Classification of Engines

[Han]

Geschrieben von: / Posted by: Han at 06 July 2000 17:21:35:
Is there anyone kind enough to pose a classification of the +-60 winboard engines into 3 categories of strength:
1000-1500, 1501-2000,2001-2600
and also it would be good if someone is able to rate the engines from a scale of 1 to 5, on the engine's positional play, and tactical play.

Tough request huh?!
Cheers to Winboard!
Han

[That Whinging Twit that g]

Geschrieben von: / Posted by: That Whinging Twit that glows pinkish purple: Dann Corbit at 06 July 2000 17:54:41:
Als Antwort auf: / As an answer to: Classification of Engines geschrieben von: / posted by: Han at 06 July 2000 17:21:35:

Is there anyone kind enough to pose a classification of the +-60 winboard engines into 3 categories of strength:
1000-1500, 1501-2000,2001-2600
and also it would be good if someone is able to rate the engines from a scale of 1 to 5, on the engine's positional play, and tactical play.
Tough request huh?!
Cheers to Winboard!

My BOC qualification listings include ELO at standard time controls. As you may or may not be aware, ELO figures are practically meaningless outside of their pool of competition. IOW, you cannot extrapolate these ELO figures to how they relate to human players. On the other hand, higher ELO generally indicates stronger play.
How about from 1 to 4 instead?
Extremely. That's because you did not tell us what time control you are interested in. More or less, your request is impossible. The programs behave differently at different time controls. Yace as a blitzer is a real tiger. It's an excellent engine at slower chess, but (so it seems) it excells even more at quicker time controls.
Welcome to the place where everybody knows your name.
Mine's "That Whinging Twit that glows pinkish purple."


My FTP site

[Aaron]

Geschrieben von: / Posted by: Aaron at 06 July 2000 18:05:29:
Als Antwort auf: / As an answer to: Re: Classification of Engines geschrieben von: / posted by: That Whinging Twit that glows pinkish purple: Dann Corbit at 06 July 2000 17:54:41:

Is there anyone kind enough to pose a classification of the +-60 winboard >>engines into 3 categories of strength:
1000-1500, 1501-2000,2001-2600
and also it would be good if someone is able to rate the engines from a scale >>of 1 to 5, on the engine's positional play, and tactical play.
Tough request huh?!

How about from 1 to 4 instead?
Extremely. That's because you did not tell us what time control you are >interested in.

At their best, i suspect about 75%-90% of the engines all fall within the 2001-2600 range. The one crown category of chessengs should be within 1500-2000. For engines as weak as 1000, well, i suppose, the programmers have to try quite hard to achieve that strenght..
It's interesting to note that the gap between a 1000 to 1500 is not as large as one between 1501 and 2000 which in terms is much smaller than that between 2001-2500..


It's Tough to rate positional play..It's quite subjective actually..


or what type of hardware you are using..

Well,for what's its worth i think for sure the 3 crowns and 4 crowns chess engines can be considered within the 2000-2600 class..Most of the 2 crowns are close to that as well..

[Claudio]

Geschrieben von: / Posted by: Claudio at 07 July 2000 00:53:11:
Als Antwort auf: / As an answer to: rating geschrieben von: / posted by: Aaron at 06 July 2000 18:05:29:

Is there anyone kind enough to pose a classification of the +-60 winboard >>engines into 3 categories of strength:
1000-1500, 1501-2000,2001-2600
and also it would be good if someone is able to rate the engines from a scale >>of 1 to 5, on the engine's positional play, and tactical play.
Tough request huh?!

How about from 1 to 4 instead?
Extremely. That's because you did not tell us what time control you are >interested in.

At their best, i suspect about 75%-90% of the engines all fall within the 2001-2600 range. The one crown category of chessengs should be within 1500-2000. For engines as weak as 1000, well, i suppose, the programmers have to try quite hard to achieve that strenght..
It's interesting to note that the gap between a 1000 to 1500 is not as large as one between 1501 and 2000 which in terms is much smaller than that between 2001-2500..
It's Tough to rate positional play..It's quite subjective actually..


or what type of hardware you are using..

