A look at the Elo ratings in the year 2021

In the years 2001-2012 the rating floor was lowered step by step. This process was guided by rating conferences to find the optimal K-Factor for the many players who got now enabled to enter the rating pool. Open discussions from that time can still be found in the Internet. When we ask nowadays if these changes have been effective, it seems to be difficult to find answers. As this subject was no longer in the focus of the open interest I took 2 years ago the chance to write an article about possible reasons of underrated juniors.

In the meantime I contacted some experts who took part in the above mentioned discussions and conferences. The FIDE Qualification Commission sees no possibility to act in case countries having an island-effect, with big parts of the rating pool being isolated from other countries. The Indian Chess Federation stays in silence. Only Jeff Sonas, who saw in former times the danger of a rating system being not in balance more on the side of inflation, sees meanwhile too a deflation-effect, “...as the younger/weaker players spent now their improving years inside the rating pool.”

As there is still silence I want to use the chance to complete my former article by making the rating pool more transparent with some diagrams. I am a fan of the Elo system which has given the chess-world a good compass in the last 50 years. This search for finding imbalances in the ratings between countries is also asking if those developments deserve more attention.  Being not an Elo expert and not a FIDE insider I hope that what I found shows correct numbers.

Maybe I give in this article too much importance to the demographic development of chess players in a country. Maybe this is because ... the 12 year old Vardaan Nagpal, my first opponent I ever played in India, went 3 years later to Europe and ended in Elo heaven, within 2 months his rating was rising from 2015 to 2501! Because of such results FIDE lowered later the K-factor for juniors depending on the amount of games being played within the monthly rating period. While I had 3 years later an unlucky start in my second Indian tournament and ended in Elo hell. In other words in the junior department of Chennai.

The Data used here is taken from the FIDE-Download area. The statistics are based on the January Lists of the years 2001-2021. There are about 8500 players with a rating but without the year of birth, those players are not in the age-based diagrams.

Table 1: the lowering of the rating floor

Don’t get confused when you see in the following diagrams the Elo level going down!

This is caused by the reduced rating floor. In the last 10 years the worldwide Elo average went down 300 points, this has nothing to do with Deflation!

But is Deflation or Inflation at work, when the level of the Top100 is changing? Here I am more with GM Bartlomiej Macieja from Poland, who gave at the FIDE rating conference 2010 the statement:

“What I would like to measure is what happens with ratings of players who keep their level. If their rating increases, I call it inflation, if their rating decreases, I call it deflation.”

The effect of the reduced Tournament-Scene in 2020/2021

The cancelling of many tournaments put the chess professionals in great distress. The effect it had on the youth can easier be made visible. In the last years about 15,000 U18-players received each year a first Elo rating. Although at the end of the year a whole group exceeds the age limit and falls out of the U18, the number of young players with a Elo rating was steadily growing. January 2021 now shows for the first time a decline, as there were in 2020 around 11,000 less new entrants in this age group than in previous years (diagram 1). In the entire period of reduced tournament scene from February 2020 to May 2021 about 14,000 juniors less received an Elo rating, compared with the trend from previous years. In view of the booming online chess, this decline should quickly be compensated when the situation normalizes again.

Diagram 1: Amount of U18 Elo rated Players 2015-2021 (active & inactive, January list)

In 2021 the active-status was temporarily exceeded to 24 months, for that reason it would make no sense to show players in diagram 1 with active-status.  Also the number of junior players would rise if they would be counted at the end of the year (Jan2019=69977, Dez2019=84790).

Obviously the young people are using the opportunity to get an Elo rating due to the lower entry hurdle, which has caused the U18's Elo level to drop each year. In order to filter out this overlapping effect of a falling overall level during the course of the year, only those players are considered in Table 2 who are in the list with “active” status at the beginning of the year. For those players the average rating after 12 months changes as follows:

Table 2 : average change of the rating by age groups during a year (players with active-status)

As the playing strength of many juniors wasn’t measured since march 2020 this may have caused in this group a gap between playing strength and rating.

The active-status and the inactive-status

The FIDE rating list January 2021 includes 362,000 players, among them there are 188,000 with the active-status. “active” means usually that the player took part at least in one Elo rated tournament in the previous 12 months. Because of the reduced tournament scene FIDE exceeded in November 2020 the period of the active-status from 12 months to 24 months.

