In case you don't believe Süchting was some 300 points below Lasker, then how large the gap actually was in your opinion?

Computer analyses generally haven't esteemed Lasker's play highly, he is known to be a quite practical player. But Lasker's combined score against Alekhine and Capablanca is almost even. And the quality of play of these players is quite high according to the engines.

For that reason I believe Lasker's level of play was equivalent of modern 2600-player, and 2400 is absurdly low for him or other great players of his era.

lambda, i clearly agree that you can't automatically say that the modern player is better in an absolute sense.

more fundamentally, i don't think the pure skills of the player from decades ago and those of the modern player can be compared in any meaningful way, because:

1) we don't know how the modern player would've performed under the <conditions> of several decades ago, and

2) we don't know how the player from decades ago would've performed under the <conditions> of the modern player

while anyone is free to keep *guessing* and making assumptions about these <hypothetical> scenarios, nothing precise, accurate or testable (falsifiable) can be claimed with any notable certainty.

the same applies equally to chess. consider modern, elite players: while they often *choose* to avoid theoretical lines in their practical approach to chess, they still need to *know* huge amounts of theory whether they want to engage in theoretical discussions or not. hence, it creates a need for a type of memory that players 100 years ago didn't need to posess in order to be successful. the fear of deep preparations has lead current top players (carlsen being very typical in that respect) to deviate from what possibly is "the most perfect chess" theoretically and objectively, which is what opening theory essentially aims at. but being "independent" and threading untheoretical waters today requires lots and lots of theoretical knowledge, which is kind of ironic.

another issue is that the number of elite players that have huge knowledge of chess and know the principles of sound, positional play, how to best defend against typical strategies, and so on, is much bigger than before. even the "average" elite players are awesome defenders today - beating a 2700-player that has the white pieces and is *content* with achieving a draw can generally *not* be accomplished by means of sound, positional play only; the modern, well-schooled 2700-player knows how to keep the position balanced - capablanca-style play with the black pieces doesn't bite on him/her as long as he/she considers a draw ok. (please find examples of draw-seeking play by 2700-players having the white pieces that still lead to a loss, and i'll provide 10 counter-examples for every example.)

this leads to another cause from deviation from "objectively correct" and "perfect" play if you're ambitious and want to win also with the black pieces against a draw-content 2700-player. carlsen & co (at the very end of the rating scale) often *have to* go fishing in muddy waters to obtain winning chances, and doing that is very double-edged against the 50 players or so in the current 2680 to 2750 rating band. they *will* punish you if you become too careless in your adventures.

the last notable change i'll mention is the time controls that have become increasingly faster, where *increments* is an entirely new concept (which requires different time spending strategies). obviously related is the abolishment of adjournments, which has created much *longer* sessions of play than what was usual decades ago; when games go on for 5, 6 and sometimes more than 7 hours at a time, without breaks, fatigue becomes a factor to a greater extent than before. this changes the odds between players, allows for new strategies based on wearing the opponent down also physically (!) - and quite likely alters the "objective quality" of the later stages of the game.

all of the above changes make it rather hard, essentially impossible, to predict how historical players would've fared under modern conditions, and similarly, hard to predict how the modern top players would've fared under the conditions 50 or 100 years ago.

hence, we have a contradiction, don't we?

the methods that rank alekhine and (in particular) capablanca highly, while being much harsher in their evaluation of lasker must be missing something - maybe the *pragmatic* fight of how to achieve the best possible results under the ruling conditions of your time.

today i think we have two rather different "schools" of thought in play at the top level, represented by

1) the players that generally aim to play the moves they think are *objectively* best

and

2) the players that generally aim to play the moves they think are *most effective* in achieving the desired result (typically a win)

the categories aren't necessarily mutually exclusive, but i think they are useful illustrations - and that this notion of "effective/practical" play represents one of the more serious problems of assessing "quality of play" by means of computer evaluations.

the ultimate goal of <competitive chess> is to win the most games and/or to score the highest number of points - and it's far from obvious that this is achieved by always searching for and/or playing the "objectively best" move (or the most "accurate" one as judged by computer chess engines).

put differently: in the very hypothetical scenario that we could device a method/system/test capable of accurately measuring the "objective chess skills" of players, it's *not* a given that the (current) player scoring highest in such a test would simultaneously be the most successful player in practical play and actual competitions.

