Why chess ratings don't mean what they used to

@GamesJoblin said ^

Very interesting analysis. I might try to play some games in Spain, if I get the chance. Of course juniors remain rightfully feared.

I'm not sure I agree with the framing of the "deflation tax" as a problem that needs fixing. Geographical fragmentation is the problem. The deflation tax is the automatic solution by which the Elo system fixes these international discrepancies. When there is geographical fragmentation, overrated nations should bleed points to underrated nations. This is a feature, not a bug. As the initial shock of pandemic skill-gain fades, ratings will find a new equilibrium that accurately reflects player skill. This is a problem that solves itself.

Indeed, every single one of these analyses I've read is written from PoW of "the robbed". Future - and rightly so! - belongs to the youth.

The diminished relevance of an entire generation of players leads to attrition. As I wrote in the book, the FIDE Rating List is currently a museum, with 40% of the rated players in Classical not having played a single rated game since before the pandemic.

If there is no incentive to mix with these younger generations and exchange points, the net effect is deflationary. So it's not only that the younger generation is stronger and progresses faster, but they also lack the opportunities to "farm" points, instead cannibalizing each other in junior events. This cannot be overstated.

@Vlad_G92 said ^

As I wrote in the book, the FIDE Rating List is currently a museum, with 40% of the rated players in Classical not having played a single rated game since before the pandemic.

Players not having played a single game for more that a year are put inactive in the rating list. Be aware that between inactive players there are a lot of deceased players. So you should only look at the active players.

@xxRudi555 said ^

If in theory there is a gab between countries. Shouldnt there be also differents in the same country and in different states of that country?

In my country Sri Lanka yes, there is a rating gap between the parts of the country even though our country is small for example the players are much stronger in the Southern (Matara) and Western (Colombo) areas most of the time due to the lack of coaches present in the area too. The easiest part of our country is the Nothern side (Jaffna) but as I said there is no big rating gap but there is at least a 100 rating gap.

@mvhk said ^

As I wrote in the book, the FIDE Rating List is currently a museum, with 40% of the rated players in Classical not having played a single rated game since before the pandemic.

Players not having played a single game for more that a year are put inactive in the rating list. Be aware that between inactive players there are a lot of deceased players. So you should only look at the active players.

Deceased players are usually removed via data requests from the ratings officers of national federations. This should not be a big factor tainting the dataset. Whether all rating officers acquit themselves of such obligations in a voluntary/unpaid position is another question!

@Vlad_G92 said ^

Deceased players are usually removed via data requests from the ratings officers of national federations. This should not be a big factor tainting the dataset. Whether all rating officers acquit themselves of such obligations in a voluntary/unpaid position is another question!

I know in Belgium the rating officers have in most cases no idea if somebody deceased. The family of the deceased player are not chessplayers. They will not inform the rating officers. It can take decades before somebody eventually removes them from the "museum" list.
Besides there are also a lot of living people which simply gave up chess. Most never return to chess. Those are kept in the "museum" list just in case they would ever decide to return.

So you should definitely remove all those inactive accounts from your research. Garbage in is garbage out.

@El_Rey_Fer said ^

I don't know if this is a global solution, but at least can help a lot.

We cannot avoid some "elo islands" form, as OTB international tournaments are not as easy and available as playing online. But two simple measures that would help this problem:

• We are not in 1970 anymore. No more "monthly ratings". We can switch to a live rating where your elo gets adjusted after every game. Those underrated 1500 winning 10 points per draw will be 1550 in the middle of the tournament, 1600 in the last rounds, limiting the damage and adjusting faster to their real rating.

• Use Glicko-2 or a better algorithm that has a variable K factor depending on the accuracy of the rating. This will make some people not understanding exactly how many points they get per game (until they get use to it) but again, helps the rating to be faster to adjust.

but don't we still have it for FIDE ratings??

@Panda_SLCA said ^

I don't know if this is a global solution, but at least can help a lot.

We cannot avoid some "elo islands" form, as OTB international tournaments are not as easy and available as playing online. But two simple measures that would help this problem:

• We are not in 1970 anymore. No more "monthly ratings". We can switch to a live rating where your elo gets adjusted after every game. Those underrated 1500 winning 10 points per draw will be 1550 in the middle of the tournament, 1600 in the last rounds, limiting the damage and adjusting faster to their real rating.

• Use Glicko-2 or a better algorithm that has a variable K factor depending on the accuracy of the rating. This will make some people not understanding exactly how many points they get per game (until they get use to it) but again, helps the rating to be faster to adjust.

but don't we still have it for FIDE ratings??

Im waiting for this Saturday to have the March ratings to register for a tournament

@El_Rey_Fer said ^

I don't know if this is a global solution, but at least can help a lot.

We cannot avoid some "elo islands" form, as OTB international tournaments are not as easy and available as playing online. But two simple measures that would help this problem:

• We are not in 1970 anymore. No more "monthly ratings". We can switch to a live rating where your elo gets adjusted after every game. Those underrated 1500 winning 10 points per draw will be 1550 in the middle of the tournament, 1600 in the last rounds, limiting the damage and adjusting faster to their real rating.

• Use Glicko-2 or a better algorithm that has a variable K factor depending on the accuracy of the rating. This will make some people not understanding exactly how many points they get per game (until they get use to it) but again, helps the rating to be faster to adjust.

but don't we still have it for FIDE ratings??

Im waiting for this Saturday to have the March ratings to register for a tournament

I am gonna be unrated from this month's update I think
And that's kind of good

@Vlad_G92 said ^

Deceased players are usually removed via data requests from the ratings officers of national federations. This should not be a big factor tainting the dataset. Whether all rating officers acquit themselves of such obligations in a voluntary/unpaid position is another question!

I checked a few (since years) deceased amateur players I know about and none of them is removed from the list.

Another challenge are the masses of youth players, who particapted in one or two tournaments. It can take years, before they are updated out of the list. One could assume that this is a general problem. Too bad, in some regions of the world youth players play FIDE rated tournaments, while in others they do not.