27

Chess ratings are a relative thing at a particular time, just like boxing. It is not absolute, compared to weightlifting for example, in which the records keep getting broken. How would say a 1700 rated player of say 50 years ago who is no longer around go against a 1700 player of today? Does this mean that chess ratings over time have grade deflation?

4
  • 6
    Computer chess programs can provide reference points of fixed ability over time. They were already being given ratings in the 80's. If these programs are still available in a runnable form, we could illuminate any drift in the rating process simply by having those same programs get re-rated today, although it is possible to exploit their weaknesses if you know how.
    – Matt
    Mar 21 at 15:03
  • Rating records are being broken but in this case we can say for sure that top players are getting better thanks to the better tools that are available to them. For the amateur players, I don't think there's any hard evidence suggesting a significant rating deflation process
    – David
    Apr 16 at 6:44
  • This reminds me of the Flynn Effect where IQ scores have been observed to be increasing over the last ~80 years.
    – shabbychef
    Sep 13 at 4:17
  • Very related question: chess.stackexchange.com/questions/1632/… Sep 13 at 20:49
21

How would say a 1700 rated player of say 50 years ago who is no longer .around go against a 1700 player of today?

Speaking as a player who was around 50 years ago and was rated the equivalent of 1800 in 1973 (my BCF grading was 150 with a generally accepted conversion formula of ELO = BCF x 8 + 600) and is rated 1718 today I'm pretty sure I would beat my 17 year old self.

OK, the me of 2021 is probably not very objective but I am sure that if I managed to reach an even endgame against my 17 year old self I would be a very strong favourite. The reason being that in 1973 almost all the chess I played was at a time control of either 30 moves in 75 minutes or 36 in 90 with adjudication after 3 hours play. The 17 year old me hardly ever played an endgame and so wasn't that good at endgames.

That kind of time control was the norm in England 50 years ago and probably in much of the non-Soviet world. Weekend tournaments where large numbers of people played 3 full games a day with a time control of all the moves in 1 hour didn't come in until the mid to late 70's.

Does this mean that chess ratings over time have grade deflation?

From my personal view absolutely, yes. However my single example of personal experience is highly subjective and one case is meaningless.

To get a more objective view of whether there has been rating inflation or deflation I think we have to look at historic data. There is FIDE rating data available going back to 2001 on the download page of the FIDE rating website and back to 1971 (unofficial data back to 1967) on the Olimpbase site. These data can be downloaded and analyzed to get more meaningful evidence.

Comparing today with 1971 is problematic because in 1971 there was a rating floor of 2200 (nobody under 2200 was listed) so there were only 547 players and obviously nobody with a FIDE rating of 1700. However we can try and maybe also, more meaningfully compare today with 10 years ago.

The January 1971 list is here on Olimpbase. There we can see that the highest rated player, Bobby Fischer, is at 2740, more than 100 points lower than today's number 1, Magnus Carlsen on 2847. There are 16 players at or above 2600, 82 at or above 2500 out of 547 at or above 2200.

The comparative figures for March 2021 are 264 above 2600, 946 above 2500, 3111 above 2400 and 17334 above 2200.

January 1971
Players above 2600 = 16 (2.9% of over 2200s)
Players above 2500 = 82 (15% of over 2200s)
Players above 2400 = 278 (50.8% of over 2200s)

March 2021
Players above 2600 = 264 (1.5% of over 2200s)
Players above 2500 = 946 (5.5% of over 2200s)
Players above 2400 = 3111 (17.9% of over 2200s)

To me that looks like it is comparatively much harder today to get to GM / elite level than it was 50 years ago but the difference in sample sizes detracts somewhat from the percentages. And, of course, it doesn't say much about what is happening at the 1700 level.

If we go back 10 years and compare numbers and percentages (using total rated players) at 1700, 2150 and 2600 we get:

March 2021
Total rated players = 363590
Players rated above 2600 = 264 (0.07%)
Players rated above 2150 = 29392 (8.1%)
Players rated above 1700 = 169425 (46.6%)

March 2011
Total rated players = 126693
Players rated above 2600 = 223 (0.18%)
Players rated above 2150 = 29575 (23.3%)
Players rated above 1700 = 111529 (88%)

Again it looks like it is relatively harder today to get to a higher rating than it was 10 years ago but that might just be a reflection of the expansion in the number of FIDE rated tournaments for lower level players. But, yes, it does look a bit like rating deflation to me.