In the rating pool of the inactive there are still players to be found who died a few years ago; this probably needs an adjustment (by the countries not by FIDE).  The world ranking-lists are based on the players with the active status. Diagram 2 compares both groups, as there is not much knowledge about the inactive block. The inactive block has less evidence, as the players who became inactive are summed up each year. As the average age is just 39,6 years it is likely that a part of the inactive-players are only temporarily inactive.

Diagram 2:  Amount of Elo rated players and Ø-Elo depending on Age

The lower rating floor led to a lowering of the Elo level. This has nothing to do with Deflation! The rating floor is a kind of filter on the worldwide chess-community, a window that should be wide open.

In diagram 2 the increasing number of Elo rated juniors is taking attention. Some of them became later inactive, maybe because of bad results, or maybe because family and job got a higher priority.

For players with the status "inactive" should be considered that some have been inactive for a long time and have frozen their rating so that the normal drop in rating did not occur with age.

The upcoming of isolated groups in the rating pool

The group of chess professionals and halftime professionals is less limited by country boarders, they play a lot among themselves in worldwide tournaments. FIDE is even supporting a mixing of countries with some rules for title-norms. However, as the rating spread to more countries and age groups, the formation of isolated areas has inevitably increased. Such isolated groups can lead to undervalued juniors.

See the recently published results of the Youth World Championships.

A kind of isolated group by choice can be found in Senior-chess, mainly in Germany, the country has the most active tournament scene in the age-groups of 50+ and 65+. If these seniors mainly play Elo rated tournaments in those age groups, then the rating level is not going down by age as usually (diagram 3 and 4). I too play those senior-tournaments but like the others not only because of the “friendly” Elo level. In view of the global population development, senior chess will become more important.

Some countries choose to stay outside the Elo rating pool, by history or just because they want to stay independent. The advantage of a national rating system is their flexibility. Here are some examples:

USA: In the year 2020 the US-Chess Federation reported 93000 members. Chess in school is booming, but still the USA has only 6701 members with an international Elo rating. The USCF rating system dates back to the year 1960 and is based on the system by Arpad Elo. FIDE adopted this in 1970, meanwhile they developed differently. It is said, that the USCF rating rules became complicated. Here an example: If an FIDE tournament is known to be a youth event, such as the World Youth Championships, then the conversion which is used to update the USCF rating is using a bonus, the bonus is getting higher when the FIDE-performance is getting lower. Which indicates that very low Elo rated Juniors are likely to be under rated.

China: On Wikipedia you can find an article called “chess in China”, which gives an overview how chess in china has developed. It is not a big surprise to find there “the big dragon project”, which has long term goals. The organisation is based on chess-academies, with more than 1000 trainers. In Malaysia I too saw a similar structure which is not based on chess-clubs but on chess-academies. The Website of the Chinese Chess Association is complete in Chinese.

England: a strong national chess-team but only 3614 players got an FIDE Elo rating?  The English Chess Federation has about 12400 members. Brian Valentine, the rating manager of the ECF told me about the specialities of the English chess scene. According to him the reason for having an own separate rating system is because FIDE does not cater for English recreational chess. English chess comes from a tradition of mid-week evening chess where law modifications make such games ineligible for FIDE rating. Many weekend events are FIDE rated but that applies to a smaller pool of players. Even here some of the lower sections are not FIDE rated to attract local players who have no interest in an international rating.

Germany: The „Deutscher Schachbund“ (DSB) reports at the beginning of 2021 about 90,000 members, with about 7500 female and 24,500 juniors. Germany established the first national rating, the Ingo system, which started in 1948. Since 1993 it is replaced by the DWZ (Deutsche Wertungs Zahl). The DWZ can be compared with Elo. Most of the juniors have a DWZ below 1000; those don’t show up in the following diagram! Also not those players who are a member of German clubs but belong to foreign nations.

Diagram 3:  Comparing of Elo and DWZ of German players

In diagram 3 you see on the left the very individual age-structure of German players. Usually only the stronger players got an Elo rating, so their Elo level is higher than the average DWZ, as shown on the right.

Diagram 4: Players with DWZ and Elo

Diagram 4 shows those players who have in the DWZ-List a FIDE-ID, so both ratings can be compared.

This connection works for about 20,700 among the 27,106 Elo rated players. Until the age of 25 both ratings are going hand in hand. With a growing age these players focused more on DWZ-rated tournaments, usually the team-championships, which are not Elo rated in the lower classes.

