It is disappointing, though not surprising, that Jeff Sonas has not responded to my concerns about Chessmetrics. Likely, because he has no answers.

Simply put, the absurd requirement that players lose rating points after a mere month's inactivity [this would be working really well during this Covid-19 lockdown...] makes Chessmetrics a nonsense. As soon as there is not continual frequent activity, Chessmetrics falls apart - as it does before 1950.

I sincerely hope that if Chessbase does the 1914 St Petersburg tournament, it will not use these ratings. Having Lasker as "12th" in the world would be very embarrassing indeed; the ratings for New York 1924 have been bad enough. Chessbase is better than this.

Thank you, Chessbase, for a great tournament series (sans the "ratings", of course).

your choose your bow tie style is your character I think and I choose Capa style

(continuing post)

Anyone with a reasonable knowledge of chess history would understand that all these cases are absurdities. I venture to say that if we had a retrospective Elo rating series (eg starting in 1870 instead of 1970), not one of these cases would occur.
Jeff, I say to you, you need to directly address and explain all these examples (and there are many lesser ones as well). Clearly there are far too many here that it can be explained away by saying there's an odd rouge result. Far more than just Lasker here! It is clear that these problems are endemic throughout the whole system. Very obviously to me, Chessmetics is quite unsound and unreliable, just a nonsense. Even if a rating looks "reasonable", we actually have no idea whether it is anything like accurate or not. I maintain that the main cause of these problems is the requirement for players to lose rating points after just one month of inactivity. And I maintain that these ratings are quite hopeless and are in effect disinformation; and that they should not be used as any sort of authority by anybody.

(continuing post)

(19) It has been known for a long time that Lasker is a problem for Chessmetrics (but obviously not the only one!). But we should be reminded just how incredibly ridiculous his Chessmetrics ratings are:
(a) 8th from December 1903 to April 1904. This was BEFORE the 1904 Cambridge Springs Tournament (ie before he had done anything "wrong"). Then after his mediocre performance at Cambridge Springs 1904, far below his usual standards, instead of going down, his rating goes UP!!
(b) Back to "8th" from November 1905 to May 1906, then "9th" from November 1906 to January 1907.
(c) An incredible 12th going into the 1914 St Petersburg tournament (just as well Chessbase hasn't done this one yet!).
(d) 8th again in April 1923.
(e) 28th in June 1934.
(f) 1st in June 1890. This is barely after Lasker started playing in tournaments. In the same list, Tarrasch is 9th(!!) This, despite Tarrasch's very impressive victory in the main event at Breslau 1889 (with 3 more big victories to come up to 1894), while Lasker struggled to win the reserve event.

(20) Tartakover 3rd in March 1921.

(21) Nimzovitch 23rd in December 1922.

(22) Makogonov 5th in July 1945.

(23) Fischer 9th in August 1966.

(24) Larsen 11th in November 1968.

(25) Keres 5th, ahead of Petrosian, Tal and Botvinnik, in July 1971 (he is 11-12= on the July 1971 Elo list, which is obviously far more accurate).

(to be continued)

(continuing post)

(1) Alekhine ahead of Capablanca going into the New York 1924 tournament. Anybody with any reasonable knowledge of chess history would know that this is absurd. There is no way that Alekhine was better than, or equal to, Capablanca in 1924, or any time before.

(2) Yet Alekhine is also "ahead" of Capablanca in September 1913!!

(3) Euwe 69th in January 1946 (and this is not just a rogue result, as he is 68th in the next list as well).

(4) Flohr 62nd in February 1943.

(5) Alekhine 13th in November 1920.

(6) Tarrasch 12th in August 1902.

(7) Pillsbury 1st, ahead of Lasker and Tarrasch, in April 1904.

(8) Janowsky 1st from May to September 1904.

(9) Bogoljubov 1st in January and February 1927.

(10) Fine 27th, behind such luminaries as Ragozin, Pilnik, Panov, Lisitsin, Bolbochan, Novotelnov when he would have been playing in the World Championship Tournament, March 1948.

(11) Fine 42nd in January 1948.

