Our dedicated chess computers in fact 300-350 elo weaker ??

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
User avatar
ricard60
Senior Member
Posts: 1285
Joined: Thu Aug 09, 2007 2:46 pm
Location: Puerto Ordaz

Post by ricard60 »

Larry wrote:I think you'll find the problem with rating the dedicateds is that only
the early games against the same person have a lot of meaning. Once
the owner figures out how to beat the comp, he can more or less go
ahead and beat at will, on any level, because he has long since found
holes in it's knowledge. It seems to me that the ratings are a fair
reflection of the strength of each comp.
If a 1500 elo player try to play lets say a machine of 2000 elo it will be a no easy task for him and if it finds a hole on its knowledge it will be for that game and in that oppening and for another game and another oppening it will be againg a not easy task and once that 1500 elo player beats that machine in all type of openings and with black and white for sure that player is no longer a 1500 elo player it will be at least a 2000 elo player or maybe more.
Larry
Senior Member
Posts: 2269
Joined: Wed Aug 01, 2007 8:42 am
Location: Gosford, NSW Australia

Post by Larry »

ricard60 wrote: once that 1500 elo player beats that machine in all type of openings and with black and white for sure that player is no longer a 1500 elo player it will be at least a 2000 elo player or maybe more.
Yes, but in practice that 1500elo player, once figuring out how to beat
the comp, does not then go on to find ways to beat it in any opening. If
he did, then yes, he would be elo2000.
But then, I've got a lazy streak and maybe I'm just assuming others
are the same.
L
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Larry wrote:
ricard60 wrote: once that 1500 elo player beats that machine in all type of openings and with black and white for sure that player is no longer a 1500 elo player it will be at least a 2000 elo player or maybe more.
Yes, but in practice that 1500elo player, once figuring out how to beat
the comp, does not then go on to find ways to beat it in any opening. If
he did, then yes, he would be elo2000.
But then, I've got a lazy streak and maybe I'm just assuming others
are the same.
L
Hi Larry,

You are right, a human has a tendency to keep playing over and over again repetitive moves against his home computer that he has had for 20 years until he wins. Then he can tell the world how much better he is. And of course the rest of the world applauds and devalues the computers playing skill. Because this person a 1600 or 1800 ELO player says "hey I beat my 2000 ELO computer all the time". And the rest of his 20 years of playing same computer buddies say "oh yeah I do that all the time too, lets devalue them all as the rating is wrong!!"

And that's how it goes.

Sadly what people seem to forget is that human's obtain ratings only on official over the table matches, in rated matches in their official rated events.

As you had observed before, put a curtain in front of the human so he doesn't know who he is playing and bring out an 1800 ELO computer to play an 1800 ELO human and then lets officially see what happens!

I have been doing some studying on online live games. It's funny how a lot of players around the 2200-2400 Bullet and Blitz range mysteriously have their ratings drop by about 400 ELO when they play a traditional standard length game. Pretty sure though these same guys tell everyone that they are 2400 ELO rated players. :P

I don't take any online human games or ratings seriously because with modern technology you just need an Ipad next to you and your computer and there you go, you become a Master. No one is the wiser. Human nature tells me that the temptation to want to hang around the top is just to great. Human's are tempted creatures. You don't have to play every computer move that would be too obvious. You just avoid playing the bad moves and voila! you have become a Master!

BTW these 2200-2400 players if you analyze their standard games, they make the same and in some cases more inaccuracies, mistakes, blunders and centipawns losses as the programs that I have playing in my tournament and in addition they resign early enough (especially after a mistake or blunder) which also avoids some major dents on their centipawn losses, whereas the games I play with our computers are almost always to the bitter conclusion.

