"Try to play after you opponent has eaten or drunk freely."

click for larger view

(As given by <Johann Berger on page 273 of Theorie und Praxis der Endspiele (Berlin and Leipzig, 1922)>).

C. N. 5536
http://www.chesshistory.com/winter/...

The position is not from Lucena and cannot be found in his book from 1497 but

This hypothesis should predict that in lower quality (error-prone) World Championship matches, or in mismatches, there should be fewer draws, meaning more wins and losses. I examined lower quality or mismatched World Championship matches and came out with the following results, which I conveniently divided into several parts.

1. The Steinitzian Era of Mismatches and Erratic Play

1886 Steinitz-Zukertort (25% draws)

1889 Chigorin-Steinitz (6%)
[Only one draw in 17 games!]

1890 Gunsberg-Steinitz (47%)

1892 Steinitz-Chigorin (22%)

1894 Lasker-Steinitz (21%)

2. The Lasker Mismatches

1896 Lasker-Steinitz (29%)

1907 Lasker-Marshall (47%)

1908 Lasker-Tarrasch (31%)

1910 Lasker-Janowski (27%)

3. The Alekhine Massacres

1929 Alekhine-Bogoljubov (36%)

1934 Alekhine-Bogoljubov (58%)

1937 Euwe-Alekhine (44%)

4. The Erratic (Drunk?) Alekhine

1935 Alekhine-Euwe (43%)

5. The Sickly Tal Disaster

1961 Tal-Botvinnik (29%)

6. The Old Korchnoi Slide

1981 Karpov-Korchnoi (56%)

Results: As predicted, the World Chess Championship matches of low quality (error-prone) or mismatches produced relatively low drawing percentages. In none of these matches did the drawing percentage reach 60%; most remained below 50%. In many cases, the loser just got massacred, thereby lessening the number of draws. The 1935 Alekhine-Euwe match was closely contested score-wise, but it features a very erratic Alekhine who would play brilliantly and then suddenly start losing badly, which also reduced the drawing percentage. The 1890 Steinitz-Gunsberg and 1892 Steinitz-Chigorin WC matches were also close score-wise, but they again feature erratic play, not what we would expect from consistently strong super GMs.

1910 Lasker-Schlechter (80% draws)

1921 Capablanca-Lasker (71%)

1927 Alekhine-Capablanca (74%)

1951 Botvinnik-Bronstein (58%)

1954 Botvinnik-Smyslov (42%)

1957 Botvinnik-Smyslov (59%)

1958 Botvinnik-Smyslov (52%)

1960 Botvinnik-Tal (62%)

1963 Petrosian-Botvinnik (68%)

1966 Petrosian-Spassky (71%)

1969 Spassky-Petrosian (57%)

1972 Fischer-Spassky (52%)

1974 Karpov-Korchnoi (79% draws) [This was virtually a WC match.]

1978 Karpov-Korchnoi (66% draws)

1984 Karpov-Kasparov (83%)

1985 Karpov-Kasparov (67%)

1986 Kasparov-Karpov (63%)

1987 Kasparov-Karpov (67%)

1990 Kasparov-Karpov (71%)

1993 Kasparov-Short (65%)

1995 Kasparov-Anand (72%)

2000 Kramnik-Kasparov (87%)

2004 Kramnik-Leko (71%)

2006 Kramnik-Topalov (55%)

2008 Anand-Kramnik (64%)

Excluding the 1950 to 1972 era, the Lasker-Capablanca-Alekhine era and the Karpov-Kasparov-Kramnik-Anand era show that chess at the highest level of <human> playing ability, as demonstrated in World Championship matches, generally has draw percentages of 60% to 90%.

That's an interesting hypothesis you use in your study, but I'd have to query the overlapping percentage sizes between he upper end of your "mismatches" list, and the lower end of your "well-matched" list.

IMO the World Championship matches have been of insufficient duration for your methodology as it currently stands to validly support the exclusion of the 1950-1972 era.

Another way of stating your conclusions from the data and your methodology would be that "the post-Fischer era showed that chess at the highest level of <human> playing ability, as demonstrated in World Championship matches, generally has draw percentages of 60% to 90%".

I think that if you rely on quantitative date, you need to systematically account for the exceptions to statistically validate your conclusion.

I am not sure how to explain the drawing percentages of the 1950-1972 era. Botvinnik, Bronstein, Smyslov, Tal, Petrosian, Spassky and in general the other top mainly Soviet Union trained players of this era were IMO playing stronger chess than the 1800s era, if one compares their games. Yet the drawing percentages as you noted <overlap> or fall in-between the 1800s and the Karpov - Kasparov era.

