Personally I feel a bit tricked by the boost I've got (1871 to 1923, 52 units). 14 months ago I started playing after 17 years total break from chess. My goal was that I, despite the older me, directly could hold a rating of about 1800 and hopefully later, with maybe 30-40 games played even reach 1900. Now I've reached my goal for free so to say and, as Kostya said, as a 1900-2000 it's maybe now easier to go higher since people that had maybe 1530 now is over 1700. Well now I have to set other goals, maybe to be stable round 1950 (live rating now is 1936) and possibly as a vision to reach 2000 before I get too old. 😎 Well, well...
Jesse - the system won't take care of itself because the Elo system itself can't handle uncertainty in rating. The model that FIDE uses to gauge someone's rating just doesn't consider the fact that people can perform fundamentally stronger or weaker than their rating if they've improved/weakened a lot since their last rated game. Essentially, the FIDE system doesn't consider an important property that should be considered. USCF does. That's why this compression had to be applied. Rather than change away from Elo itself to something like Glicko-2 (like USCF does), FIDE decided this was a smaller change to help address the underlying issue.
The K factor in the FIDE system is the equivalent of RD in Glicko and expresses the uncertainty in the rating. The update rules for K are much simpler and the opponents' K isn't taken into account at all, but to say the property isn't considered would be incorrect.
Also he's been focused on the post pandemic when this problem has been here for over a decade at this point, it's just been put into hyperdrive in the last few years. I don't think he grasps that what's broken isn't the "deflation" but rather the relative spacing which is what ELO was supposed to be about. It's irrelevant if there is inflation/deflation but that the gap is not commeasurate with the model and it is widening. The way FIDE has chosen to address this and where they injected the points is the closest to trying to be what Jesse is asking, and in fact is likely to not really solve the problem because of it. There is no "eventually this fixes itself" when new cohorts of of newly rated players is being generated every year. Eventually ratings become meaningless and higher rated players simply refuse to play lower ones BEYOND the actual tapering off due to age (which the system can handle).
@@user-nh9so3sf3uThe K factor was there before and remains unchanged. Glicko models modify the K factor based on recency of last play which ELO never has before and since. I took the OP observation to be about how ELO doesn't modify the K factor with time (other than an age-based component), not be a complaint of K factor at all.
Also there is a prototypical player whose elo is the average of all elos. As more players enter the game, if you are below that average you will slowly drift down in elo unless you improve, but the suck is that improvement has less elo value as more players come into the game
Jesse is the Kramnik of the ratings rise. No matter how many times you explain it to him, he thinks Carlsen will be 3100 in a couple of weeks. Credit to the "ChessNumbers" dude for his patience.
The problem seems to be that nothing in this change will combat future deflation. People will still keep entering, improving, and then leaving with those extra gained elo, and deflation will occur again. You have to introduce bonus points to stabilize the system (the USCF does this). Otherwise reaching 2500 for GM will mean different things at different times and eventually become impossible. FIDE relies on rating floors to slightly reinflate numbers, but this seems artificial and imprecise
@@ThortheMerciless that doesn’t prevent deflation. There has to be a way to feed points into the system when people improve, and remove points when they degrade.
@@Derrickthepeng It does prevent some deflation, by preventing overly low ratings. Obviously the whole system needs to be changed, but equally obviously FIDE are not going to do that.
Adding points helps the stronger players, because they won't lose as many points if they lose to underrated players. Kids get better faster because.... engines and databases. Plus the ability to play constantly. Information is everywhere. During the Fischer era, you had to fight for info and you never knew if the info was correct. You saw this when better players defended their judgment against engines.
If the issue is that rating gaps weren’t resulting in the expected outcomes, instead of changing the ratings, why not just change the expectations. In other words, just continue to play under the paradigm where 1900s beat 2400s 10% of the time
Also imagine playing 10 open tournaments 10 rounds each, with 1 section only. Each tournament has 100 players of all skills. At the end of all 10 tournaments, there will be a bell curve distribution of players' win/loss. Then apply elo based on that distribution where the best player is 2800 and the worst player is zero
Although adding FIDE rating points all at once seems clunky to me, it was probably more-or-less the right idea. My USCF rating is 1692. My FIDE rating boosted from 1387 to 1642. 1642 FIDE seems much more realistic to me.
Can someone tell me if I’m understanding how this works? I think the problem is that when someone starts they bring with them so many points, a thousand or twelve hundred in the past, but if one person joins with their thousand points but a skill level equal to two thousand, then we need two people to join with skill levels equal to five hundred points to provide the points for that other guy to go up. If enough people aren’t joining with skill levels under that starting point to offset the people coming in with skills over the starting point then there will be a deflation in the rating system. This basically serves to offset the effect of so many stronger players coming in over the past few years by giving more points for strong players to win and adjusting for the decrease in players with skills below their initial point bucket. Is that basically accurate?
You're close but not quite. People can learn and get better, but FIDE's system isn't really built for that. Your first FIDE rating is your performance rating once you've played 9 FIDE games, but after that you move up too slowly. And in many places there aren't many FIDE tournaments.