In the DWZ rating rules there can be found a junior speedup-factor, a stepped lifting of the first rating when it is below 1000, and a non age-based braking parachute. All rules to keep up with the increasing playing strength of juniors or beginners. The English system also tackles this problem but from a different angle.

The organizer of an Open-tournament in Germany has to pay 3 Euro for each participant to get the tournament FIDE rated, so for an Open with 300 players 900 Euros have to be paid just to get FIDE rated. Such coasts seem to be a hindrance for getting more Elo rated players. Asking FIDE why they charge such high fees, I got told that FIDE takes 1 Euro per player in a tournament. So 2 Euros go to the DSB, which shows that the national associations are not a good supporter of spreading the Elo system.

Probably the only problem caused by isolated groups can be found in some upcoming chess-countries.

The lower rating floor and an active tournament-scene inside those countries enabled now many juniors to receive an Elo rating. Juniors who stop playing because of bad results may add points to the pool.

But mainly the remaining juniors (K=40) need to play against adult opponents (K=20) to rise the rating level of such a country until those juniors become grown-up.

The following diagrams take a look at the countries with the highest amount of Elo rated players, including the USA and China as they have very strong national teams.

Diagram 5: The countries with the highest amount of Elo rated players (January 2021)

Russia, India, Germany, Spain and France are the countries with the highest amount of Elo rated players.

About 60% of the Elo rated chess players are Europeans. The low number of Elo rated players in the USA and China don’t correspond with their national teams. A strong increase in Elo holders can also be seen in Turkey (8414) and Iran (9719), which are in 5th and 6th place according to the U18 share, ahead of Poland and Germany. Surprisingly Spain and France have worldwide the most active Elo holders!

Diagram 6: Countries with the highest amount of Elo rated players – youth part in % (January 2021)

China doesn’t show up in diagram 6 as they have a high percentage of juniors but a very small amount of Elo rated players. Russia has a higher percentage of young Elo rated chess players than India!

Diagram 7: Countries with the highest amount of Elo rated players: Elo level in dependency of age

After France and Spain have drawn level with Germany in the top 10, the gap in the overall level is surprising. This could be due to the DWZ alternative, if Germany had more Elo holders, the Elo-Ø would also decrease. But even a non-expert can see that the increase in India is less dynamic and ends much earlier. If this is a consequence of the high U18-proportion, why can something similar not are seen in Russia? Perhaps such age-dependent average values are not suitable for deriving causes from them, it should be borne in mind that the lower the evaluated game levels, the lower the overall average.

The influence of the rating floor and the K-factor on the ratings of juniors

This report is more based on adding up numbers and building averages. For those who like it more scientific, the paper “Understanding Distributions of Chess Performances” (by Regan, Macieja, Haworth) is recommended. It was predicted as early as 2011 that the worldwide number of players with an Elo>2200 would hardly have any more increase potential. This was based on a model for forecasting population growth, or growth in general in biology. In fact, since then the number of players with an Elo>2200 has remained just below 21000 (active & inactive). But how does the U18-group get along since 2002?

Table 3: How many U18-players managed since 2002 to rise above Elo 2000? (Status=active&inactive)

Table 3 shows the countries that have already been considered, each with the total number of U18s and how many of these juniors managed to exceed Elo 2000. The threshold value 2000 is about the same value as the rating floor in the years 1993-2000.

The highest number of U18 players with an Elo> 2000 was around the year 2005. The lowering of the minimum Elo led in the following years to a significant decline of players above Elo 2000 – this seems to be the obvious explanation. If they previously had mainly adults as opponents, there are now more encounters among young people. The youth naturally shows the increase in skill level with a time lag in the Elo rating. If Elo rated juniors play against each other, this time lag doesn’t really get repaired, maybe it even get stabilised.

This development was counteracted in 2014 by increasing the K factor for young people and beginners to K=40. Since then the values have recovered, but so far there are still less juniors above Elo 2000 than in the year 2005. The sharp decline of U18 players in Russia is noticeable, reason unknown?! The only country with a steady growing amount of Juniors>2000 is the USA, this is not caused by the 40 Chinese-sounding names in the Top 100 list of the US-Juniors, the simple explanation is a low-rated Elo level in the USA. It is said they have not many opportunities to play Elo rated tournaments. so their rating is behind their increased playing strength.

Is there deflation or inflation going on?