It's possible that Karpov and Kasparov (and those that followed) in general represent an improvement of playing strength by BOTH contenders in WC matches that is a notch higher than the previous eras. Whereas before, one or both contenders would play relatively poorly, beginning in the Karpov era, we see WC matches in which BOTH contenders would generally play near the limit of the human capacity for chess.

If we take this view, how do we explain the 1910 Lasker-Schlechter, 1921 Capablanca-Lasker, and 1927 Alekhine-Capablanca matches? The drawing percentages fall fully within the Karpov era WC matches. Furthermore, if one takes a close look at the games these three pre-WW2 WC matches produced, it's even possible that they are generally of higher quality than the games from the more recent WC matches, surprising as that may sound but the games are there for any one to analyze and compare.

If so, another way of stating the conclusion above is in this manner:

It's an interesting conundrum. I get the impression, without having deeply analyzed the 1950-72 era, that the standard of play during the WC matches of that era was very high - perhaps excepting the second Tal match.

If we look at the situation the other way around, perhaps the relatively high ratio of results was because the few mistakes that were made were ruthlessly taken advantage of by the ultimate winner of each game and match.

This is purely hypothetical of course, but it is theoretically possible for a result to arise from only one mistake per game, and it doesn't have to be an obvious mistake. The all-Soviet matches are probably a direct consequence of the rise of the Soviet School of chess which raised the Soviet Union to absolute pre-eminence in chess, presumably because the standard of play had markedly increased through theory and rigorous training.

The extent of Soviet dominance was one reason why the Spassky-Fischer match had become such a spectacle attracting world wide interest. The Cold War paradigm was also prominent in the subtext from the popular viewpoint, but I believe from the point of view of chess, it had much more to do with breaking the monopoly of the seemingly invincible Soviet School of Chess.

The fact that it took someone like Fischer to do so maybe even underscores this aspect of Soviet dominance, namely that it took an oddball fanatic genius, a once-in-a-century phenomenon, who lived only for chess, to break the Soviet monopoly and <introduce a new dimension to chess>. What Alekhine brought to chess with his work ethic, Fischer raised to new heights by turning it into a life ethic, and raising chess to higher standards.

Fischer was rated nearly 2800 before anyone else even reached 2700.

I guess the point is there are qualitative issues like this to try and account for, as well as raw data involving results. The standard of training undertaken in the Soviet Union became somewhat more accessible to everyone with the rise of the machines.

Nevertheless, the Soviet legacy has endured almost to the present day. Anand is the first World Champion, undisputed or otherwise, since Fischer who was not from the Soviet bloc or from one of the countries of the ex-Soviet bloc.

Maybe one way of quantifying this is to use an engine to run routine blunder checks on all the matches (say 15-17 ply per non-obvious move), and weight the games and matches according to obvious blunders, and how they affected game and match outcomes.

That's an excellent suggestion. (I don't have a chess engine though.)

The 1951 to 1972 era bothered me because I expected, when I started computing for the draw percentages, that this era would conform to the Karpov - Kasparov era, with draws above 60%.

Explanations? Maybe it is just that the games of this era simply represent the lower range of results for high quality play. Take note of the following:

1951 Botvinnik-Bronstein (58%)

1954 Botvinnik-Smyslov (42%)

1957 Botvinnik-Smyslov (59%)

1958 Botvinnik-Smyslov (52%)

1960 Botvinnik-Tal (62%)

1963 Petrosian-Botvinnik (68%)

1966 Petrosian-Spassky (71%)

1969 Spassky-Petrosian (57%)

1972 Fischer-Spassky (52%)

All the draw percentages except the 1954 Botvinnik-Smyslov (42%) match are still above 50%. The following:

1951 Botvinnik-Bronstein (58%)

1957 Botvinnik-Smyslov (59%)

1960 Botvinnik-Tal (62%)

1963 Petrosian-Botvinnik (68%)

1966 Petrosian-Spassky (71%)

1969 Spassky-Petrosian (57%)

could be regarded as 60% and above, since there isn't much difference between, 57%, 58%, and 59%. If so only the following:

1954 Botvinnik-Smyslov (42%)

1958 Botvinnik-Smyslov (52%)

1972 Fischer-Spassky (52%)

are significantly below the 60% mark. These would then represent exceptions to the rule: <that chess at the highest level of human playing ability, as demonstrated in World Championship matches, generally has draw percentages of 60% to 90%.>

I don't think the conclusion can be signficantly refined further at this broad level of methodology.