Sorry but I just tend to be too suspicious by nature....regards
Nick
User avatar
ricard60
Senior Member
Posts: 1285
Joined: Thu Aug 09, 2007 2:46 pm
Location: Puerto Ordaz

Post by ricard60 »

spacious_mind wrote:
Larry wrote:
ricard60 wrote: once that 1500 elo player beats that machine in all type of openings and with black and white for sure that player is no longer a 1500 elo player it will be at least a 2000 elo player or maybe more.
Yes, but in practice that 1500elo player, once figuring out how to beat
the comp, does not then go on to find ways to beat it in any opening. If
he did, then yes, he would be elo2000.
But then, I've got a lazy streak and maybe I'm just assuming others
are the same.
L
Hi Larry,

You are right, a human has a tendency to keep playing over and over again repetitive moves against his home computer that he has had for 20 years until he wins. Then he can tell the world how much better he is. And of course the rest of the world applauds and devalues the computers playing skill. Because this person a 1600 or 1800 ELO player says "hey I beat my 2000 ELO computer all the time". And the rest of his 20 years of playing same computer buddies say "oh yeah I do that all the time too, lets devalue them all as the rating is wrong!!"

And that's how it goes.
If we talk about serious chess lets say we are going to study chess this is not true. We know that if a beat a 2000 elo computer lets say on a certain line of the spanish oppening over and over againg because i found a hole on this particular line on its knowledge i can not say that i am way over this machine. Been a 2000 elo player means you have to beat that machine in different kind of positions that comes up in different kind of oppenings and there maybe you can say lets devalue the elo machine or maybe you have begun a more skillful player to do this and you are no longer that 1500 elo player.

Getting skillful regards
Ricardo
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

ricard60 wrote: If we talk about serious chess lets say we are going to study chess this is not true. We know that if a beat a 2000 elo computer lets say on a certain line of the spanish oppening over and over againg because i found a hole on this particular line on its knowledge i can not say that i am way over this machine. Been a 2000 elo player means you have to beat that machine in different kind of positions that comes up in different kind of oppenings and there maybe you can say lets devalue the elo machine or maybe you have begun a more skillful player to do this and you are no longer that 1500 elo player.

Getting skillful regards
Ricardo
Yes that's true you have improved your skill. But, you haven't lowered your opponents rating right? The problem even then is that you have become familiar with just your one opponent so to prove that you have improved your skill, you would need to now play some other opponents. :) Preferably those that won't be lead into one of those 1 or two dozen openings that you have learned to perfect with old computer :) Same applies to middle and endgames you have gotten to know pretty well what move tendencies your old computer has in situations and you have learned to anticipate them better in your plan. But that same plan may not work with your next opponent :) But you are right, doing all this makes you a better player and you are improving to get closer to your computer level of play and someday even surpass it. You have to do that though with many opponents in order to reach a steady and consistent improved rating. Dedicated computers are definitely static unless you increase the amount of thinking time.

Well, you could also have a program that grows in strength with you :) As you get better it gets better! For example take Sargon 3 or Chris Whittington's Checkmate or Chess Player 2150, Chess Simulator, Chess Champion 2175, Chessmaster 2000, Chessmaster 2100, Colossus X and so on. You start playing it on an Amiga 500 where it plays like a 1600 ELO player. Then as you get better you play it at 68020 where maybe it plays at 1800. Then you progress and play it at 68040 where it plays at 2000... And then finally you play it at 68060 where it plays at 2200+. Now the computer has improved with you at your pace of progress. :)

You can also do the same with DOSBox :) or Portorose, Lyon, Vancouver and London :)

Best regards
Nick
User avatar
ricard60
Senior Member
Posts: 1285
Joined: Thu Aug 09, 2007 2:46 pm
Location: Puerto Ordaz

Post by ricard60 »

spacious_mind wrote:
ricard60 wrote: If we talk about serious chess lets say we are going to study chess this is not true. We know that if i beat a 2000 elo computer lets say on a certain line of the spanish oppening over and over againg because i found a hole on this particular line on its knowledge i can not say that i am way over this machine. Been a 2000 elo player means you have to beat that machine in different kind of positions that comes up in different kind of oppenings and there maybe you can say lets devalue the elo machine or maybe you have begun a more skillful player to do this and you are no longer that 1500 elo player.

