03 Dec 08
1 edit
Just had an interesting thought.
Running a top box against the games of the greatest chess players
in history and findling a low match does not mean the box is correct.
In Human v Human games there are other factors to be considered
that cannot possibly be programmed into a computer.
Tiredness
Nerves
Fear
Keeping the draw in hand (looking at other games in a tournament
to see if you really need to win the game)
Having to play sharply to get a win.
Relaxing because you have won the tournament/league match etc.
Revenge
Playing the man (playing not the best move but one you know
will make your opponent feel uneasy - Lasker's ploy).
Playing a move you feel has a refutation but the position is too
difficult for a human to calculate out. (A Tal ploy).
Shooting for the brilliancy prize (a Street Fighter ploy)
Playing a very dodgy move and hoping (praying) your opponent
stumbles into a trap (a Greenpawn ploy)
I'm sure you can come with other differences that affect only
Human v Human games.
So a good Human v Human move will be looked at by a box without
taking in to consideration any of the above factors that do matter
in H v H encounters.
So Kep's analysis of the 1922 tournament was pointless.
The computer was looking at a different game.
That is why a top box v a good GM will win the majority of the games.
The GM comes with baggage.
so....
That is why the top three match up works.
The box will recognise it's own moves.
You set a thief to catch a thief, you have a computer looking for
computer moves.
You just need to stabalise the search (30 secs seems accepted but
why I do not know).
Agree on the sample of games. 20, 30 40.
Then settle on the % 90% 80% 70%.
I think then you need a strong player to look at the highest
match up's. Despite what Kep says a good player can smell silicon.
I just wanted to clarify that though the Greatest Chess Masters in
History score low when a top box looks at their games.
It's not because these guys were bad players - they were brilliant in
Human v Human games. It's because the box is calculating coldly
and nothing else. It's NOT thinking.
- Joined
- 07 Jun 05
- Moves
- 5301
03 Dec 08
Originally posted by no1marauderFischer being "the best of all time" is a matter of opinion. I happen to have other players in mind for the position. He was, as a matter of fact "a very good player". You saying he was not will not diminish his ability in the slightest.
BS. Somebody showing up on an internet site is not going to duplicate Fischer at the very height of his playing form. Fischer in 1971 was not "a very good player" he was the best of all time.
People like you can keep insisting that it is perfectly common for "good" internet players to duplicate such results. But you're kidding yourselves.
The point is not how common it is, or how good "very good" needs to be. The point is that given a sufficiently good player, pure mechanical "match-up rate" is not an adequate reason to ban them.
"People like me" do not insist that it is "common" at all.
"People like you" asked for evidence, and then whined when you got it.
03 Dec 08
1 edit
Originally posted by no1marauderYou should not be Fischer to play good against weaker opposition (Fischer reached his matchup playing against strongest world GMs of his time) when your opponents don`t make you to solve so much problems (and mistake).
BS. Somebody showing up on an internet site is not going to duplicate Fischer at the very height of his playing form. Fischer in 1971 was not "a very good player" he was the best of all time.
People like you can keep insisting that it is perfectly common for "good" internet players to duplicate such results. But you're kidding yourselves.
Also "higher matchup" not always means "better play" because of:
1) Despise of fast progress engines are still far from perfect - some kind of positions they don`t understand so good as humans;
2) You can make 40 good moves (supported also by engine) reaching good position and then lose due to blunder in 41th move.
03 Dec 08
Originally posted by gezzaYou provided NO evidence that players on RHP can achieve, without cheating, in a large sample the match up rates that Fischer did in a small sample of his games when he was playing at his absolute best. No, Fischer isn't a "very good player" like some players on RHP might be "very good players"; he was a super elite GM who is, at the very least, in the discussion of the greatest players of all time.
Fischer being "the best of all time" is a matter of opinion. I happen to have other players in mind for the position. He was, as a matter of fact "a very good player". You saying he was not will not diminish his ability in the slightest.
The point is not how common it is, or how good "very good" needs to be. The point is that given a sufficiently good p ...[text shortened]... n" at all.
"People like you" asked for evidence, and then whined when you got it.
Gatecrasher's analyses set an absolute upper limit of something like 83% match ups over a large series of games for super GMs and the best correspondence players of all time. Yet, somehow, some way, a few people here want to insist that "good" players on RHP can achieve match up levels significantly higher than that in a large sample of games by non-cheating means. That is absolute nonsense.
03 Dec 08
Originally posted by KorchI don't see how either one of these points is relevant to the discussion. As you know, Game Mods here do not use games played by engine suspects against much lower rated opposition or even short tactical games. So the first paragraph, while true, isn't really on point.