(12) Fine goes from 26th in the world in February 1935 to 36th in the world in May 1935 without having played a single game!

(13) Reshevsky 12th going into the World Championship Tournament, March 1948.

(14) Stahlberg 3rd in March 1948.

(15) Nimzovitch 2nd above Capablanca and Lasker from February to July 1913.

(16) Marshall 2nd above Capablanca and Lasker in August 1913.

(17) Capablanca 10th in February, May and June 1913 (after being 3rd in the world in 1909).

(18) Capablanca 18th in January and February 1935.

(post to be continued)

@Jeff Sonas:

I take your point regarding Edo ratings. Therefore, they are not suitable for use for New York 1924 either.

And before going further, I would like to commend your attempt in Chessmetrics to deal with the dreadful rating inflation problem. This is the feature I like most about both the Chessmetrics and Edo systems. However, I do feel a mechanism could also be inserted into the Elo system to deal with this problem.

And, yes, better data would obviously improve the results. "For the 19th century game results, I actually went through Jeremy Gaige's first Crosstables book and entered all of them, along with other sources for matches, and so the 19th century data is more complete. We would need to do something comparable for 20th century data". I wonder if something/some program could be set up online, and volunteers enter the results.

However, although better data would undoubtedly improve the results, I don't think this is the main cause of the problems. I still think the main cause is the requirement, in an n0n-professional era, for players to lose rating points after just one month's inactivity. This is what causes most of the absurd results.
You did not really address the main points I made. Jeff, really, how can a system that produces results like Euwe 69th in January 1946 and Fine 27th in March 1948 have any believability, integrity, or reliability? SERIOUSLY??
I said in my previous posting "But in fact, in earlier eras, before 1950, there are major problems everywhere". And there are some pretty questionable results from 1951 to 1975 as well. To prove these assertions, I provide about 30 examples:

(post to be continued)

You can check the rating [if that is displayed] to distinguish between Emmanuel and Edward

*saw

Could you distinct somehow Emanuel and Eduard Lasker? Sow in some books they used Em and Ed instead. It would be great to do so. Thank you!

The C.B coverage of this event has been magnificent. I'd like to see the paring method they used (players did not know who they were playing or what colour they would be till 15 minutes before the start of the game) used in today's tournaments.

We keep having to remind ourselves that up until about 1996 players were allowed to adjourn their games and use seconds to analyse their games. This was an advantage over modern chess to the death games and should be considered when comparing era's especially when we sometimes guffaw at a GM blowing a won ending that one of the old era's boys converted with ease. The old boy had a rest and often someone else worked out the win for them.

(maybe Chessmetrics should only use games that were never adjourned - good luck trying to sort that one out....)

I am also thoroughly enjoying the play-by-play of New York 1924! Certainly a remarkable achievement for Lasker.

With regard to ratings, I'd like to point out that around New York 1924, Chessmetrics and Edo are in pretty close agreement. Pre-tournament (Feb. 1924), Chessmetrics has Alekhine at 2776, Capablanca at 2769, and Em. Lasker at 2739. But, post-tournament (Apr. 1924) it has Lasker 2833, Capablanca 2824, and Alekhine 2775. The current Edo Ratings for 1924 (which also take the results of New York 1924 into account), have Lasker 2754, Capablanca 2742, and Alekhine 2686. Apart from the difference in scale, these are comparable to the post-tournament Chessmetrics ratings (subtract 83 points from the Chessmetrics ratings and they are very close).

All rating systems have to make assumptions and where the assumptions don't hold, the conclusions may be not be reliable. In general, where data is dense, different rating systems will give similar results, but the different assumptions made by different rating systems will have more effect where data is thin (e.g., in periods of inactivity for Chessmetrics, and where players are weakly connected to the main pool of players for Edo). No rating system is perfect!

I believe the Edo ratings look forward in time as well as backward, when trying to assess the strengths of players at a given point in time. So for instance the January 1st 1924 ratings would consider the results from New York 1924 and Moscow 1925 in the Edo calculation, and of course such an approach is going to benefit from its knowledge of the future. Would you have said the pre-match FIDE Elo ratings for Kasparov and Kramnik in 2000 were ridiculous because Kramnik clearly outperformed them? I could certainly give you much better ratings if I too was using my knowledge of the future to inform them. The Chessmetrics rating system uses only the game data from the previous 48 months for calculation of ratings, and so it is more appropriate for hypothetical "pre-tournament published ratings". And it also means that for semi-retired or mostly-inactive players, who don't have many results in the preceding 48 months, they are going to look like players lacking much evidence from recent play that they are still world-championship caliber. It should also be noted that for the game data for most of the 20th century, I used the Mega Database for my source of game results, and there are lots of lesser events where the results were known but not the game scores, and many of those events were not present in the Mega Database from around 2005. For the 19th century game results, I actually went through Jeremy Gaige's first Crosstables book and entered all of them, along with other sources for matches, and so the 19th century data is more complete. We would need to do something comparable for 20th century data in order to really get good historical "pre-event" ratings.

While I'd like to agree with AVRO1938, the ratings for any period of time are just an average temperature reading of various players strengths not an absolute by any means of standing. Today we have a far greater measure because nearly every, far more frequent, event counts!?

(continuing)

(1) In the January 1946 list, Chessmetrics has Euwe rated 69th in the world!! Any rating system that produces a result like this is obviously completely unviable (but we are told it is "better" than Elo...). It is inexcusable that results like this weren't checked for before publishing.
(2) For the 1948 World Championship tournament, the 6 players correctly selected for this event (Botvinnik, Reshevsky, Keres, Fine, Euwe, Smyslov) were, apart from Euwe who suffered a serious loss of form, clearly the 6 strongest in the world at the time. Yet for March 1948, when the event was held, Chessmetrics has Najdorf second on its list and Stahlberg third; while Reshevsky, who was considered to be the most likely to challenge Botvinnik for first place, was, wait for it, ...12th; and Fine was, um, believe it or not, ...27th. This is just completely stupid utter nonsense.
(3) In the January 1927 and February 1927 lists, Bogoljubov is given as first, ahead of Lasker, Alekhine and Capablanca. I'm sure that if anyone in 1927 had suggested that Bogoljubov was stronger than these three, they would have been laughed off the planet.

Chessmetrics ratings before about 1950 are complete rubbish, just absolute junk.
There is an alternative historical rating system (Edo ratings) that goes to 1928. Why not use that for New York 1924 instead? Although I think that there are also some issues with Edo ratings, these are as nothing compared with those of Chessmetrics. At least Edo ratings for 124 look sane. I would ask Chessbase, and anybody else, PLEASE do not use Chessmetrics ratings. It is very disappointing seeing these being used without a thought being given as to how bad they are. Not only are they quite useless, but they are, in effect, disinformation.

Has Chessbase already done Hastings 1895?

Ty Riprock is certainly correct when he says the ratings are a joke - they are an utter joke.
One of the main things on which Chessmetrics is based is the really absurd premise that players lose rating points after just one month of inactivity. While this might work, sort of, when play is frequent (say after 1975), it is obviously completely inappropriate for a less professional era when players, through no fault of their own, may not play for many months, even years.
Mr Sonas would have us believe there is just a liitle problem with Lasker, and maybe Botvinnik and Steinitz. But in fact, in earlier eras, before 1950, there are major problems everywhere if one bothers to look. Already, for this New York 1924 event, Alekhine is supposedly higher rated than Capablanca, which for 1924 is of course ridiculous. Three other examples, among very many:

(to be continued)

Agree with anthonyy. More of these great tournaments with the background setting would be very welcome. What’s next?

I do hope there will also be
Nottingham 36
and Zurich 53 !
i have the books, but it would be nice to have pictures and new comments

I am enjoying this very much. The very first Tournament Book I bought was New York 1924. Ironically Norbert one of the organizers was a relative. At the time I did not understand just how strong the players were, though some were at the end of their careers. Except for Nimzowitsch, Rubinstein and Spielmann not playing this easily could have been a Candidates tournament of the best players of the era. It also began an era where opening theory got turned on its head.