Well,for what's its worth i think for sure the 3 crowns and 4 crowns chess engines can be considered within the 2000-2600 class..Most of the 2 crowns are close to that as well..

1000Elo engines... that's sure!
or the operating system. If you use Windows NT your engines go 3 times faster or more than if you use Windows 95/98. I suspect changing some parameters in the priority settings of the Windows 9x dos emulation can let you get closer to the NT performances. My chess engine averages 45kn/s on my p200MMX Windows NT, and 33kn/s on my father's pII 450 Windows 98.
I also ignore everything about Windows 2000...
Elo ratings are incremental. They hold as well as you consider score differences, not the exact value IMO. Anyway I guess ratings of engines below 2100 are a little overstimated. My impression is that a couple of my friends, who are rated 1850-1930elo in Italian FIGS, can easily beat 1900-2000 chess engines. Just an impression, I could be wrong..
Regards

[Aaron]

Geschrieben von: / Posted by: Aaron at 07 July 2000 05:34:54:
Als Antwort auf: / As an answer to: Re: rating geschrieben von: / posted by: Claudio at 07 July 2000 00:53:11:

It's interesting to note that the gap between a 1000 to 1500 is not as large >>as one between 1501 and 2000 which in terms is much smaller than that between >>2001-2500..
Well,for what's its worth i think for sure the 3 crowns and 4 crowns chess >>engines can be considered within the 2000-2600 class..Most of the 2 crowns >>are close to that as well..

Elo ratings are incremental. They hold as well as you consider score >differences, not the exact value IMO.
Anyway I guess ratings of engines below 2100 are a little overstimated. My >impression is that a couple of my friends, who are rated 1850-1930elo in >Italian FIGS, can easily beat 1900-2000 chess engines. Just an impression, I >could be wrong..

Well you are right. But it seems really hard to progress /improve as you go up the levels..


Probably. My feeling is that those chessengs rated around 2000 are occasionally vulnerable to Average to strong club players. Such engines are usually quite weak positionally playing moves that even i believe 1500ish players won't consider, eg moving pieces twice in the opening, weakening pawn structures for no reason etc..
But their tactical skill more than makes up for it..I have many games, where onmce out of book, the 2000ish engine starts to play weird moves, and I start to gain the advntage (+2 or +3) , but usually I get cocky, and sooner or later, one tactical shot, evens or wins the game for the program..
But in my defence, i usually play at time limits 2 12,and most of the games are played half heartedly. In fact, I take rated FICS games even more seriously (and those are not very serious ), which tells you how serious the pratice computer games are..
I'm currently 1800 standard (varies from 1800 to 2000) and a lousy 1500 Blitz..at FICS
Your friends, are stronger than me..so I suspect, someone who is sharper in tactics then me can probably beat up such engines ..Since most engines have small books, it's quite easy to outbook them, eg It's hard for even Crafty to play the Two knight's defence, Max Lange attack or Evans without a good opening book..
At least 50% of the time..

[Dann Corbit]

Geschrieben von: / Posted by: Dann Corbit at 07 July 2000 06:40:57:
Als Antwort auf: / As an answer to: Re: rating geschrieben von: / posted by: Aaron at 07 July 2000 05:34:54:

It's interesting to note that the gap between a 1000 to 1500 is not as large >>as one between 1501 and 2000 which in terms is much smaller than that between >>2001-2500..
Well,for what's its worth i think for sure the 3 crowns and 4 crowns chess >>engines can be considered within the 2000-2600 class..Most of the 2 crowns >>are close to that as well..

Elo ratings are incremental. They hold as well as you consider score >differences, not the exact value IMO.
Anyway I guess ratings of engines below 2100 are a little overstimated. My >impression is that a couple of my friends, who are rated 1850-1930elo in >Italian FIGS, can easily beat 1900-2000 chess engines. Just an impression, I >could be wrong..

Well you are right. But it seems really hard to progress /improve as you go up the levels..