You should not be Fischer to play good against weaker opposition (Fischer reached his matchup playing against strongest world GMs of his time) when your opponents don`t make you to solve so much problems (and mistake).
Also "higher matchup" not always means "better play" because of:
1) Despise of fast progress engines are still far from perfect - some kin ...[text shortened]... (supported also by engine) reaching good position and then lose due to blunder in 41th move.
I agree with the second paragraph as well. But again I don't see what relevance it has to the discussion. I'm not claiming Fischer was a great player because his engine match ups are high; I'm saying he was a great player because his results were great.
03 Dec 08
5 edits
Originally posted by no1marauderFirst point is very relevant to the discussion. If Fischer has reached his high matchup against other top GMs (which playing strenght I presume could be about 100-200 rating points weaker) then I would not exclude possibility that 2400 rated human could reach similar matchup playing against 2200-2300 opposition. Games against 200 points lower rated opposition are (or at least were) analysed by Game mods.
I don't see how either one of these points is relevant to the discussion. As you know, Game Mods here do not use games played by engine suspects against much lower rated opposition or even short tactical games. So the first paragraph, while true, isn't really on point.
I agree with the second paragraph as well. But again I don't see wha engine match ups are high; I'm saying he was a great player because his results were great.
Second point was more overall remark.
P.S. But it does not matter for our guy - his matchup was much higher than Fischer, so there is no place for doubts.
03 Dec 08
Hi Gezza,
Quote:
"Fischer being "the best of all time" is a matter of opinion.
I happen to have other players in mind for the position."
Not looking for a squabble just curious. Who?
Think Fischer is certainly amongst top three along with Kasparov and Karpov.
But remembering that these players stand on the shoulders of
Alekhine, Capa, Lasker and Nimzovitch. Take these players games
out of chess history and we would be back in the 50's regarding chess
understanding.
03 Dec 08
Originally posted by no1marauderThis does not make any sense. You are comparing apples and oranges. The match-up rates for Fisher and Anand are for OTB games, not CC.
Anand, Fischer, et. al. don't play at RHP. Therefore, someone reaching their match up levels at RHP is cheating. Period.
What would be their match-up rates if they played CC?
A grandmaster or even an IM much weaker than Anand or Fisher should obtain a rate much higher than playing OTB. How much higher? Unfortunately there is limited or no data available to make this determination.
But the fact is that you simply don't know if there are IMs or GMs playing at RHP. There could be some in principle.
03 Dec 08
1 edit
Originally posted by smaiaTheir match up rates would be about the same in CC, perhaps slightly higher. Please review Gatecrasher's analyses earlier in the thread, last post on page 5 and several posts on page 6.
This does not make any sense. You are comparing apples and oranges. The match-up rates for Fisher and Anand are for OTB games, not CC.
What would be their match-up rates if they played CC?
A grandmaster or even an IM much weaker than Anand or Fisher should obtain a rate much higher than playing OTB. How much higher? Unfortunately there is limited or no data ...[text shortened]... you simply don't know if there are IMs or GMs playing at RHP. There could be some in principle.
Sure, there's whole bunches of super elite GMs just as good as the World Champion and the greatest players of all time playing on RHP.
🙄
04 Dec 08
2 edits
Originally posted by greenpawn34That would all be true if I was just looking at one game or one player. I'm not, so the tiredness/grumpiness/nastiness/need-a-draw factors should average out over the games and players. However, it is not particularly surprising that the test I was asked to perform on those two samples doesn't find a difference. There isn't one to find. In this particular case, match up rate is not particularly useful. In the case of a sample of games from an individual then high match up will be a good indicator of engine use if a high match up is obtained. A low match up doesn't actually tell us anything at all so the verdict should be the Scottish Not Proven rather than Not Guilty in such cases. It just happens to be less useful distinguishing between two samples of games by several players and engines.
Just had an interesting thought.
Running a top box against the games of the greatest chess players
in history and findling a low match does not mean the box is correct.
In Human v Human games there are other factors to be considered
that cannot possibly be programmed into a computer.
Tiredness
Nerves
Fear
Keeping the draw in hand (lo games. It's because the box is calculating coldly
and nothing else. It's NOT thinking.
In the particular case of the two samples I was asked to test, there are things that scream DIFFERENT but it isn't match up rates. I showed a summary of both samples to a very strong player I know. He was able to pick out the engine tournament from the human tournament after a quick glance. The factors he noticed? Game length and number of moves in book. "Engines don't get bored and aren't creative" was his comment. Not very useful as a test of an individual but that was not what I was doing.