Getting skillful regards
Ricardo
Yes that's true you have improved your skill. But, you haven't lowered your opponents rating right?


Yes that is right and that is why i do not agree on lowering 350 or whatever elo on the machines. I know for sure that if you beat this 2000 elo machine on different kind of openings that leads to different kind of middle and endgames for sure if then you go into a human U2000 tournament you will have a good tournament.

It is good to know that there are programs that they can get better as you can get better but when you do it only changing the hardware or increasing the speed the program gets better on tactics but does not improve its knowledge.

Hardware and software regards
Ricardo
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

ricard60 wrote:
It is good to know that there are programs that they can get better as you can get better but when you do it only changing the hardware or increasing the speed the program gets better on tactics but does not improve its knowledge.

Hardware and software regards
Ricardo
Yes, but it depends a little on how you interpret knowledge. If the extra speed stops it from previously playing a weaker move then this kind of knowledge is increased considerably. It also makes less mistakes in middle and endgames.

Best regards
Nick
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Time for a story.

We are in a fantasy world where 3 Species cohabitate and where all of the world's knowledge is captured in one big bucket known as the "Know-All" bucket. Since the Know-All bucket was originally made for Species 1 only, it grows in size very slowly by minuscule amounts that are hardly noticeable to the human eyes of Species 1 and Species 1 is not even aware of this fact. All the knowledge is measured and then distributed back to the 3 species everyday by an Abacus called Com-Pewter (yep Xanth).
Species 1 is a slow evolving species whose evolution is measured by years, centuries and millenniums. But, they think they are in charge of all and everything, well after all, rightly so since they think they have invented everything and they proudly think that they can prove it through years and years of carefully documented history written on parchments made from now extinct animals skins, carefully preserved in cool subterranean caves, in secret locations which only their leader the Supreme Grandmaster has access to.

Species 2 was created by Species 1 (its Master, its God) to become servants and slaves and cater for every whim that Species 1 can devise. Species 2 was already evolved at birth to the maximum albeit limited abilities. Knowledge carefully measured out of this Know-All bucket, is precisely distributed to everyone at birth in predestined measurements. This knowledge is administered by the Supreme Grandmaster’s chief accountant an Abacus that everyone calls Com-Pewter (yep, it’s a Xanth novel)

Species 3 was also created by Species 1. Species 3 is a fast evolving species whose evolution can be measured in days, weeks, months and years. Species 3 in the last few years has surpassed its Creator in knowledge and knowledge taken from the Know-All bucket. This was not as the Supreme Grandmaster had planned for the evolution for Species 3. "It just happened so" was all that he could say to the Grandmaster Senate when questioned. And this was an absolute truth as he had no idea what happened. Even Com-Pewter was totally clueless about how it happened. The complaints that were discussed in this Senate meeting were that more and more Species 1 people complained that Species 3 have gotten too smart for their boots and this needed to be addressed by the Senate. No one knew what happened, except one disgruntled Species 2 servant who got tired of being told where to go and what to do all the time, who had a moment of temper tantrums and accidently switched the red potion with the green potion that were used to create the first Species 3 brain. Well, this Species 2 servant thinks it was his fault but he also is not really sure since a Species 2 temper tantrum lasts only about one seventy fifth of a millisecond. Too fast really, for anyone to know for sure. So he is going to keep his mouth shut and not tell anyone what he suspects might have happened. Worse still, no one even knew that the knowledge taken from the Know-All bucket needed for Species 3 was beginning to dumb down the other two species. Except for Com-Pewter who was not ready to tell yet….