Probably. My feeling is that those chessengs rated around 2000 are occasionally vulnerable to Average to strong club players. Such engines are usually quite weak positionally playing moves that even i believe 1500ish players won't consider, eg moving pieces twice in the opening, weakening pawn structures for no reason etc..
But their tactical skill more than makes up for it..I have many games, where onmce out of book, the 2000ish engine starts to play weird moves, and I start to gain the advntage (+2 or +3) , but usually I get cocky, and sooner or later, one tactical shot, evens or wins the game for the program..
But in my defence, i usually play at time limits 2 12,and most of the games are played half heartedly. In fact, I take rated FICS games even more seriously (and those are not very serious ), which tells you how serious the pratice computer games are..
I'm currently 1800 standard (varies from 1800 to 2000) and a lousy 1500 Blitz..at FICS
Your friends, are stronger than me..so I suspect, someone who is sharper in tactics then me can probably beat up such engines ..Since most engines have small books, it's quite easy to outbook them, eg It's hard for even Crafty to play the Two knight's defence, Max Lange attack or Evans without a good opening book..
At least 50% of the time..

Your feelings are right for two reasons. First, the ELO in a pool of computers has absolutely no connection to the ELO in a pool of human players. ELO ratings are calculated and have meaning only in connection to the rated competition. If you try to extrapolate outside those bounds, all bets are off.
Second, with an ELO difference of 200 points, you will definitely see upsets.
Win expectency for a difference of 200 points is 0.240253 which means that the 200 lower rated opponent will still get 1/4 of the points *even* if we are talking about two humans in a rated competition. [On average, of course -- it won't predict the outcome of any single experiment].


my ftp site

[Han]

Geschrieben von: / Posted by: Han at 07 July 2000 12:35:14:
Als Antwort auf: / As an answer to: Re: rating geschrieben von: / posted by: Dann Corbit at 07 July 2000 06:40:57:

It's interesting to note that the gap between a 1000 to 1500 is not as large >>as one between 1501 and 2000 which in terms is much smaller than that between >>2001-2500..
Well,for what's its worth i think for sure the 3 crowns and 4 crowns chess >>engines can be considered within the 2000-2600 class..Most of the 2 crowns >>are close to that as well..

Elo ratings are incremental. They hold as well as you consider score >differences, not the exact value IMO.
Anyway I guess ratings of engines below 2100 are a little overstimated. My >impression is that a couple of my friends, who are rated 1850-1930elo in >Italian FIGS, can easily beat 1900-2000 chess engines. Just an impression, I >could be wrong..

Well you are right. But it seems really hard to progress /improve as you go up the levels..

Probably. My feeling is that those chessengs rated around 2000 are occasionally vulnerable to Average to strong club players. Such engines are usually quite weak positionally playing moves that even i believe 1500ish players won't consider, eg moving pieces twice in the opening, weakening pawn structures for no reason etc..
But their tactical skill more than makes up for it..I have many games, where onmce out of book, the 2000ish engine starts to play weird moves, and I start to gain the advntage (+2 or +3) , but usually I get cocky, and sooner or later, one tactical shot, evens or wins the game for the program..
But in my defence, i usually play at time limits 2 12,and most of the games are played half heartedly. In fact, I take rated FICS games even more seriously (and those are not very serious ), which tells you how serious the pratice computer games are..
I'm currently 1800 standard (varies from 1800 to 2000) and a lousy 1500 Blitz..at FICS
Your friends, are stronger than me..so I suspect, someone who is sharper in tactics then me can probably beat up such engines ..Since most engines have small books, it's quite easy to outbook them, eg It's hard for even Crafty to play the Two knight's defence, Max Lange attack or Evans without a good opening book..
At least 50% of the time..

Your feelings are right for two reasons. First, the ELO in a pool of computers has absolutely no connection to the ELO in a pool of human players. ELO ratings are calculated and have meaning only in connection to the rated competition. If you try to extrapolate outside those bounds, all bets are off.
Second, with an ELO difference of 200 points, you will definitely see upsets.
Win expectency for a difference of 200 points is 0.240253 which means that the 200 lower rated opponent will still get 1/4 of the points *even* if we are talking about two humans in a rated competition. [On average, of course -- it won't predict the outcome of any single experiment].

I am not interested in relating the computer Elo to human Elo. I think they should be strictly kept apart. They are 2 different things altogether.
If you prefer to rate engines on a scale on 1 to 4, go ahead! Let the time setting be 2 hrs for all moves.