time control 2+2)

https://sites.google.com/site/motoresde ... lo-compleo

I included in the list a dedicated machine, Mephisto Roma32. This machine has an Aktiv Elo of 2076.

However, the machine only got on the list about 1600 points and this is roughly the strength of the Darky engine.

So I thought that the difference of Elo between a list like the Aktiv and another list like the CCRL could be about 466 (2076-1610) points and that is the value that I used in the parameter discussed in the previous post.

In a forum they faced the engine against Mephisto Roma32 (games of 15 minutes) and the force seemed to indicate about 2000 points.

For that reason I thought that the calibration could be good despite the huge difference between lists.

It would be great to be able to face the engines against humans with a contrasted Elo and really know their opinions.

I still have doubts if your engine plays too hard and mine loose. Perhaps one thing that greatly influences is time control.

You may be interested in seeing the following post:
http://foro.meca-web.es/viewtopic.php?f ... 7&start=10
(You can see the big difference between setting up danasah as an engine or as a computer)

[Ferdy]

I will try to explain why I used the values ​​I had.

I made a list of Elo (with Hispanic engines) and more or less calibrated the list with CCRL values (Time control 2+2)

https://sites.google.com/site/motoresde ... lo-compleo

I included in the list a dedicated machine, Mephisto Roma32. This machine has an Aktiv Elo of 2076.

However, the machine only got on the list about 1600 points and this is roughly the strength of the Darky engine.

So I thought that the difference of Elo between a list like the Aktiv and another list like the CCRL could be about 466 (2076-1610) points and that is the value that I used in the parameter discussed in the previous post.

In a forum they faced the engine against Mephisto Roma32 (games of 15 minutes) and the force seemed to indicate about 2000 points.

For that reason I thought that the calibration could be good despite the huge difference between lists.

It would be great to be able to face the engines against humans with a contrasted Elo and really know their opinions.

I still have doubts if your engine plays too hard and mine loose. Perhaps one thing that greatly influences is time control.

You may be interested in seeing the following post:
http://foro.meca-web.es/viewtopic.php?f ... 7&start=10
(You can see the big difference between setting up danasah as an engine or as a computer)

It is good to know the story behind your setting.

While looking a couple of games live, I can see that CDrill is using more time.
Calculating the average depth and time I get this,

Code: Select all

 nr                       player  games  pts%      sumTime aveDep    aveTime

  1                  CDrill 1800    100  92.5  00:06:33:04   7.08  00:05:279
  2           DanaSahZ 0.4b 2000    100   7.5  00:05:45:25   6.17  00:04:683

file         : danasahz.pgn
score window : [-1000.0, +1000.0]
notes        : 1. games and pts% are not affected by score window.
               2. Table is sorted by aveDep in descending order.
               3. aveTime is the average time/move in m:s:ms
               4. sumTime is in d:h:m:s

which verifies what I have seen.

I will read the link you give later.

[Ferdy]

I will try to explain why I used the values ​​I had.

I made a list of Elo (with Hispanic engines) and more or less calibrated the list with CCRL values (Time control 2+2)

https://sites.google.com/site/motoresde ... lo-compleo

I included in the list a dedicated machine, Mephisto Roma32. This machine has an Aktiv Elo of 2076.

However, the machine only got on the list about 1600 points and this is roughly the strength of the Darky engine.

So I thought that the difference of Elo between a list like the Aktiv and another list like the CCRL could be about 466 (2076-1610) points and that is the value that I used in the parameter discussed in the previous post.

In a forum they faced the engine against Mephisto Roma32 (games of 15 minutes) and the force seemed to indicate about 2000 points.

For that reason I thought that the calibration could be good despite the huge difference between lists.

It would be great to be able to face the engines against humans with a contrasted Elo and really know their opinions.

I still have doubts if your engine plays too hard and mine loose. Perhaps one thing that greatly influences is time control.

You may be interested in seeing the following post:
http://foro.meca-web.es/viewtopic.php?f ... 7&start=10
(You can see the big difference between setting up danasah as an engine or as a computer)

Do the rating in aktiv are based on human vs machine, or they also include machine vs machine?

[pedrox]

Elo Aktiv is a list of dedicated chess machines and they are operated by humans, the control time I think remember is 30 minutes for the game (with a time for the operator) or 30 seconds per move.

I think that to establish a starting point for the Elo was based on the old lists of the SSDF.

In some forum I have seen that they still do some tournaments between these machines and humans. Of these tournaments some concluded that if you add the Aktiv Elo about 70 points looks like FIDE ELO, but I do not know about the number of games or circumstances that led to that conclusion.

[Ferdy]

Elo Aktiv is a list of dedicated chess machines and they are operated by humans, the control time I think remember is 30 minutes for the game (with a time for the operator) or 30 seconds per move.

I think that to establish a starting point for the Elo was based on the old lists of the SSDF.

In some forum I have seen that they still do some tournaments between these machines and humans. Of these tournaments some concluded that if you add the Aktiv Elo about 70 points looks like FIDE ELO, but I do not know about the number of games or circumstances that led to that conclusion.

I also found this out.

