Top 50 Chess Tournament Cities (Modern times)

(; Updated monthly; Last update 2014.10.01)

   The City ranking is based on a formula that we came up with to evaluate the significance of chess tournaments that happened in each city in the world. First, we came up with a formula that evaluates the significance of each FIDE-rated chess tournament. We have then sorted the cities in the world according to the sum of the significance of the tournaments that took place in them. Calculating the significance of a tournament need to be done with good care, that's why we have spent some time on this and came with a formula that is not that straightforward to understand, but aims to be as accurate as possible in determining such rating.

 N        City    (Click on city name for details) Country Significance*
2Khanty Mansiysk3096112
3Saint Petersburg2869818
6Wijk Aan Zee2240525
20Cappelle La Grande1215985
30New York850565
37Las Vegas699697
47Buenos Aires597043

* Evaluation criteria

The cities are sorted according to the sum of the significance value of the tournaments that took place in them. Before we come to how we calculate the tournament significance value we first want to list some considerations that led us to the formula we have choosen:
  • We do not want to take the average ELO rating per tournament as a metric for its significance (Hence, for example, we would avoid using metrics similar to FIDE tournament categories, based solely on average rating). The reason for this is that a tournament with several super grandmasters which also allowed participation of weaker players would rate much lower than tournament with only super grandmasters. We believe in such cases the tournaments should be "similarly" ranked, as strength of the participating players is one of the main criteria to have a quallity tournament.
  • If a city has hosted several super-GM tournaments plus many "weaker" tournaments it should rate higher than a city that has hosted a single super-GM tournament and no other tournaments. That's why we use the sum of tournament values (and not e.g. the average of all tournament values)
  • We do not want to just take something like the "top 10" tournaments for each city and rank by this. We believe a city that hast hosted many "medium-rated" tournaments, or tournaments with big amount of participants deserves *some* credit for this.
  • Due to the goal in the previous bullets the "scale" of tournament values becomes quite important. We want to avoid a situation where a city that has hosted many "medium-rated" tournaments gets better rank than a city that has hosted high quallity tournaments but falls behind just because it has not hosted "enough" tournaments.
  • A tournament like for example the World Youth Chess Championship (U8-U18), or the World Women Chess Championship would not have many top (or in youth tournaments even medium) players from the general ranklist as participants (due to the age or gender restrictions), but it still has to be considered of very high value. We would like to take this into account.
  • Finally, some grouping of players within tournament is needed, to avoid effects like World Chess Championship match (consisting of only two players) being rated much lower than a single super-GM tournament with 10 players.

Let's now also take an example. How would one rate a tournament with Magnus Carlsen, Vladimir Kramnik, Levon Aronian participating versus tournament with dozen 2600-ELO-rated players? It's natural that the former one gets a higher value, but how high? To make it more interesting, should a city that has hosted a single tournament with Magnus Carlsen participating be rated higher than a city that has hosted five 2600-rated tournaments? What about 10? To find this out, we base our calculations on the following: We determine how "rare" or "valuable" players are by looking into the ELO rating distribution (Normal distribution peaking around ELO rating 2000-2150). As of December 2012 there are only 3 players with rating greater than 2800, about fifty players with ELO rating greater than 2700, 245 players with rating greater than 2600, about thousand with rating greater than 2500 and over 8,000 players with ELO rating greater than 2300. Thus, we would argue that a tournament that has attracted ten 2700+ rated players would "count" roughly as much as five tournaments with ten 2600+ rated players.

Thus, we came up with the folowing estimation formula:
  • We calculate the ELO distribution per age group and gender.
  • The tournament significance value is the sum of the "significance" values of each participating player group (Group of players consists of the average of each 10 players in the sorted list of participants in a torunament).
  • The "significance" value of a player group is calculated as 1/X, where X is the probability of a player to have a higher ELO-group rating than the average in the group being valuated (ELO groups are multiples of 100 ELO points, i.e. 2000-2100, 2100-2200, etc.)

To summarize our analysis, the city ranking would be determined mostly by the amount of strong players that took part in tournaments organized in this city (or top players in age/gender restricted tournaments), nevertheless, this is further "fine-tuned" by all other tournaments in this city (both considering their sizes and players), and the latter becomes much more significant for the cities outside of the "top" range.

Data set (And current limitations)

We have evaluated all (more than 60,000 as of April 2013) tournaments submitted to FIDE. As only FIDE-evaluated tournaments are considered, we are missing in this ranking famous past tournaments, including world chess championship matches. Factoring those in the ranking would require further algorithmic research, thus we named the ranklist currently as the "Top chess tournament cities (modern times)". Additionally, we are still improving the evaluation, as about 5% of all tournaments could not easily be assigned to a single city or the information for the venue was unclear (We expect because of this that mostly several German and French cities are affected in a negative way, however we do not expect the ranklist above to change drastically). Finally, and quite important, the gender considerations is not yet implemented, which could actually drive some changes to the published ranklist.

References and Future work

Our future work includes providing supplementary metrics next to our primary index of city chess tournament significance; providing an overview of tournaments that has taken place in the city; providing extended ranklist (top 100, 200 or more cities); including gender considerations in the calculations as well as fine tuning and making sure all data and calculations are accurate and of highest possible quallity.
Reference: [1] FIDE Tournament Archive.