There's not much doubt that modern tournaments are much more rigorous affairs than those from early last century, as I think it was GM Nunn and some others ran the blunder check methodology on selected tournaments in the 1910s and found an astounding level of basic mistakes you simply don't get in major tournaments anymore.

I'm looking at a WC game (Zukertort vs Steinitz, 1886) using the blunder check system, trying to figure out what constitutes a blunder. I'm also mapping the game with 14-16 ply evaluations, and getting a feel for how to weight evaluation shifts.

Is a losing move, say one that delivers a game to the opponent, or do we look at the cumulative affect of two or more inferior moves, that may turn a win into a draw or a loss? Should a faulty strategy be evaluated according to each move, or given a global weighting based on the accumulated effect of a sequence of inferior moves?

The problem is that any game with a result is going to result in an evaluation shift from close to 0 at move 1 to over 1.50 or , often much more.

So, some moves can be excluded from evaluation if they are nothing more than desperate moves in a lost position. But where to draw the line can be tricky.

I don't want to get too fussy with this for all sorts of reasons including time constraints, but I'm thinking a basic approach of weighting the quality of a game according to the total number of large evaluation shifts in either direction. Keeping it simple, I'll look at a perfect game as scoring 0, and adding two points for blunders (evaluation shift of greater than 1.40) and a point for bad moves (evaluation shift of between 0.80 and 1.40), working off a 16 ply engine evaluation baseline.

I think this may be useful in distinguishing between "quality" draws and draws resulting from mutually cancelling blunders.

&&&&

I recommend an engine as even some very powerful engines (ELO 2800+ standard) are now comparatively cheap (mine was less than $100), and they're really useful to use as blunder checks.

Instead of ordinary tournaments, IMO GM Nunn should have run the blunder check on the event that represented the topmost competition in 1910, the Lasker-Schlechter World Championship match.

<I'm looking at a WC game (Zukertort vs Steinitz, 1886) using the blunder check system>

I guess I'm curious now. I'm just letting the engine run for a minimum of 14 or 15 ply per move, and mapping the results. There's plenty other things I can do while the engine's running...

I'll see if I can construct some sort of methodology.

<GM Nunn should have run the blunder check on the event that represented the topmost competition in 1910, the Lasker-Schlechter World Championship match.>

That's good. However I think they should also have the same with

Lasker-Schlechter World Championship Match (1910)

I might do that myself when I put together something resembling a viable blunder-check methodology. What do you think of my preliminary ideas about using an engine's evaluation shifts based on a fixed ply analysis? THe first problem I've encountered with this idea is whether to stick to say 16 ply, or use it as a minimum. Leaving an engine on overnight to wrack up 20 or more ply can change evaluations significantly. Maybe that concern is too fussy.

I don't really want to start on any project like this until I feel that I have the basic principles of the methodology organized.

It sounds fine.

If you are going to push through with this project, it would be just great! I would suggest the first two matches to be the 1910 Lasker-Schlechter (80% draws) and 1921 Capablanca-Lasker (71%). I believe these two matches are crucial in our understanding of the modern World Championship Math because these are the matches that started it all - the first examples of the high quality nearly errorless closely fought World Championship matches that we see with regularity since the beginning of the Karpov era.

Here are some suggested caveats which you already stated above, stated in different words:

1. Desperate moves in a lost position that are regarded as 'blunders' by computers may not be real effective blunders.

2. Often when faced with multiple winning lines, an advantaged human chess player chooses the humanly simplest to follow, but which is regarded as a blunder by a computer. For example, there is a sure mate in 5 in a position, but follows a very complicated path from the viewpoint of the human chess eye. A human player would often choose a simpler but longer path; something like winning a piece instead, or transposing into a clearly won endgame. This is especially true when under time pressure. Such a decision to choose a longer winning line may be a blunder in the viewpoint of a computer, but in effect is perfectly fine, as it surely brings the full point anyway even though it takes longer.

I will do it (and it will take time), but rather than these two early matches I'd rather compare and contrast Schlechter-Lasker with one of the "massacres", preferably one of the recent ones such as the second Botvinnik-Tal natch or the second Korchnoi-Karpov match.

What do you reckon?

I've finished the preliminary mapping of the Zukertort-Steinitz game, and I'll deal with it on my forum, rather than here. I think your project should stay on page 1 for a while at least.

It gets even more mysterious considering that it is his NN opponent and not Lucena himself that had an Elo of 2720 back in the 15th century: