Current Rating System needs a change

@ytroitsky said in #19:

But first of all, many people in forum think that the current ELO system is completely fine and it's usage should be continued, while others want to get rid of it. The ELO system is, in fact, a very accurate way of measuring strength and a very useful meter - much more useful than other systems like Kyu and Dan (which was also mentioned in the forum). ELO is probably one of the best, if not the best, measurement of strength.

Glicko is not Elo. Glicko is absolutely ridiculous, no doubt about it.

Elo is only used by FIDE but is better, although it has it's flaws. It's certainly not an accurate measurement of strength at all
which is also not possible.
Still the rating systems could be much improved.

There is no need for titles. It's just an ego and moneymaker thing and does not reflect the strength of the player quite often.

@ColossusChess said in #21:

Glicko is not Elo. Glicko is absolutely ridiculous, no doubt about it.

Elo is only used by FIDE but is better, although it has it's flaws. It's certainly not an accurate measurement of strength at all
which is also not possible.
Still the rating systems could be much improved.

There is no need for titles. It's just an ego and moneymaker thing and does not reflect the strength of the player quite often.

I didn't know what Glicko was, and I searched it up online - it was meant as an improvement for ELO (according to Wikipedia), but I didn't look in-depth into it. It does seem cool though.
I agree that ELO does have it's flaws, but currently there isn't anything much better than it. I have to disagree though about your second statement about ELO, I think that ELO is an accurate measurement of strength, but it's just that people are inconsistent.

About titles, you may be right. But it works at a high level, so I was wondering if it could work at amateur levels. However, you are right that it would often not accurately reflect a player's strength. After thinking about it I think you're right that it might not be the best ranking system, but it's hard to find something where fear of rating loss wouldn't be an issue.

Any ideas on how the rating system could be improved? I have to admit that I am definitely not knowledgeable enough about this topic, and you seem like you know much more than me. Do you have any opinions?

This thread is a train wreck. Op should be muted for a week just for starting it, and another week for enjoying playing opponents rated 1000 points lower than he is. I lost all respect after reading that.

And ps, there is no world where I would want to play chess with a woman I was in a relationship with. Zero.

@dboing said in #18:

I have not raed the whole thread. but it depends what you want what people really want any measure of this game experience in some population to be used for.

There is the performance pyramidal tiering goal (or is it method, competition below and above, game is single event competitive rule, but then there is the social more than 2 people game of tournaments.

Each level has its needs. I can only talk about the single game experince one, as I do not have any motivation for tournament game. But in both cases we are talking about something to be repeated many times over ones life time (some significant subset of it).

It is in that many game context that I find it still makes sense, but for me it is about learning efficiency about my pairing experience. For others it might be target and motivator for their performance (win ratio basically) "carreer".

For the learning only carreer, I just want to be able to advertise to the pool I might be randomly paired with, a number that would allow the others, that are in my optimum pleasure and learning range (a band, it is said, perhaps +- 200 on lichess, although why not anyting, but knowing the slope of struggle quality that every move in that singel pairing might require, the work given the time control. etc.).

I can't talk about tournament life and meaning of rating system.. I like that it is a real number.. I would not yet imagine countable rating system, what would be a quantum of chess play quality? Maybe NNue knows.

Where I would pull your question into would be about not just having competitive pairing based statistics but having board feature based metrics that could decompose our progression and actually offer some "map" of our skill sets gaps.

our chess foresight fog density given a battery of standardized challenge positions sets that would explore and test our response in all cases toward a feature space decompositoin of the human chess learning problem from newborn to expert. it might have some relation with the engineX engine chess sub-space, but I think there is no science there yet.

The human chess science is needed. And there are some pre-cursors tools, that are underestimatd or underused data wise. I think for example. we could look into the better define chess theory of tactical patterns (static and dynamic).

lichess has been accumulating a lot of big chess puzzle data with a thematic decomposition. A lot of that data must be either continusally lost to the ehter, or hidden in their live databases.... Their priority as developper is service to individual pairs of players, so it is not a priority to develop in that direction. The imagination of futher lichess code development given the resources is not going to be spontaneously exporing such questions.

So i think you question has merit. (any question has, but I mean it can justify our continue many heads efforts to understand what you or anyone could mean or read through your first salvo.

i do have a possibly abnormal understand of what the word "forum" might mean). I will find time to read others takes.

I guess what my Idea is;

A.} Keep the point system because it has merit, is useful, and people like it.
B.} Introduce a new 'Player Pool' based on a term, not points, for Casual players. This wouldn't use/impact points at all. It would, effectively, be a bunch of 'unrated' games. What would set them apart would be that Casual players could select a general area that fit the quality of the opponent they want to play against.

Like Beginner, Intermediate, Expert, Pro, etc etc.

Then, like in my case, I meet another chess player that wants to play we can setup a game that won't have an Elo Rating displayed, so no one is gonna be intimidated, or ashamed, as a result of their rating.

@ytroitsky said in #19:

I've read the whole thread and it's a mess. Almost all of the comments were people arguing, but not about the actual question. Anyways, I have an opinion on this question and I would like to hear comments about it.

But first of all, many people in forum think that the current ELO system is completely fine and it's usage should be continued, while others want to get rid of it. The ELO system is, in fact, a very accurate way of measuring strength and a very useful meter - much more useful than other systems like Kyu and Dan (which was also mentioned in the forum). ELO is probably one of the best, if not the best, measurement of strength.