These have really been great. I know this tournament pretty well, and I'm still learning things here. Thank you!

@VerdajedrX
A report about round 7 will follow soon. For round by round reports about the previous rounds see the links at the end of the live-announcement. Or scroll down the website.

Is there footage for this one?

Was there round by round footage for the previous ones?

@Yayo Thanks!

@Johannes Fischer  

Mr. Fischer, many thanks for the prompt reply! 

The idea of replaying old tournaments and presenting them as if they were live is truly fantastic. It gives us a chance to study and enjoy the classics of the great Masters of the past in a new and fascinating way. I hope many more tournaments will follow!

@Yayo
The ratings follow the historical ratings of February 1924, calculated by Jeff Sonas on chessmetrics.com. (http://chessmetrics.com/cm/CM2/SingleMonth.asp?Params=192010SSSSS3S000000192402111000000000000010100). These ratings reflect the strength of the players before the New York Tournament 1924 started.

Please, one question on the Elo rating calculated for each player: is this calculated as an "average" Elo that each participant had during his entire career, or does it reflect the real strength at the exact moment the New York 1924 tournament was played? Thank you.

@niteguide
Yes, but they seem to be correct now. Two Laskers in one tournament - that is confusing indeed!

Ah, thank you jo-henry!
It seems the pairings were wrong. :-)

@niteguide
Thanks for writing. It is indeed confusing. But in round 6 Emanuel Lasker had White against Edward Lasker.

Sorry, but this is too confusing, who is who? https://ibb.co/vHs7tfv

Thoroughly enjoying this coverage. The fact that these great players played without engines and depend totally on their wits and study takes me back to what made me fall in love with Caissa back in the 70s. To be honest I’ve spent more time on following and studying this tournament than on most modern tournaments.

The USCF has Chess Life & Review Archives for previous years including 1974 (50th Anniversary of this tournament) and interested readerscould download and peruse them.

For the 50th anniversary of the tournament in Chess Life and Review Edward Lasker, as the last living participant, wrote a tremendous round by round discussion of the turnament that very much has that "You are there" feel. I still have my copy. It would be interesting if someone like Chessbase could get permission to republish it.

The Chessmetrics historical ratings (from 2005) were designed with a sort of "regression to the mean" behavior, where your rating would drift a bit toward an average player's rating when you were inactive. And it only considered your games from the previous 48 months, and ignored everything prior to that. In general this works well - and was an intentional feature of the ratings since it helped predictive ability in general - but in the specific case of Emanuel Lasker, it admittedly looks odd. For him and maybe a few others like Botvinnik or Steinitz, it might have been better for the ratings to include some amount of "regression to the player's previously-established level", in which case his pre-New York rating wouldn't have been so low.

Love New York 1924, a big thank you for the coverage of this fascinating tournament.

I agree with the other post: please clarify the description of E. Lasker by adding the First Name.

I don't believe the ratings can be called a joke. They appear to come from the Chessmetrics web site for the time of this tournament. One point of possible confusion is that both Emanuel and Edward Lasker played in the tournament, but with very different ratings. Today's game featured Lasker the Younger, not Lasker Preferred.

I have the Dover edition of the tournament book. Besides extensive notes by Alekhine it includes the treasurers report. Total cost was $13508.58. This was met by subscriptions, $9978.90; Tickets, $3476.00, and picture sales, $53.68. If you see it grab a copy of Edward Lasker's book, _Chess for Fun, Chess for Blood_. An interesting look at the era.

A quibble: Lasker was hardly a pre-tournament favorite. He was 53, 4 years past losing the title & only played 1 other tournament in 6 years (although he won it). His victory was hailed as a major feat due to his age. The ratings are also a joke.

I do appreciate this series; however, the Round 2 coverage comes across as a lazy production with a photo and a bunch of games - not really an article. Given that you are taking the trouble to reenact a classic tournament, it would be nice if you included some paragraphs as if you were there - similar to some previous articles of this nature. Thank you.

This was a nice tournament; It would have been better had Rubinstein and Nimzowitsch played.