Chess.com Proposal: How to Handle Growing Bullet Ratings

Ratings are not comparable in time or between styles.  Time inflates ratings as they tend to be lognormalish - so new players get and averagish rating which contributes to the "pool" of ratings. Players leave behind bad ratings but re-enter at average and keep good ones inflated.  Maybe when a player leaves the pool, or doesnt play for a time,  you can remove their entire effect from the ratings by marginally degrading all of the players?

Just leave everything the way it is. You can not compare ratings across different systems, don't try to do it, it will never be perfect. Ratings are designed to compare players in the same system. So I don't see a problem with growing bullet rating. 2100 in bullet here will beat 1900 60-70% of the time, despite the fact that in reality the could have 1800 and 1600 OTB ratings respectively. The pools of players are really different in OTB and online systems.

Second reason the rating scale for bullet is really distorted. On popular sites like ICC and Playchess some GM players have bullet rating of 3600! Other GMs  have 2600  (1000 points less). In reality their FIDE/USCF ratings could differ by only 200 points. First, reason for it. Is that bullet is not real chess. It is hard to compare to OTB because the bullet rating is influenced by completely different factors such as speed of thought, knowledge of offbeat/tricky openings, etc. These factors might not be that important in OTB chess.

Second reason, most top players on these sites play only other 3000+ players, thus they essentially create a subpool of players inside of a larger pool. This makes the rating scale distorted from 2500 - 3600 rating.

It's funny to see how chess.com admins try to match uscf/fide ratings with bullet. It can't be done!

I agree, bullet rating are meant to show how good you are at bullet chess, not chess.

What about this- 

Restart everyone's bullet ratings at 0.

If you win against ANYONE you get 1

If you draw, you get 1/2

If you lose, you get -1

Also, players should have to play at least one higher-rated player every session.

ahhh! What kind of system is that?!?!

Well, it's only fair. 

Vyomo - I see your idea - why dont you run a little simulation of it and then you can see some of the issues.  I think you will find your system does not allow for intercomparibility of player strength.

Well, then let's say you get an asterisk next to your name if you defeat a higher leveled player

Also, all such games will be showcased.

For bullte, if you chop off 200 points then the "pros" will be upset unless you reset the glicko rd which will do nothing.

Chop!

Better to rip the band-aid off than to take it off slowly.

If the formulae are modified so that ratings will go down by design, the players are going to think that they have started to become patzers. Not good.

Chop.

If you chop 200 pts off, TigerLilov will be mad. D:

What it will do is that to tell the bullet players "Youre not so hot" and leave it at that.

I agree with the chop approach, however by a % of someones rating rather than a given number for all.

Speaking from the wrong side of 40 this topic may be for the young guns, with nimble fingers, speed of thought and a liking for pre moves.

Can the higher bullet to blitz ratings be attributed to technique and youth, I would say yes.

However if going for the chop 9% would align to blitz, but should this be increased to 18% to build in some rebound effect.

>:)

Chopping a percentage doesn't make sense, given what I know about ratings.

Adjusting by a given % is the only reasonable approach. However I believe the increase in bullet ratings can be explained by technique and youth, so the proposal to reduce the ratings needs further justification.

>:)

That's not how ratings work; two peoples' win chances against eachother are based on the difference, not ratio of their ratings; i.e. A 2100 beats a 1400 with a much higher probability than an 600 beats a 400.

If the inflated bullet rating is considered "un-attractive", then how to chop the values has nothing to do with how rating works. Many ways can be done, and all are fair. The best way is the one with less complaints.

Reducing by the same percentage for all players will dismotivate many players, especially those in lower level. And it is wrong theoretically, because their rating is deflated anyway.

Trying to be more "technical", you may try to find a way to reduce the spread in rating, by increasing lower rated players' rating and reducing higher rated players' rating. But this is not good from user perspective.

A good and easy way is to find a middle rating point where zero reduction will be applied, and then to (linearly) increase the percentage of the reduction. The maximum reduction is probably a value that will give equivalent rating with standard rating.

After that then a new formula can be applied (e.g. by allowing RD to come close to zero).

How about just reset all the ratings but you start with a lowish (lets say 25) glicko rating andas you go higher you gain less when you beat a person 200 points above you than a person that is rated much below you and beats a person 200 points above them.

I still think this is the most sensible thing suggested so far.

http://www.herbs2000.com/h_menu/det_history.htm

BLOODLETTING

Bloodletting was perhaps the favorite of the ancient treatments, but it was not used on the very young or old. The purpose of bloodletting was to cleanse and balance the humors by removing "bad blood." Bleeding was used as a cleansing technique in the case of abscesses, swelling of the spleen, fever, diseases of the mouth, eye, and head, headaches, and gynecological disorders. In the case of hemorrhage, whether from a ruptured blood vessel, wound, or childbirth, bleeding was used as a "balancing" treatment.
The most common method of bleeding was to open a vein to divert the blood from the problem area or to use freshwater leeches, a milder method of bloodletting. At first, ancient physicians performed all of the bloodletting. When the university schools of medicine were organized around 1000 A.D., bloodletting became the task of the barber-surgeons. Both bloodletting and surgery were considered beneath the dignity of the university-trained physicians, and surgeons and barber-surgeons were below them in status. Below them were the apothecaries and bathhouse keepers who frequently rented the leeches to their clients.
Cupping was another favorite bloodletting method. A small piece of hemp (tow) was burned in a cup. As soon as it had burned out, the cup was placed over a cut on lightly greased skin. Suction from the cup caused it to fill with blood. If the skin had not been cut, the cup was left in place until it fell off, producing a blood blister. A cupping glass over the stomach was considered to be an infallible cure for seasickness.
Bloodletting began centuries before the birth of Christ and remained popular into the 19th century. Every civilization in the world has practiced bleeding. It is still practiced today, even in North America, but on a very reduced scale, and for more practical reasons. Leeches are valuable for removing blood from bruises and black eyes, and for removing the congestion from around a reattached amputated limb.

Not sure if my last posting has anything to do with this topic at hand, lol

1. leave it. this isn't a good option in my opinion. ratings will continue to bloat and be more and more silly. 

2. slowly bleed them back down and then stabilize with adjusted formulas. 

3. chop them by 200 points and implement new formulas (we would also chop the "highest rating", "average opponent", etc)