That said, ELO does have some problems. Despite being a very accurate measurement of strength, it can have many psychological issues and can lead to people misreading it. For example, many people see ELO as a completely accurate picture of their strength, and will not play rated games because they do not want their ELO (which they see as their actual strength) to drop, or they will be intimidated with facing higher-rated opponents because they will lose rating, and in doing so, they subconsciously think that they will lose actual strength. The counter-argument is that someone along the lines of who I described should stop caring about their rating - however, fear of rating loss is usually subconscious and difficult to get rid of (this has happened to me; it has taken a lot of effort to change my habits and I still sometimes come back to them).

I think that the chess community should keep the ELO system, because it is one of the best measurements of strength. However, I think that we should replace it as the main ranking marker. In many rating systems today in e-sports, there is a system of leagues where people grind through them and see how far they can go. However, this system is useless for chess because chess isn't a sport where people should have to grind to get to a certain rank. Personally I think that it would be an interesting idea to have many amateur titles which cannot be lost - once you reach a certain rank, even if your rating drops, that mark still remains. In essence, this tries to make the ELO system a secondary measurement - which people do not care as much about. I will say truthfully that I have no statistics on how titles would change people's thinking, but I think that because there is no way to lose it people will play much more, because would be no fear of rating loss. However, this idea does nothing against intimidation - it would still be intimidating to play against someone with a much higher title, but in that case even if you lose the game there would still be no possibility of losing your rank.

However, there is a problem on how to reach the rank. If the rank has to be reached via ELO, this brings us back to the original problem where people fear rating loss because if they lose they cannot reach the title. This means that there would have to be another system of how to reach the title, but that would confuse everything and could lead to an unfair system (for example, if number of games is involved, then people who have played for longer would have an advantage over people who haven't played as much). Personally I do not actually know what the best way to make usable amateur titles is, and I would be interested so hear some suggestions.

In conclusion, I think that we should keep the ELO system as it is accurate and simple, but we should make it secondary to another ranking system where there is no fear of rating loss, such as a system where people strive to reach titles which they cannot lose afterwards. There are a lot of professional titles, and it does seem to work for them, so it would be interesting to implement them in amateur ranking systems as well to see what the effect would be.

Any opinions?

Yeah, that was what I was musing for.

@Rally_Vincent said in #20:

Even if this was a serious thread - which it isn't - the 1.000 lower rated GF would have been demolished in game 1 and then not played a second, so if this indeed was a serious thread - which it isn't -, OP would have lost a single game.

This is before applying internet memes about chess nerds.

That is a logical fallacy - there is no way anyone could know that would have been the outcome.

@ColossusChess said in #21:

Glicko is not Elo. Glicko is absolutely ridiculous, no doubt about it.

Elo is only used by FIDE but is better, although it has it's flaws. It's certainly not an accurate measurement of strength at all
which is also not possible.
Still the rating systems could be much improved.

There is no need for titles. It's just an ego and moneymaker thing and does not reflect the strength of the player quite often.

I always thought the point system as used is wonky. Sometimes I gain +5 and my opponent loses -24, sometimes I gain +4 and they lose -6, and of course the reverse. Just always trusted that there was valid math reasons.

If there is merit for overhauling the existing points system then that is kinda a different topic, as this thread is primarily about creating a new category for casual play that doesnt use points.

@ytroitsky said in #22:

I didn't know what Glicko was, and I searched it up online - it was meant as an improvement for ELO (according to Wikipedia), but I didn't look in-depth into it. It does seem cool though.
I agree that ELO does have it's flaws, but currently there isn't anything much better than it. I have to disagree though about your second statement about ELO, I think that ELO is an accurate measurement of strength, but it's just that people are inconsistent.

About titles, you may be right. But it works at a high level, so I was wondering if it could work at amateur levels. However, you are right that it would often not accurately reflect a player's strength. After thinking about it I think you're right that it might not be the best ranking system, but it's hard to find something where fear of rating loss wouldn't be an issue.

Any ideas on how the rating system could be improved? I have to admit that I am definitely not knowledgeable enough about this topic, and you seem like you know much more than me. Do you have any opinions?

I was thinking general categories like Casual Beginner, Casual Intermediate, Casual Pro, whatever. Though I suppose Casual Pro would be silly as those players will almost always be playing rated point games anyway.

@V1g1yy said in #23:

This thread is a train wreck. Op should be muted for a week just for starting it, and another week for enjoying playing opponents rated 1000 points lower than he is. I lost all respect after reading that.

And ps, there is no world where I would want to play chess with a woman I was in a relationship with. Zero.

Some are of the opinion that my topic {might} be trollish. Fair enough.

{This} reply definitely {is} trollish.

If it wasn't intended to troll and start an argument I would like to point out that, at no time, where you forced to read, reply, react, or in anyway become in involved with this thread.

So, if the experience was negative for you then that is a 'you' thing, not anyone else's fault.

@DeimosRuhk said in #27:

I always thought the point system as used is wonky. Sometimes I gain +5 and my opponent loses -24, sometimes I gain +4 and they lose -6, and of course the reverse. Just always trusted that there was valid math reasons.

There is a valid statistical reasoning for the asymmetry in points gained/lost, which most often has to do with the difference in rating between the players, and also with parameters called Rating Deviation and Volatility - you can easily look this up.

Even if you don't want to go into the details of this, there is an easy way to verify that the Glicko system works in practice: Just look at a sufficiently large sample of games against opponents that were in your rating range (say, +- 50 points before the game). In this sample, losses and wins should be balanced (around 50:50).

If you do this, you'll notice that the Glicko system on this site actually works really well.