lichess.org

Rating Hyperinflation

@Morozov In the full explanation above, I outlined that the use of the term hyperinflation was perfectly appropriate as:
(1) a cautionary nod towards historical incidents we would be wise to avoid
(2) an accurate quantitative descriptor where both
- [i] a strict criteria of 5% hyperinflation per day is used.
- [ii] two players choose to exploit the inflation algorithm for maximum self benefit.

Your counterargument relies upon:
- (A) A partial reading, since your rebuttal ignores the reasons for word choice that were explained fully in part two of my hyperinflation exposition.
- (B) An ad hominem attack referring to part one of my exposition as "wikipediaesque" or to be "copypasting", where [i] the reason for part one of the exposition was to compassionately remind everyone of historical precedent that we are liable to forget. It was not as a reason for the use of hyperinflation as a purely quantitative descriptor, and more importantly [ii] the term hyperinflation was justified as a quantitative descriptor in part two of the exposition.
- (C) You state that you prefer the definition of >50% per month.
- (D) You focus upon the historical average inflationary rate with respect to the whole pool, whereas I focused upon the potential for inflation where individual (or small groups of) players are incentivized to maximize self benefit as they are in any rating or monetary system.

Points (A) and (B) can be ignored as irrelevant or as requiring you to re-read the full exposition.

Addressing point [C], your preferred definition of 50% is much more modest than my >5% per day criteria. 5% per day converts to a 332% increase per month. Your 50% per month would only require a 1.36% inflationary rate per day. In the example in my previous post where 2 players choose to play each other exclusively to maximize inflation gains, Player A would only need to play Player B for 240 hyperbullet games per day to meet this rate, compared to my stated 1040 games for the stringent 5% per day rate.

Addressing point [D] in a monetary or rating system, players will always be incentivized to exploit the conditions for maximal self benefit, if they are aware of them. Incidentally, this awareness factor is also likely to be the reason it was wise not state publicly that the inflationary algorithm had been implemented or how it worked. While it is true we only meet the criteria for "inflation" rather than "hyperinflation" in terms of historical pool average, it is better to look at worst case scenarios given the incentives at work (an average chess players significant emotional investment exploiting the conditions to maximize his own rating). As an example in regular society, most people don't commit violently acts, but we still need systems in place as preventative measures as well as systems of justice to reach fair consequences when if they do. Likewise, in the glicko2 system it is equally important to take preventative measures against those players who may seek to exploit the system by accumulating 1vs1 match derived hyperinflation points and unashamedly peak sitting with them.

However, since all of this was an intentional adjustment that is about to stabilize, none of the above is relevant.

Warm regards, Burrower 🙏

Meanwhile on ICC the raiting is stable. I talked with one friend recently and he said that server is not suffering same phenomena there.

@Burrower no hard feelings but you are trying too hard to sound smart. Just admit that the use of "Hyperinflation" was not appropriate and leave it at that. If I wanted to read pages worth of pretentious nonsense I would review my own Uni papers.

Dear @Morozov , I wrote a rigorous analysis to show the word choice was appropriate. You have responded with yet another attack on character. I love science, so I am simply not interested in mudslinging.

A quote comes to mind that you may enjoy. Prolific American novelist Upton Sinclair once remarked "It is difficult to get a man to understand something when his salary depends upon his not understanding it".

Likewise, if we could transport dear old Upton through time to Lichess in 2019, he might remark "It is difficult to get a man to understand rating hyperinflation when pride in his rating depends upon his not understanding it".

For example, an analysis of hyperinflationary trends in your historical rating graph: https://i.imgur.com/vBnOH1L.png

We're all in the same ship of course, since everyone in the pool has inflated more-or-less equally.

However, we must also remember that since this was an intentional algorithmic adjustment that is close to stabilizing at p50 = 1500, any previous concerns and analysis about continuing inflationary trends are no longer relevant.

Warm regards, Burrower 🙏

#7 Agreed, this seems to be a correct conclusion (consistent with #2 and #14):

"However, since all this was an intentional adjustment that is about to stabilize, we can all sleep easy."

Usually I am eager to jump into rating-related discussions and #1 sets the bar high since proving a negative is difficult, doubly so because I vigorously advocated for rd decay changes so it may seem that I'm biased:

> We suspect that changes to rd decay earlier in the year unbalanced the glicko2 system and may have caused points to be generated out of thin air, or at least a trend towards a much higher stabilization point.

I have counter argument:
personally, only been playing chess for 2 years and 10 months, and constantly trying to improve, so far went from 1125 to 1850 in blitz.
lichess.org/@/hash11011/perf/blitz
i reached 1818 on 2019-03-19
i reached peak 1857 on 2019-11-08
now i am around 1820

bullet:
lichess.org/@/hash11011/perf/bullet
1673 peak on 2019-02-14
now still around 1630

my last 7 blitz opponents (picked randomly):
@lousdad
no enough old rating to get any useful data
@islandification
no increase, relatively steady over 4 years
lichess.org/@/Romantik1900/perf/blitz
did get better recently
@ricardo807
no enough data
lichess.org/@/radiohead1965/perf/bullet
no increase, relatively steady over 9 months
@Schachmakan
did get better recently
@AlfredoSaenz
steady stable increase over 2 years




What recently surprised me is getting paired with a FM in the 3+2 pool and he offered 3 rematches, probably due to my rating being about 200 points higher than I thought it was (which is itself humorous: some players claim not to get attached to their ratings, and I can't even remember my own rating).

I guess my rating can be a barometer for the entire system; if I'm rated 2200 and dropping pieces every game, everyone's 200 points "over-rated".

@Burrower Yes invoking science or whatever sounds nice does not make your "analyses" right. And you are severely mistaken if you think i give half a crap about bullet. Again please explain how you justify using the term "Hyperinflation" instead of trying to attack me für my bullet rating

This topic is now closed.

Reconnecting