Obviously the methods that ignore practical play are incomplete. Trying to cause opponent to play weaker is a vital part of chess. You are absolutely right. But I'm not yet convinced that measuring practical play is absolutely impossible.

<hence, my conclusion once more is: trying to compare players across eras is rather futile, for multiple reasons, the most important one being that the ultimate goal of competitive chess is to be successful under the current conditions. and the conditions have always been changing. >

As long as all factors that determine the level of play are not taken into account, your presumption might be true, but imagine a hypothetical scenario where all of these factos and their effect on different players have been successfully described, in that case you wouldn't object it?

<2) the players that generally aim to play the moves they think are *most effective* in achieving the desired result (typically a win)>

as i said, the two "approaches" aren't necessarily mutually exclusive - maybe not even at all exclusive - as i certainly agree that those "always" aiming for the "objectively best" move probably consider the resulting move(s) the most effective too.

the point is the difference in philosophy or belief, if you will: playing a move that you consider "objectively worse" (than other candidates) because you either think (or hope!) that it's more effective in getting you what you want *does* represent a deviation from the the line of thinking represented by the first category.

in fact, some players might even go as far as playing moves based on the reasoning that these moves will <statistically> - or on average and in the long run - provide them the best results, and thereby they may even slightly dismiss features of the current, concrete position partly based on considerations of a probabilistic nature.

some would use the word "gamble" about such a decision, which reminds me what my general manager of development (r&d) here in mdcn once said; he's not a chess player, but he's good at understanding people, and based on very little interaction with carlsen (some years ago, when fast was carlsen's sponsor for one year) he made this observation, communicated to me over lunch: "the guy is a gambler!"

nimh, the problem is that this scenario will forever remain *hypothetical*. hence, what i or anyone thinks about it isn't all that relevant, is it?

<their effect on different players>

<I'm not yet convinced that measuring practical play is absolutely impossible.>

well, i'm more or less convinced, because it would require the kind of meta-knowledge about players that carlsen & co *do* posess, but that "analysis systems" don't.

people may choose practical approaches based on:

- general knowledge about the opponent
- current state of opponent
- current state of oneself

"current state" may refer to
- standing in tournament
- recent form
- observed health/fatigue
- state of preparation (for a match)
and much more.

capturing all of the above info from the games only is clearly impossible, and even capturing and meaningfully representing the above in other ways seems essentially impossible to me.

in terms of the game score itself (available for analysis), both complicating and simplifying the position (over "objectively" stronger moves) can be done for practical reasons - based on one or several of the factors/meta-data mentioned above.

obviously, the best one can imagine would be some coarse-grained heuristics that possibly can improve the overall perceived quality of computer analysis of players, in the sense that the results will correspond better to our already conceived prejudices of the matter (for current and historical players) - but would that really take us further?

i think the fundamental problem with these types of "measurements" is that we're making up what's being measured as much as measuring any actual, real substance. the methods available for *veryfying* a given approach are few and debatable (predicting or comparing to results being imperfect in my view). hence, the creator of the measuring device will always have notable room for refitting the measure to preconceived notions, to the point that he/she is "happy" about it - and the naive users will over-interpret the measurements anyway.

so, to zoom quickly back to the topic of this page: anand and gelfand are both great chess players, and the winner of the match will be the "world chess champion" - because this is the competition where the winner receives the right to use that title.

But the thing is I don't believe this remains hypothetical forever. And hence the rate of relevance increases a bit, at least for me. But I'm not forcing you, if you don't want to answer, you don't have to.

<our already conceived prejudices of the matter>

What prejudices exactly can you see in the computer analyses by me and other researchers? I thought my approach has been fairly objective so far.

<hence, the creator of the measuring device will always have notable room for refitting the measure to preconceived notions>

This sounds like refitting is something negative... As long as methods used are clearly imperfect, it's necessary to make amendments as to make methods more trustable. A some kind of etalon have always been used to calibrate yardsticks and measuring methods. In case of move analyses, one has to use historical rating systems and head-to-head scores as an etalon.