The rating pool is a closed system. The amount of rating points that the winner gains are usually equal to the amount of points that the opponent lost. An inflationary effect arises if a player enter the pool with a starting Elo that is above his skill level, or if a player leave the pool with a lower Elo than his first rating, or if a player with a higher K-factor than his opponent exceeds his performance expectation. A deflationary effect arises in the opposite case. Or to put it more simply, adding points to the rating pool creates inflation, withdrawing points creates deflation.

Most of the Elo holders are quite young, so their time of leaving the rating pool is out of sight. Therefore the starting Elo and opponents with different K-factors are more important to keep the system in balance. As you can see in diagram 2 the Elo level of the inactive is above the level of the active when players get older. One reason may be the higher rating floor in former times, but it is also a human behaviour to use a kind of "window dressing" for finally getting inactive. Therefore, only players with status = active are considered here for the question of inflation / deflation. Taking the whole group of “inactive” not into account, ignores the part of the "temporarily inactive", but this would need a special definition of the players status.

The topic of inflation has been so far a bit often discussed in connection with the rating pool. Who would speak of inflation in the 100m run just because there are more athletes today who run this distance under 10 seconds? In athletics one speaks of a professionalization of the sport. In chess the number of players in the Elo range 2600-2699 is still increasing (2002 = 86, 2021 = 210). There is a motivation to reach this level, because at this level you can make a living out of chess at least for some years, not getting rich but being independent and see the world! If you consider the millions of members that chess.com reports, it is rather surprising that all other areas above 2100 with status = "active" show a stagnating or declining number of players.

If you imagine the entire Elo range as a band that is held at one end at Elo 1000 and at the other end at Elo  2900, then this band is thinned in the area of Elo 2100-2400, the number of players (with status=active) in this area has decreased in recent years. For women the same happened in the range 2000-2300, for U18 players in the range 1800-2100. To name the reason for that more examination would be necessary. Without knowing the deeper reasons I would speculate that a kind of stretching effect is caused either because the top players are rated too high, or the players at the bottom are rated too low. But it is more likely that the players at the bottom caused this effect because of the high number of players.

An attempt to check for deflation / inflation:

Here is a look at Elo rated players in the year 2015. In that year the players shown in table 4  all have status=active and aged 20-25 years. Because of the age this group should be able to keep their Elo level in the following 5 years, it is even more likely that they will increase their level. In the year 2020 the same players are shown again, just 5 years older. How did their Elo-Ø changed in these 5 years?

Table 4: Elo Development in the Age-Group 25-30 years

The significant decrease among women is surprising. In 2020 there are 17,362 women worldwide with Status = active, 59.8% of them are U18 youngsters! This is mainly caused by Asian countries. Among the female players of Russia and India more than 70% are juniors.

In the last 5 years, the number of Elo holders in the 1000-1500 range has tripled to around 120,000. The size and fast change of this number is perhaps meaningless, the decisive question is whether players start here with a rating that is too low?

Conclusion

The look at just a few chess-countries shows already a lot of individuality! It is not about levelling those differences, as individuality is not a hindrance for a system in which ratings of the same value from year to year and from country to country can represent the same proficiency of play. Overlapping effects allow this or that interpretation; one may see inflation the other one deflation. What does then “fair evaluation” means?

In the first step it means probably to examine "this or that interpretation" and to discuss it publicly. There seem to be some questions waiting for an answer: Which influence has the demographic development on the rating pool? Is it helpful to spread Elo rated tournaments if beside the FIDE-fees chess nations take an extra fee for a tournament being FIDE rated? Why do players become inactive? Does internet chess increase the gap between skill level and Elo rating?

The youth was not in the rating pool from the start; this integration process may need further adjustments. In addition, there are enormous differences from country to country in the age structure, in membership growth, in the proportion of players who only have a national rating, even in the form in which tournaments are held. It will also be interesting to see how the chess professionals got along with the reduced tournament scene.

Certainly this is not an easy task for FIDE through all these years. But in the rating system it seems as if FIDE focus too much on the chess elite. Maybe they are aware of the whole spectrum, but they don’t report about it. The regularly published rankings are always interesting to look at. More background information would also be of interest to the general public. What is FIDE's vision for the spreading of Elo ratings and what progresses and hindrances showed up along the way?

There will always be overrated and underrated players. The joy of chess should always be above the Elo number!                                                                                                   

Links

Are Elo ratings going down?