It took just 2 decades for Species 3 to catch up to its Creator and 1 decade to surpass its Creator as a result of its mysterious ability to increase its learning speed with a velocity that now exceeds Warp speed. There is no slowdown of acceleration in sight any time soon. So the Supreme Grandmaster had another Species 3 problem that he didn’t know he had yet.
A former Supreme Grandmaster several years prior to the birth of Species 2 commissioned the creation of an Abacus which he named Com-Pewter (sorry Mr Piers Anthony :) ) that could measure quite accurately the ability of each member of its specie and cast them according to a new cast system from Absolute Novice (the idiot) to the Supreme Grandmaster (savant). This was a fiendishly clever idea! This meant that the knowledge that was shared from the all-encompassing "Know-All" bucket could now be distributed precisely by using this cast system. The cast system was an idea that was originally used by a now extinct foreign culture and adopted by this former Supreme Grandmaster and used by every Supreme Grandmaster since then. The idea was if the Supreme Grandmaster did not like someone, they got less knowledge from the "Know-All" bucket and if he really liked someone like his billiard playing buddies and many girlfriends, well then they just got more! And when he got tired of a girlfriend (he mostly preferred blondes) well then he simply ordered Com-Pewter to do a redistribution of knowledge from the Know-All bucket. And of course being the Supreme Grandmaster he always got the most knowledge from the "Know-All" bucket. So this more or less system was built into Com-Pewter in order to manage the Know-All bucket. It was brilliant, the system worked perfectly until now.....
Com-Pewter moves on roller blades and talks. No one really knows how it talks and why it talks and no one really cares since this is the world of Xanth and no one really questions anything. Spread across the roller blades is this wooden framework with lots of strings and even more lots of stone pebbles hanging on the strings. No one really knows how all this works but this is how Com-Pewter does all his compewting for the distribution of knowledge to all the species from the Know-All bucket. Any way no one really cares about the how's and why's, it just is.

When Species 2 was created as slaves for Species 1 everything worked just fine. Com-Pewter had enough pebbles to calculate around with to account for the various casts in both species. It was simple really as all Com-Pewter had to do was to dumb down everyone in Species 1 by just a little to give Species 2 just enough knowledge from the finite supply contained in the Know-all bucket. So Species 2 could do all the slave work and no one would really notice as Species 1 would have more play time would not need to think as much anyway. Com-Pewter who is a fiendish little schemer carefully avoided removing knowledge from the Supreme Grandmaster and his Grandmaster Senate as he really did not want his boss the Supreme Grandmaster to get mad with him. So anyway for many years everything worked just perfectly, Supreme Grandmasters changed and no one knew and no one noticed the gradual dumbing down of the original species and their slaves.

But... as happens to all fiendishly good plans, they have a way of going wrong... Com-Pewter's predicament and nemesis became Species 3. For many months and years, Com-Pewter spent nights deliberating on what to do. He couldn't dumb down Species 1 any further as there would be more and more idiot distributions by the day, by the hour, by the minute, by the second emanating from the finite supply available in the Know-All bucket. Eventually Species 1 as well as Species 2 would become complete utter idiots leaving Species 3 to fight amongst themselves for bigger and bigger shares of knowledge from this finite Know-All bucket.

So finally one day, Com-Pewter decides that he has no choice but to tell all this to the Supreme Grandmaster, fess up and provide some options for solutions. Here are the choices that Com-Pewter offered to the Supreme Grandmaster:

1) Sorry Supreme Grandmaster. It is not my fault as your predecessor did not give me enough string and pebbles to calculated properly the changes needed as a result of the existence of Species 2 and Species 3 (blaming a previous Supreme Grandmaster is of course perfectly acceptable in Xanth) I must inform, I have been dumbing down the Species 1 population recently (no need to elaborate to the Supreme Grandmaster as to how many years this has been going on) in order to provide the other species their much needed knowledge to do their tasks and duties. Now I have no choice but to inform you that I am forced to have to dumb you down as well by 350 knowledge points. Therefore I am afraid as of tomorrow your total knowledge intake will be 2500. Also starting next week, I am going to have to remove an additional 1 point every single day for 12 months and next year increase this to two points. So based on my calculations in 4 years’ time you will become a complete and total idiot and for everyone below you I have no word in my Xanth Com-Pewter dictionary as to what to call them because their intelligence will move into less than zero knowledge calculations. Species 2 will also no longer function, so you must learn to bathe and feed yourself and that is going to become more difficult for you to do as you will be becoming dumber each day. Oh and by the way I quit! I have been offered a job by Species 3.

2) Supreme Grandmaster, we all live in this magical world of Xanth, and I am pretty sure that you don't like option 1, so why don't we immediately magically invent a much much much bigger "Know-All" knowledge bucket where you can keep your much needed Super Grandmaster knowledge of 2850 points that I distribute to you every day. You can magically make this new Know-All bucket so huge that it will cater for all our species needs for many years to come. And I can go back to correctly distributing the knowledge to all the people in Species 1 and of course our Species 2 servants and slaves! And as for Species 3, why not let them evolve to whatever they want to evolve to and I will manage the distribution accordingly and everyone is happy!

Well! Xanth’s Supreme Grandmaster is not a stupid man when it boils down to self-preservation, he liked his privileges and of course he had noticed that his recent batch of girlfriends had this vacant look in their eyes when he tried to explain this and that to them. He hated to admit it but it had been making him more and more frustrated in dinner conversations. So in Xanth the decision was easy to make. He snapped his fingers... poof..

Everything went back to normal in Xanth. The Know-All bucket became gigantic. Com-Pewter received a brand new set of roller blades and many more strings and pebbles to manage the knowledge distributions correctly for all 3 species. Conversations between the different casts of Species 1 became more intelligent. Species 2 now loves their work and status.... and Species 3 evolved to the stars......

The End......

Oh! And now back to earth... I am just wondering.... hmmm.. Still wondering what Earth will do?
Nick
User avatar
ricard60
Senior Member
Posts: 1285
Joined: Thu Aug 09, 2007 2:46 pm
Location: Puerto Ordaz

Post by ricard60 »

spacious_mind wrote:
ricard60 wrote:
It is good to know that there are programs that they can get better as you can get better but when you do it only changing the hardware or increasing the speed the program gets better on tactics but does not improve its knowledge.

Hardware and software regards
Ricardo
Yes, but it depends a little on how you interpret knowledge. If the extra speed stops it from previously playing a weaker move then this kind of knowledge is increased considerably. It also makes less mistakes in middle and endgames.

Best regards
This in part is true but not all of it. Look at this position:

[fen]8/8/3k4/8/8/3K4/8/2BN4_w_-_-[/fen]

This is a well known ending, surprisingly some 2000 elo machines does not know how to win in this position and no matter how much you increase their speed or improve their hardware they will not win. But there is a low elo machine U1500 that has the knowledge to win in this position. It was the second world dedicated chess computer champion the SciSys Chess Champion Mark V and it does it really quick.

Sometimes is just software knowledge and not hardware regards
Ricardo
User avatar
Scally
Full Member
Posts: 970
Joined: Sun Jun 05, 2016 5:44 pm
Location: Bermondsey, London
Contact:

Post by Scally »

Hi Ricardo,

In the recent High Performer Chess Computer Tournament, the Mephisto Montreux rated at 2259 failed to win a KBN v K ending against the Rev2 London.

Al.
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Scally wrote:Hi Ricardo,

In the recent High Performer Chess Computer Tournament, the Mephisto Montreux rated at 2259 failed to win a KBN v K ending against the Rev2 London.

Al.
You'd be surprised how many high rated humans would struggle with the same if they had not learned the principle and practiced it out of a book or in a club enough to understand the method. Over a chess board with a clock ticking without that prior book learning they would be just as useless if they had not remembered the principles or learned them at all. Therefore a draw in this situation does not really devalue that humans overall rating and neither does it a computers.

It's just another example of a missed opportunity for a win..... sounds very humanlike to me :)

Oh...btw... experienced dedicated computer players know these gaps in computers and they reflect this towards tending to let the computers play to the very end in order to obtain a conclusive result. Human's perhaps out of respect of the other human tend to resign. Therefore it is rare that a possible incompetence is put to a test. But if you search the record books through history you will find good examples of grandmaster's struggling in certain endgame situations and KNB v K is one of them.

Therefore if you wish a computers rating is actually more accurate than a humans because humans often resign too quickly giving their opponents a lot of cheap points. Many even resign as soon as they lose a pawn out of disgust at themselves. Quitters!
Nick
User avatar
spacious_mind
Senior Member
Posts: 4000
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

ricard60 wrote: Sometimes is just software knowledge and not hardware regards
Ricardo
It is not really true for your example. Let's assume the search algorithm for your 2000 ELO computer is programmed correctly, maybe it is not. But let's assume it is.

Without looking at Chessbase to see exactly what the fastest checkmate solution is, accept my guess of say 30 Full moves or 60 Half moves. Your 2000 computer in the above position has no chance of finding it unless someone to assist it has programmed into it the move sequences needed to trap the King in the correct corners to be checkmated. Therefore the programmer would have to program if "If White Bishop then do this" or "Black Bishop then do that". So that is the knowledge that you are talking about as your expectation for more knowledge.

Well lets put the same program and assume it has a perfect search engine into a superfast fast computer. Make this an imaginary computer if it doesn't already exist today that can calculate (without endgame tablebases) 80 ply in 10 seconds. Well don't you think it will find the right moves immediately without having to program the extra stuff that was programmed in some programs in the past to compensate for its inability to search deep enough?

In the above example finding checkmate in 60 ply should be pretty easy and quite fast as there are only 4 pieces on the board.

What I am saying is that the extra width search knowledge that you are expecting as a solution today for to compensate for depth knowledge will become more and more obsolete with time.

So move 20 years into the future and lets assume that chess is still not been cracked where a computer can solve the whole game starting at move 1. Search depth in the next 20 years gets fine tuned to make it even more precise than say Komodo 11.01 64Bit 4 CPU has today. Let's give this extra fine tuned search program a name. Let's call it Ricardo! :) Ricardo! has nothing other than it is a perfect depth searcher, so good even that if hardware were fast enough it would actually solve the game of chess from start to finish!

Anyway it plays Komodo and Komodo loses 100-0. It can't even get a draw its not good enough. In the eyes of the future human Komodo is actually crap at chess! ;)

So Komodo is now an imbecile, an absolute beginner right?.... since its rating in this 100 game match would be what? A 1000 ELO? It doesn't matter what evaluation you use today. Komodo is crap it has 1000 ELO...
:P

Best regards
Nick
User avatar
scandien
Member
Posts: 206
Joined: Mon Sep 12, 2011 1:15 pm
Contact:

Post by scandien »

Hello,
this is an interesting topics, but it is difficult to answer corectly. i have perform some test on strength and weakness of computer chess, and the machine are quite unbalanced.

So even if the overall level of play indicated by the rating list are relevant with result in man-machine match ( and machine machine match as well), you can always found computer weakness to say that the machine is not at expert level ( or any other level indicated bu the rating list).

Let me explain :
For the MEPHISTO NIGEL SHORT the overall level of play is EXPERT LEVEL ( between 2000 and 2200 FIDE ELO equivalent). If you play a match with Expert player you will get a balanced match ( specially if the player are not aware that they are playing an old dedicated computer chess).
The Khelmnitsky test show that this 'EXPERT' level is more or less relavant but the test show too that :

During the opening stage the machine plays as a grandmaster.During the other phase of the game, the MEPHISTO NIGEL SHORT is playing as a EXPERT player.

The strength of the machine rely mostly in tactical skill ( MASTER level ) rather than in strategic skill ( Class D player Level) . This mean that a player should take care about tactical threat during the game, and should try to exploit positional weakness of the machine.

The NIGEL SHORT is really strong in DEFENSE ( International Master level), Threat Detection ( International Master Level) and in counter attack (EXPERT LEVEL). This mean that a player will have to think really before launching his attack because he may suffer from Nigel Short Skill.