0
16

Everybody seems to agree that "ELO inflation" is real (I found an article from 20 years ago claiming this exists)...except scientists. Here is a 2011 paper that vehemently denies the phenomenon; the abstract says that only little inflation happens, and the players really get better. https://ojs.aaai.org/index.php/AAAI/article/view/7951

3
  • 2
    I would be careful about saying everyone seems to agree ELO inflation is real, since there are good reasons to believe it isn't real - in particular, players of today understand chess much better than players of the last century. It shouldn't be surprising if Carlsen were to crush Steinitz in a match, for example. Give Steinitz enough time to study and work with computers and he might very well beat Carlsen, but the Steinitz of 1885 logically should lose badly.
    – Allure
    Mar 23 at 6:33
  • @Allure: "everybody seems to" was a subjective distillation of my google attempts. I found no "no, careful research shows..." hit - except the paper I linked. Mar 23 at 7:49
  • 1
    "Everybody seems to agree..." [citation needed]. This claim is highly dubious without justification given the dogmatic wording. Apr 23 at 16:43
8

This is actually a very interesting question! Actually, there were not many 1700 rated players 50 years ago because most ratings started above 2000. With the evolution of computers, it is very plausible that a 1700 player now is much stronger than 50 years ago.

Nowadays, openings are spoon-fed and many references are available whereas those days every day you get someone playing novelties. And players had to read books and analyze on their own. Imagine having to learn Russian to learn about your favorite opening.

But it is important to note that playing styles have also evolved. 50 years ago, as openings analysis was limited, players depended on attacking and pressuring the opponent, whereas nowadays players have many resources( thanks to engines ), and attacking is very dangerous.

Talking about grade deflation, currently, there are a lot of players and a very high gap in terms of skill compared with an active Grandmaster. Considering the number of players and mix of people involved in playing chess, there is no doubt that grade deflation will keep on continuing. Although this can have a negative impact on bringing new players into the field.

Finally, I would like to point out the strength of players differ according to the geographic region too. For instance, countries in south Asia have a lot of skillful yet underrated players.

But in general, a 1700 player now has the upper hand over a 1700 player 50 years ago. Back then chess was an ART of its purest form. Nowadays computers have molded it into something completely different.

5

According to Dr. Ken Regan, no. Players at the same rating even across different eras are approximately equal skill. In other words today's Hikaru Nakamura is stronger than the 1972 Bobby Fischer. This shouldn't be too surprising, given that Nakamura was able to study Fischer's games and learn from Fischer, but not vice versa.

The methodology is by matching the players' moves against computers.

Only players alive during the same period can play each other. But since Regan's method compares moves to a common standard (the engine), rather than the results of games, he can objectively relate player abilities across eras. What he found was that rating inflation does not exist. Between 1976 and 2009, there has been no significant change in IPR for players at all FIDE ratings. Figure 5 shows, for example, how the IPR for players rated between 2585 and 2615 has remained relatively constant over time. Today's thousands of grandmasters and dozens of players rated over 2700 indicate a legitimate proliferation of skill. Thus one may conclude that Hikaru Nakamura's peak FIDE rating of 2789 beats Bobby Fischer's peak of 2785 for best American chess player of all time, and Magnus Carlsen's peak rating of 2881 places him as the best human chess player of all time.

Source

2
  • so shoulder of a giant basically?
    – BCLC
    Nov 20 at 0:39
  • 1
    @BCLC yes. 12 chars
    – Allure
    Nov 20 at 5:34
2

My pet theory is that players in isolated communities (e.g. from different eras, or from one platform to another) can be compared by considering the probability of a draw. We see in high level play that draws are quite common, with more than half the games ending in a draw. High skill players are less likely to make mistakes that lead to a loss. In my study of games played, I found that the probability of a draw was increasing in Elo for evenly matched players. This kind of analysis might be usable to compare current to past players.

1

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.