Thanks for this, I was wondering if underrated juniors are a problem globally or only a problem in certain regions like India? Here in Germany we have the 2nd highest average age in the world and FIDE ratings feel a bit inflated here to me now.
It's clear why Kostya is positive and Jesse is negative about it. One wants to become a GM and have a fair and proportional shot at it comparing it to older generations. The other is a GM and wants his elite group to stay as small as possible, since this of course also has massive implications for the amount of money you can charge for lessons etc. The more GMs there are, the less special it is (compared to now with decades of deflation), even if it is proportional. This move by FIDE will have a very real impact on the financial situations of pros.
I just joined the dojo and was also confused because of this. My fide rating is now more than 200 points higher than my german rating (dwz) - so I switched my preferred rating to dwz and downgraded my assigned cohort - I think its better to start lower...
Of course it's in FIDE's interest to keep the percentage of titled players constant. That way they can stay relevant by giving out titles. Their function now is organising the WC and titles.
I get, and applaud, the anger towards FIDE. At the same time, keeping the percentage of titled players constant is only natural. Your absolute skill doesn't determine whether you are a master or not, rather your skill in comparison to the skill of others. If only the absolute skill was taken into consideration then you could easily argue about retroactively removing the "master" titles of old players. How can you say that Anderssen was a master when compared to today's players? So, keeping the percentage constantly is logical. The fact that it aligns with FIDE's wishes is... unfortunate.
They address this during the video. No, USCF is not based on Elo and instead uses a system that can better handle uncertainty of ratings. So deflation hasn't been as pronounced in USCF as FIDE and therefore there is no fundamental compression that needs to be applied to the USCF ratings.
@@michaelf8221Saying Glicko isn't based on Elo is a bit weird. The underlying models are the same, the update rules for the uncertainty are just more advanced (and complicated). Same for USCF.
In terms of ChessDojo cohorts: Maybe you should recommend to base the decision of choosing a cohort on their online ratings, since those haven't changed. Since people play much more online then OTB, I assume they are more accurate anyways.
This, among so many other reasons is why you CAN NOT base education on performance ratings. Imagine skipping from geometry to calculus because your high math bowl scores qualified you for a higher cohort .... may work for some, almost certainly a terrible idea in general. You guys have the correct method staring you in the face with your logo - use kyus (belts), not ratings, and NEVER equate the two. Just because I happen to defeat experts regularly doesn't mean that I don't need to learn how to checkmate with a rook and a king, right??? Knowledge builds on knowledge, not on competitive results. Martial arts do not earn belts through competition, the belts reflect a standard level of knowledge and determine whom they compete against ... there are plenty of black belts who have never won a martial arts competition.
Years ago I got onto the FIDE rating list as new and unrated. Never having played OTB since, I am decades later listed as new and unrated. This boost huh? I become ultra new, super unrated like some high grade engine oil?
Why can't there just be a z-score over whatever rating system is in place and just place people into classes based on that? That would also normalize all federations
Is it possible to calculate present ratings based on games databases? And if so, can you then calculate multiple rating types? ( so you can have a recalculated Elo and even Glicko for any player in history up to the present)
The point Sonas argues is not that the number should remain constant, it's that if the total population of players is increasing then the number of 2600s should (slowly) increase as well. In fact it was decreasing for many years, hence he concluded there was a problem.
I agree. My point was to Jesse. Jesse seems to believe that the if the number of 2600s increase (or the number of GMS) that it dimishes the achievement. The reality is the number of GMs should be proportional to the number of players. Just because there are more GMs than 20 years ago wouldn't indicate that it is easier to become one.
@@bruceelio1606Well, it's obvious that Jesse wants it to be harder, because he already achieved it. He's already part of the club. And this also has financial consequences. If more people become GM, he is going to be less "special", meaning he might not be able to charge as muss for lessons etc. That's why you see the difference in attitude between Jesse and Kostya: Kostya wants top become a GM, Jesse wants to stay part of an elite group that is as small as possible. Financial and other interests can clearly influence your opinion on this rating change, and this video is a perfect illustration.
so they raised the rating floor and made it possible to farm players again instead of adjusting the deviation or provisional metrics? I don't see how this stops people progressing in their national federations not playing many FIDE games, or the 8 yr old at the world blitz who has only been playing for 3 years. Sounds to me like they want to appease players like HIkaru complaining about losing points to underrated players. The olympiad was pretty shocking and its probably why Hikaru didn't attend the last one, but all it proved to me was ratings are politics, easily manipulated and chess is too exclusive. This is the only sport I know of where ratings even determine rankings, its absurd and I believe its a major reason western society doesn't sponsor it. I'm with jesse, I'd rather let everyone's ratings crash and burn rather than ratings and titles becoming even less meaningful then they already are imo. Let guys like Hikaru cry a river. let the top pros stay scared of open tournaments. and let all these young kids come bulldoze their fake inflated ratings imo. The queens gambit and covid boon is long over with. I don't see what the problem was.
Kostya: Will the points trickle up? I think the points will trickle up. 8yr old 1900 at Kostya's next tournament: Not so fast; Drop them drawers, bend over and spread!
It's the first time ever I had the feeling you didn't say anything. You were bringing "facts" without saying what you felt was right, like you were affraid of some back hits?