The Nigel Short ( as many other machine) is rather weak in ATTACK and SACRIFICE (Class D player for both skills). This mean that if it is difficult for the NIGEL SHORT to identify weakness in opponent position and to mobilize his pieces to launch a long term attack on the identified weakness. As well sacrifice to get a positional advantage will not be considered by the machine.

An EXpert level player wil consider that the weakness in ATTACK disqualify the machine for the EXPERT LEVEL , because, as an Expert , it should be able to plan such an attack!

I have the feeling that in tournament the awesome skill in tactics , Defense and counter attack balanced the weakness of the Nigel Short , and Expert player will face problem in front of this machine.



The Nigel Short have approximately the same profile ( the Nigel Short is better indeed) and the same remarks can be done.
The MM V is very difficult to attack ( very strong in defense, great skill to detect threats, master like level in tactics), and a human will always take the risk to suffer a counter attack if he try to defeat the machine ( the MMV is EXPERT Level in counter attack).
If he doesn't launch an attack ( and if he is good enough to be able to keep a good position without weakness) he will probably be able to hold the draw or even win in the end game. But i am quite sure that an average club player will soon or less do an error, and be crush due to tactical strength of the machine.


I can resume like this :
The machine are really able to perform at the level indicated in the list, but Most of them have weakness in long term attack, and those weakness are really important at a very good level. Those weakness are balanced by an awesome power in Tactics, defense etc...

Best regards


Nicolas
User avatar
Peter Grayson
Member
Posts: 207
Joined: Sat Aug 04, 2007 3:23 pm
Location: South Wales, UK

Post by Peter Grayson »

In the 1980's and 1990's, the strength of the available dedicated chess computers fell within the range from beginner to FM with top end 1990's machines perhaps hitting the bottom end of IM strength. Consequently there were sufficient recorded games against players within these strength bands when the allocated Elo for each machine bore some resemblance to the Elo strength of similarly rated human players.

Having personally owned dedicated machines from the Sargon 2.5 through to the London 030, the Elo ratings allocated to machines were somewhat optimistic in the early to mid 1980's but became more accurate as time passed when for example, the Selective Search rating list was a very good reference to identify the likely performance for prospective purchasers of these machines.

Entering into the new Millennium, it became clear that with most PC engines being commercial and with the engines starting to compete with the higher end of chess players, their sales people were all too ready to exploit good results against established IM and perhaps some GM players that was seen to be detrimental to those players especially if they were seeking to make a living from chess. The consequence was a dearth of games between PC engine and players with the loss of any correlation today between an engine's allocated Elo and the FIDE published Elo's. Since that time PC engine Elo's have continued to be calculated using the Arpad Elo system but with reference to other engines and not to chess players. Therefore today, the Chess engine rating system has no stable reference with respect to human players and neither then to those dedicated machines of yesteryear.

Thus there is no reason to suppose those dedicated machines have become 300 to 350 Elo weaker than they were 20 to 30 years ago. The issue is more likely to be the mechanism of performance measurement of the PC engines is no longer valid, especially when we need to be reminded the only accurate Elo rating using the Arpad Elo system is when two players have equal results against each other it can be confirmed with a high degree of certainty that those players are equal in strength. The performance probabilities using that system seem to work well enough for chess players within a certain band of Elo but once the Elo differential exceeds 200 Elo it is doubtful there is much accuracy from the resultant calculation. As I recall from my playing days there was a cap on the maximum Elo points loss or gain from any single result.
Larry
Senior Member
Posts: 2269
Joined: Wed Aug 01, 2007 8:42 am
Location: Gosford, NSW Australia

Post by Larry »

I just happened to be browsing u-tubes today and happened upon
this one. While it won't be the final word in this discussion, it is
relevant to it. It's old, circa 1993, and features the well known
Brit, Eric Hallsworth:
https://www.youtube.com/watch?v=74xHd4A ... e=youtu.be

L
Post Reply