04 Dec 08
Originally posted by no1marauderYou stated that the ceiling has nothing whatsoever to do with the test engine. I disagreed; I stated that a lower match-up to a weaker engine would be indicative of engine use since weaker players have more variability in move selection and subsequently provided an example that, in my opinion, makes sense. I stated that 45% match-up Top 1 Match to a 500 rated chess engine would be indicative of engine use but 60% match-up Top 1 Match to a much stronger engine may not be evidence beyond reasonable doubt of engine use.
That doesn't make any sense at all. Making the same assertion over and over and over again is not providing evidence to support that assertion.
Here’s why: two 2800 rated players are more likely to play the same move than two 500 rated players in a given position. I should have stated that deduction to support my point.
Something interesting:
From Gatecrasher’s statistics; these statistics show the matchups in which the moves humans were compared against those suggested of an engine and matchups in which the moves of engines were compared against the same engine:
{ Anand, Viswanathan - ELO 2779 (Games: 14) }
{ Top 1 Match: 224/355 ( 63.1% )
{ Fischer R (Games: 21) }
{ Top 1 Match: 378/602 ( 62.8% )
{ Aronian, Levon - ELO 2744 (Games: 14) }
{ Top 1 Match: 188/306 ( 61.4% )
{ Svidler, Peter - ELO 2728 (Games: 14) }
{ Top 1 Match: 189/311 ( 60.8% )
{ Rittner, H. (Games: 15) }
{ Top 1 Match: 203/337 ( 60.2% )
{ Junior - Intel Dunnington, 12 x 2.67Ghz (Games: 9) }
{ Top 1 Match: 281/469 ( 59.9% )
{ Cluster Toga - Cluster, 24 cores (Games: 9) }
{ Top 1 Match: 275/469 ( 58.6% )
{ Shredder - Intel Core 2, 8 x 3.16Ghz (Games: 9) }
{ Top 1 Match: 271/467 ( 58.0% )
{ Sjeng - Intel Core 2, 4 x 2.8Ghz (Games: 9) }
{ Top 1 Match: 266/459 ( 58.0% )
There is a disparity in the strength of the human group and engine group; the engine group is stronger.
Increased chess strength corresponds with a decreased match-up with an engine for the entities above.
I think that part of the reason for this result is because the test engine is closer in strength to the human players than to the chess engines. So I think that the strength of the test engine is an important factor in match-ups; if you disagree [and I hope that someone does because I like knowing how I am mistaken], then please explain why based on the above statistics.
04 Dec 08
2 edits
Originally posted by no1marauderAssume Anand or other top GMs were playing CC (same number of games as in the OTB universe used in the analysis) and spending - say - average of 5 hours per move in very complex games. 5 hours/move average vs. 3 min/move average (OTB).
Their match up rates would be about the same in CC, perhaps slightly higher. Please review Gatecrasher's analyses earlier in the thread, last post on page 5 and several posts on page 6.
Sure, there's whole bunches of super elite GMs just as good as the World Champion and the greatest players of all time playing on RHP.
🙄
You think his match-up rate would be just "slightly" higher than his OTB rate?
And I'm not saying whole bunches, but there can be a few. How do we know?
04 Dec 08
Originally posted by smaiaBecause a Fide Master who is rated 2300 OTB is #4 on the list & only 100 points off the #1 spot.
I'm not saying whole bunches, but there can be a few. How do we know?
That is a fairly good idea of the standard of play here.
A GM would very quickly rise to the top (unless they intended to throw games or blitzed too many in progress) and stay there.
04 Dec 08
Originally posted by smaia5 hours per move?? LMAO!
Assume Anand or other top GMs were playing CC (same number of games as in the OTB universe used in the analysis) and spending - say - average of 5 hours per move in very complex games. 5 hours/move average vs. 3 min/move average (OTB).
You think his match-up rate would be just "slightly" higher than his OTB rate?
And I'm not saying whole bunches, but there can be a few. How do we know?
Get real.
04 Dec 08
Originally posted by YugaYou're cherry picking the results to fit into your preconceptions. Once ALL the player results and ALL the computer results are compared, the computer match ups are predictably higher. Contrary to yet another one of your baseless assertions, there's zero evidence to suggest that GMs are stronger than the engines mentioned.
You stated that the ceiling has nothing whatsoever to do with the test engine. I disagreed; I stated that a lower match-up to a weaker engine would be indicative of engine use since weaker players have more variability in move selection and subsequently provided an example that, in my opinion, makes sense. I stated that 45% match-up Top 1 Match to a 500 rated ...[text shortened]... ecause I like knowing how I am mistaken], then please explain why based on the above statistics.
BTW, who the hell would use a "500 rated engine" (not that such a thing exists)?? Your "arguments" are a mixture of fantasy, poor comprehension skills and selective and superficial reading of the actual analysis.