Code: Select all

Mephisto Lyon 68030 36 MHz                  2151 ssdf new
Mephisto TM Lyon 68030 36 MHz               2251 ssdf old
Mephisto TM Lyon 68030 36 MHz               2297 activ

The new ssdf is lower by 100.

Later I will try myself to play Danasahz 0.4b fide elo 2000 the setting that I played vs CDrill 1800.
BTW in that Danasahz fide elo 2000 what hardware did you use?
I will be playing it on my hardware of 3.4 Ghz.

[pedrox]

Elo Aktiv is a list of dedicated chess machines and they are operated by humans, the control time I think remember is 30 minutes for the game (with a time for the operator) or 30 seconds per move.

I think that to establish a starting point for the Elo was based on the old lists of the SSDF.

In some forum I have seen that they still do some tournaments between these machines and humans. Of these tournaments some concluded that if you add the Aktiv Elo about 70 points looks like FIDE ELO, but I do not know about the number of games or circumstances that led to that conclusion.

I also found this out.

Code: Select all

Mephisto Lyon 68030 36 MHz                  2151 ssdf new
Mephisto TM Lyon 68030 36 MHz               2251 ssdf old
Mephisto TM Lyon 68030 36 MHz               2297 activ

The new ssdf is lower by 100.

Later I will try myself to play Danasahz 0.4b fide elo 2000 the setting that I played vs CDrill 1800.
BTW in that Danasahz fide elo 2000 what hardware did you use?
I will be playing it on my hardware of 3.4 Ghz.

When it says "TM" I think it refers to a special version of hardware used to play in tournaments and world championships (Tournament Machine)

I did the calibration on a computer that has 2.4 GHz, the same speed more or less that uses CCRL reference. Then, once the force has been calibrated, the engine should play the same, even if it plays on a computer or an android phone, since the force is mainly based on the nps.

In the test you did previously danasah played with openning book?

Code: Select all

limitstrength true 
elo 2000 
elo_type human 
style 3 
diff_engine 0 
diff_computer_engine 466 
diff_human_computer 70 
easylevel 0 
ownbook false 
tourbook ./book/none.bin 
mainbook ./book/super.bin 
hash 128 
egbb_path 
egbb_men 4 
egbb_cache_size 4 
egbb_load_type 1 
noise 0 
castling KQkq

To play with book should be ownbook true and then the engine would play a few book moves, I think remember it was elo/100 plys (20 plys or the first 10 moves)

[Ferdy]

Elo Aktiv is a list of dedicated chess machines and they are operated by humans, the control time I think remember is 30 minutes for the game (with a time for the operator) or 30 seconds per move.

I think that to establish a starting point for the Elo was based on the old lists of the SSDF.

In some forum I have seen that they still do some tournaments between these machines and humans. Of these tournaments some concluded that if you add the Aktiv Elo about 70 points looks like FIDE ELO, but I do not know about the number of games or circumstances that led to that conclusion.

I also found this out.

Code: Select all

Mephisto Lyon 68030 36 MHz                  2151 ssdf new
Mephisto TM Lyon 68030 36 MHz               2251 ssdf old
Mephisto TM Lyon 68030 36 MHz               2297 activ

The new ssdf is lower by 100.

Later I will try myself to play Danasahz 0.4b fide elo 2000 the setting that I played vs CDrill 1800.
BTW in that Danasahz fide elo 2000 what hardware did you use?
I will be playing it on my hardware of 3.4 Ghz.

When it says "TM" I think it refers to a special version of hardware used to play in tournaments and world championships (Tournament Machine)

I did the calibration on a computer that has 2.4 GHz, the same speed more or less that uses CCRL reference. Then, once the force has been calibrated, the engine should play the same, even if it plays on a computer or an android phone, since the force is mainly based on the nps.

In the test you did previously danasah played with openning book?

Code: Select all

limitstrength true 
elo 2000 
elo_type human 
style 3 
diff_engine 0 
diff_computer_engine 466 
diff_human_computer 70 
easylevel 0 
ownbook false 
tourbook ./book/none.bin 
mainbook ./book/super.bin 
hash 128 
egbb_path 
egbb_men 4 
egbb_cache_size 4 
egbb_load_type 1 
noise 0 
castling KQkq

To play with book should be ownbook true and then the engine would play a few book moves, I think remember it was elo/100 plys (20 plys or the first 10 moves)

Given:

Code: Select all

limitstrength true 
elo 2000 
elo_type human 
style 3 
diff_engine 0 
diff_computer_engine 466 
diff_human_computer 70 
easylevel 0 
ownbook false 
tourbook ./book/none.bin 
mainbook ./book/super.bin 
hash 128 
egbb_path 
egbb_men 4 
egbb_cache_size 4 
egbb_load_type 1 
noise 0 
castling KQkq

Then run it in computer with 2.4 Ghz and compare it with
computer with 3.4 Ghz, I believe the strength or force is stronger for 3.4 Ghz. So in this case elo 2000 is according to your reference of 2.4 Ghz. The question is what parameter should I change so that it plays elo 2000 on my 3.4 Ghz machine?

In danasah.txt I have this. I believe it is turned off.

Code: Select all

ownbook false

I tried to determine the difference between the old and new ssdf rating, so I guess they lowered it by 100 rating points overall.