<i'd think there are numerous ways. which "two ways" did you have in mind?>

Agreed - but two massive blunders, (losing a piece & losing a rook) in the same tournament, just a year before his absolute prime? I only remember one instance of Karpov dropping a whole piece unforced (to Christiansen) in his entire career.>

Blunders occur at all levels. It is pretty obvious by study of the games of Capablanca and others that chess hasn't got "better" (well perhaps the :'level has but the players aren't inherently better). What has happened is technique is better ...

I had put quite lot of Fischer's openings to the test - now in many he was quite wrong in his moves BUT at the time recall there were no computers. Also even today some of the "mainlines" are not so good on computers so by using them well a player can hugely improve (o.k. some lines are pretty well worked out..but even then...) technique...that was obvious with Kasparov...

I think it is pretty safe to say that beyond a certain level (forgetting that ratings inflate - similar reason money does, population increase as someone said) ... but there is no way intelligence has increased in small (evolutionary) time period and the stats don't necessarily favour better chess...no, I all of Tarrasch's best games (183 of them, that is excluding some other games he had in his book first published in German I think which Jonathan Safarti (ex NZ Champion and was a member here) has studied)) and he was, in his prime, as good as any player ever alive. There are other examples. I am currently playing over Kasparov's and Karpov's games (he points out quite few times where he himself miscalculates (but it is true that he was very alert as young player as were Kasparov and Korchnoi), and we have to recall Rubinstein was very nervous and perhaps as excitable as Korchnoi was (is?)...also Korchnoi was (still is?) addicted to time trouble) and in the Korchnoi matches both he and Korchnoi blundered in a few games, I recall when that was on...)) and others... Just now I think it is between Anand Aronian Topalov Kramnik and Carlsen...of course there are other young players emerging.

You cant prove anything backwards (proleptically)..all you might be able to do is to compare players alive now at a similar age.

Was Galileo a greater scientist than Newton? Or was N greater than Einstein? (These are really silly questions, but I say this in the interests of truth!!

Or how to you say who is "the greatest, say, in boxing? Ali? People say that as both he and Fischer hit the headlines and with good reasons and they were bloody good...but... It is ultimately subjective. Blunders mean nothing. Kramnik overlooked a mate in one vs. Fritz. Carlsen asked for a move back after blundering....

But knowledge of chess and its subtleties and improved coaching techniques mean players have a greater average de facto level of chess. In other words, as that great rival to Tesla, Edison said of "genius":

"It is 99% perspiration and 1% inspiration."

Of those who worked very hard...well most if not all of the World Champions were big "perspirators" (laugh!) - even Capablanca studied hundreds of endings as a young man, and it is clear that such as Alekhien had a huge memory (by going through them)of many types of tactical procedures and so on....Soltis is good on this..he points out in his book on improving chess that the e.g. the Polgar sisters and for example Seirawan (who I actually had game of chess with once (of course I lost!), and indeed I had games online with Susan Polgar and once got a draw!)that these people had books with hundreds and hundreds of position types and or tactical examples which they studied and remembered..not to play rote chess but to be confidant in combination or positional situations....

This is not to downplay their ability to calculate etc

So what has improved is coaching and learning methods and perhaps we have more players at an age they keep their concentration etc and have great grasp of opening theory and endgame technique etc but we have NOT moved (inherently) "forward"...and in fact we will still find much to criticize in games by Anand etc great as he is...

You can statisticate until you go blind but you will prove nothing about chess or any other ability.

Also, some might say that X is the greatest ever as he / she makes such wonderful and imaginative and creative moves and combinations, and plays the most beautiful games...

But then you will find another player you are "amazed by", and the convergence of your excitement will shift to that person: and you will be very much in wonder and amazed...

Estimating the level of play scientifically by engine analysis of moves is a fairly new phenomenon, the first study was reported no earlier than 6 years ago. So, rigorous methods for estimating the validity of such analyses do not exist yet. All one can do is to compare one's findings against historical data.

I use the following 3 rules:

1) players from the same era, must retain ther relative difference based on chessmetrics ratings and classical head-to-head scores

2) players who had simpler positions are more likely to be overrated, and vice versa: players who had more difficult positions are likely to be underrated

3) if two players from different eras have the same relative strength, then the modern player must be ahead

<Analysis of moves and games is always interesting. But stats like 'this decade has seen more losing moves by top players on average' 1) are tricky to establish 2) lead to highly speculative conclusions.>

Strictly speaking, it's the average difference between the best move and actually played move, not the number of losing moves.