Elo's Book: The Rating of Chess Players
Sam Sloan
In 1978, Professor Arpad Elo, who established the Elo Rating System,
published a book entitled "The Rating of Chessplayers, Past and
Present". A description and review of the book can be seen at
http://www.amazon.com/exec/obidos/ASIN/0668047216/slavesofthomasje
This book, which is long out of print, contains the historical ratings
of 476 chess players, including Paul Morphy, who is rated 2690.
The oldest players on the rating list are Alexander Petrov
(1794-1867), who is rated 2590, and Jozsef Szen (1800-1857) who is
rated 2450. There are 41 players on the list who were born before
Morphy.
I have just posted the list on my web site. The address is
http://www.anusha.com/eloslist.htm
The list is in text format delimited by tabs. Therefore, it is easy to
take the list, insert it into any spread sheet program and sort it by
name, birth date, rating or country.
I have sorted it in name alphabetical order.
Perhaps the most interesting question which the list addresses is: Who
is the strongest chess player in history, before Fischer of course.
According to Elo's list, the strongest chess player was Capablanca,
who is rated 2725.
After that comes Lasker and Botvinnik, both rated 2720, Tal, rated
2700, and Alekhine, Morphy and Smyslov, all rated 2690.
These are all five year peak averages. There is also a list of the
ratings of every chess player still alive on 1/1/1978, the final date
of the list.
The list contains surprises. For example, Gustav Neumann (1838-1881),
an almost unknown German player, is rated 2570, making him one of the
strongest players in the world in his day. Neumann rarely played
outside of his home town, but frequently played with Adolph Anderssen
(1818-1879), who is rated 2600.
I need to thank Jeff Sonas je...@kasparovchess.com for e-mailing me the
list. I used to have it in my computer, but I lost it in October, 1997
when I took my computer with me on a trip to see my girlfriend,
Passion Julinsey, and somebody stole my computer off of the Greyhound
Bus.
Sam Sloan
http://www.anusha.com/elosbook.htm
Sam Sloan
wrote:
>Sam Sloan wrote:
>
>> I have just posted the list on my web site. The address is
>> http://www.anusha.com/eloslist.htm
>>
>> The list is in text format delimited by tabs. Therefore, it is easy to
>> take the list, insert it into any spread sheet program and sort it by
>> name, birth date, rating or country.
>
>Look again, the tabs apparently disappeared when you converted the list to
>html.
>
> -Paul
I use the <PRE> and </PRE> tags
If you save the html file to disk, the tabs should remain. Then, using
a text editor like Wordpad, you should be able to remove the headers
and footers and insert the remaining into a spread sheet for sorting
by rating for example.
Sam Soan
Sam Sloan
wrote:
I just tried this just to make sure and it works fine. Just delete
everything before and including the <PRE> tag and everything after and
includiing the </PRE> tag.
Sam Sloan
Bill Smythe
>In 1978, Professor Arpad Elo, who established the Elo Rating System,
>published a book entitled "The Rating of Chessplayers, Past and
>Present". A description and review of the book can be seen at
>http://www.amazon.com/exec/obidos/ASIN/0668047216/slavesofthomasje
______________
That review, by the way, is mine (brag, brag).
Bill Smythe
Sam Sloan
Yes. I noticed that and I was trying to decide whether to mention
this, but I decided not too, not that there was anything wrong but it
would distract attention away from my subject matter.
Neil Brennen
Bill, in your review you didn't mention that the data used to compute
the historical ratings was flawed. Elo rated a match that was never
played, as pointed out by Edward Winter in Chess Explorations, p 162.
--
Sincerely,
Neil Brennen
Editor and Publisher, DELAWARE VALLEY CHESS
Sent via Deja.com
http://www.deja.com/
Bill Smythe
>Bill, in your review you didn't mention that the data used to compute
>the historical ratings was flawed. Elo rated a match that was never
>played, as pointed out by Edward Winter in Chess Explorations, p 162.
________________
Interesting information. How long (how many games) were in this non-match?
Are there other flaws in Elo's data, as well? How many games, matches,
tournaments, etc are there altogether in Elo's data? How many historical
ratings are significantly affected by the flawed data?
Bill Smythe
Neil Brennen
Winter says Elo includes a match Capablanca-Kostic, 14 games played in
1915. No such match was played.
Also, Winter points out, on page 272 of Chess Explorations, that Elo
mixes up Samuel Tinsley and E.S. Tinsley, and assigns the "historical
rating" of one to the other.
I don't know the total number of results Elo used in the book. But as
in the current USCF rating system, bad results affect everyone. Just as
the submission of my 14 game match with Bill Smythe will eventually
affect the entire rating pool. :-)
Bill Smythe
> .... bad results affect everyone. Just as
>the submission of my 14 game match with Bill Smythe will eventually
>affect the entire rating pool. :-)
_________________
Hmm. Do you have the game scores? What were the opening 5 moves of each
game?
Bill Smythe
Anders Thulin
Bill Smythe wrote:
> Interesting information. How long (how many games) were in this non-match?
14 games. According to Feenstra Kuiper's book 'Hundert Jahre Schachzweikämpfe'
(Amsterdam, 1967) the details are:
1915 Buenos Aires J. R. Capablanca -- B. Kostic +12 -2 =0
It seems reasonable enough to assume that Elo relied on F. K. here, but it
should be noted that Elo published a series of retrospective ratings already
in 1964 in Chess Life. I hope there's no mention of this 'match' in that
article.
However ... it's difficult to prove that this is incorrect data. All
that can be said is that it doesn't appear anywhere else. I don't think the source
of the problem has been found: did F. K. mistype, misremember, or did he, in his turn,
rely on a flawed source?
Kostic says in an article in Kagans Neueste Schachnachricten (1928, p. 176)
that he developed as a chess player in the US during WW1. That doesn't help much,
I know.
> Are there other flaws in Elo's data, as well?
As far as I know, no other error has been verified. (The Tinsley error mentioned
by Neil Brennen is not clearly over anything else than the name.)
I may be overinterpreting, but in the referenced note Winter says that if
"Elo's historical data are largely based on Feenstra Kuiper many of his oft-quoted
retrogradings will be way out", and so seems to give Elo's work the benefit of
the doubt, suggesting that there were no more problems known at the time, while at
the same time creating the impression that Feenstra Kuiper is largely unreliable.
He may be to some extent -- his early match records seem to rely a lot on
Bachmann's data, and they contain several problems. Bachmann does not list the
Buenos Aires match, incidentally, in his article "Übersicht der Spielerfolge der
bedeutendsten Schachmeister ..." in the Teplitz-Schönau 1922 tournament book.
Elo cites Feenstra Kuiper's both works as well as Gaige's crosstables as sources.
However, as he did include exhibition games for the period 1846-1862 to pad out the
scanty data for that period, we certainly have more than one source of errors to cope
with. If he included also such games that are clearly non-serious, there may be
several. Is it possible to check? Is the raw data deposited somewhere, perhaps?
> How many games, matches,
> tournaments, etc are there altogether in Elo's data?
I don't think he says ... as far as I can find, the closest statement is
found in 5.43: "Much time and effort have been devoted to the collection of every
bit of published data, every master tournament crosstable, and match result."
> How many historical
> ratings are significantly affected by the flawed data?
What is significant here? 1 rating point? 5? or more?
Capablanca is probably slightly overrated as a result, but the error margin of
the entire retro-rating (which seems to be left undocumented) is, at a wild guess,
on the order of +- 10 rating points already. To that comes the almost certain
overrating of those early players whose exhibition games were used as serious games
could not be found.
I doubt that the error from the inclusion of that single match affects
Capas rating more than to 5 points, and any indirect effect on other players to be
much less. But it *is* a just wild guess. The computations would have to be repeated,
with the flawed data removed to give an accurate evaluation.
--
Anders Thulin a...@algonet.se http://www.algonet.se/~ath
Cdunn3
>> Are there other flaws in Elo's data, as well?
I have both Elo's book and "Oxford Encyclopia of Chess Games,1485-1866"
and some of the ratings Elo gives for
the early players seem too high. I cite
4 examples:
Lasa (2600) - The only serious games were
1853 match with Staunton (+5-4+3)
won by Lasa. Staunton was past his
peak. Lasa was +2-2 vs. Anderssen
1846, and was even with Hanstein and
Janisch 1836-1842 (non serious games)
Hanstein(2480)- games with Lasa as
given and +5-0 vs. Janisch(2360) in
1842 when Janisch had probably not
reached his peak.
Petrov (2530) born 1794, undoubtedly
a great player, but only 9 games vs.
Janisch(again) and D. Urosov, 1844-
1853, neither top flight competion.
DuBois (2550), also surely a good player,
but his only tournament was London 1862
(+5-4=3) where he finished below
Owen (2380) and just ahead of
MacDonnell (2410). He lost a match
to Steinitz 1862 (+3-5=1) and this was
before Steinitz was anywhere close to
his peak. DuBois had some non serious
games with Anderssen 1862 (+1-3)
None of the above results translates to
a 2550 overall rating.
Are any other of the Lasa, Hanstein, or
DuBois games available any where ?
CarlDunn
StanB
Bill Smythe <chic...@enteract.com> wrote in message
> Hmm. Do you have the game scores? What were the opening 5 moves of each
> game?
In the games in which you opened 1. d4, Neil played 1..., f5
StanB
Bill Smythe
>In the games in which you opened 1. d4, Neil played 1..., f5
_____________
Okay then, what was my second move in each case?
Bill Smythe
Neil Brennen
I'll have to hunt for the scores. All I remember is I lost them all.
And of course the rating points as well.
Neil Brennen
Bill Smythe
>I'll have to hunt for the scores. All I remember is I lost them all.
_______________
If you had truly played me, you wouldn't need to hunt for the scores. My
second move in games beginning 1. d4 f5 is sufficiently unusual that you
wouldn't have forgotten.
Bill Smythe
Neil Brennen
the point incorrect results corrupt the rating system. Just as an
imaginary match will corrupt Elo's historical ratings, so will an
imaginary match between Brennen and Smythe corrupt the USCF rating
system.
I have never had the pleasure of meeting you, let alone playing you.
Perhaps one day I will.
Anders Thulin
Cdunn3 wrote:
> I have both Elo's book and "Oxford Encyclopia of Chess Games,1485-1866"
> and some of the ratings Elo gives for
> the early players seem too high.
Too high in comparison with what?
We should be very careful about comparing them with FIDE or USCF
ratings. Personally, I don't think that can be done at all. It's possible
to compare them with other ratings for the same period, and so get an idea of
relative strengths during that time, but to compare a rating from
the 1850's (and a rather wonky one, at best) with one from the 2000's
needs some kind of strengthening argument to show that it can indeed
be done meaningfully.
Another interesting problem is how Elo got his retrospective ratings
in sync with modern ratings -- I don't think he explains it anywhere.
In one example calculation he just adds 2000 (3.46), a magic number that
is not explained -- why 2000? why not 1950 or 2061?
> [...]
> None of the above results translates to
> a 2550 overall rating.
A rating is only a result of a mathematical process. Unless you're sure
you use the same data and the same process as Elo, the difference may not
mean anything, as the measures would simply be incommensurable.
The relative differences in strength between players in the same period
may be a better choice of measure: is Dubois's rating in comparison
with those of von der Lasa or Hanstein wrong?
Whether one is better than another is another and more difficult question.
I would probably prefer to refer to Prof. Glickman's ratings for getting
the relative strengths for these older players, as the selection of games
seems to have been more severe -- Morphy did not get a rating, as there
were too few games for a reliable rating. (I forget the place of that paper
-- a rather technical one somewhere on http://math.bu.edu/people/mg/ratings.html.
I think it's one of those PostScript papers ... )
The Glicko method of rating seems more attractive actually, as it
incorporates the uncertainty of the rating.
Anders Thulin
I forgot the important question:
Cdunn3 wrote:
> Are any other of the Lasa, Hanstein, or
> DuBois games available any where ?
Early years of Schachzeitung and other early German chess journals (such as
Deutsche Schachzeitung) for the German players. Dubois I would expect to find
more of in French chess journals, which I unfortunately know only of, but
very little from -- although I think I can recall one or two games printed
Schachzeitung as well.
Perhaps a check with Jay Whitehead might produce some?
Jeremy Spinrad
in pgn notation, so some of these are easily available.
Jerry Spinrad
Bill Smythe
>Bill, I fear you are taking me seriously on this.
______________
Nope. But it was fun while it lasted. I was hoping to provoke somebody
into asking me what 2nd move I play as White against the Dutch. I'm proud
of my weird openings.
Bill Smythe
Todd Durham
Actually, I was going to ask and forgot. Is your second move 2...b5?
Todd
Todd Durham
Duh. Not paying enough attention. Since I think I'm paying a suitable
amount of attention, I will now ask if your second move is 2 g4.
Todd
Neil Brennen
Okay, I'm game. 1.d4 f5 and Black wins by force. Isn't this so Bill? :-)
Bill Smythe
>Okay, I'm game. 1.d4 f5 and Black wins by force. Isn't this so Bill? :-)
_____________
Aha, I provoked you, as hoped. Todd Durham was close. One of my favorite
games EVER went like this:
1. d4 f5 2. Qd3 e6 3. g4 fxg4 4. h3 Nf6 5. hxg4 Nxg4 6. Rxh7 Qf6 7.
Rxh8 Qxf2+ 8. Kd2 Qxg8 9. Qg6+ Ke7 10. Kc3 Qxf1 11. Qxg4 Qxc1 12. Qf3
e5 13. Qxf8+ Ke6 14. Qe8+ Kf6 15. dxe5+ Kf5 16. Rh5+ Ke4 17. Nd2+ 1-0
This game illustrates two important chess principles: (1) "With an attack
like this, who needs development?" and (2) Kings should be moved off the
first rank as soon as possible.
Bill Smythe
StanB
Bill Smythe <chic...@enteract.com> wrote in message
> StanB wrote:
I don't know what you play, I only know what Neil plays. He loves the Dutch
Leningrad.
StanB
Todd Durham
Very cute! But what about 12..d6? I'm sure I'm missing something, but I
keep seeing lines like
13 Qxf8+ Kd7
14 Qxc8+ Kc6
15 Qe8+ Nd7
16 Qxa8 Qe3+
17 Kb4 Qxd4+
18 c4 Qxb2+
19 Ka5 Qxa1
20 Rh3 Qxb1
21 Qh8 Qxa2+
22 Kb4 Qb2+
23 Ka5 b6+
24 Ka6 Nc5+
25 Kxa7 Qa2+
26 Kb8 Nd7+
27 Kc8 Qa8#
(Yes, I know 24 ... Qa2+ wins faster, but how can one resist 24 ...
Nc5+?) Actually, White can avoid all of this with 15 Qxe6, but Black
still has some fight in 'im, I think. Anyway, it's a pretty game, and (I
think) a pretty variation I have given. There are lot's of cute litte
sidelines, too, for those who want to play with it some. Also, I didn't
run any of it by a program, so it's likely choke full of holes other
than the one I mentioned.
Todd
PS Thanks for the game. This was fun!
StanB
Bill Smythe <chic...@enteract.com> wrote in message
> Neil Brennen wrote:
Give us a hint. Does the h3 square have anything to do with it?
StanB
Bill Smythe
>Give us a hint. Does the h3 square have anything to do with it?
_____________
By now you've seen the answer, so you realize that h3 plays, shall we say, a
cameo role.
Bill Smythe
Bill Smythe
>> 1. d4 f5 2. Qd3 e6 3. g4 fxg4 4. h3 Nf6 5. hxg4 Nxg4 6. Rxh7 Qf6 7.
>> Rxh8 Qxf2+ 8. Kd2 Qxg8 9. Qg6+ Ke7 10. Kc3 Qxf1 11. Qxg4 Qxc1 12.
Qf3
>> e5 13. Qxf8+ Ke6 14. Qe8+ Kf6 15. dxe5+ Kf5 16. Rh5+ Ke4 17. Nd2+
1-0
>
>Very cute! But what about 12..d6? I'm sure I'm missing something, but I
>keep seeing lines like
>
>13 Qxf8+ Kd7
>14 Qxc8+ ....
______________
Stop right there. 14 Qe8+ is mate.
12...d5 looks no better, either.
Bill Smythe
StanB
Yes. I am aware of that line but hardly familiar with it. It looks like a
lot of fun and I think I'll fool around with it for a while. Thx.
StanB
Todd Durham
Crap. Talk about missing the obvious. The rest of it was a great deal of
fun, however.
> 12...d5 looks no better, either.
No. 12...d5 just invites the Q to take on e6 with check, and then
slaughter the Black King. Ah, well, it was fun for a while, and it shows
one more reason why I'm a class B lifer.
Todd
>
> Bill Smythe
Bill Smythe
> .... It looks like a
>lot of fun and I think I'll fool around with it for a while.
______________
Good luck. Unfortunately, there are ways Black can make it quite a bit less
fun.
Bill Smythe