Rating floors for RHP

Continued...

The problem is inflation. Just as in the high street your dollar, euro, or whatever no longer is worth what it was 5 years ago, so rating points become inflated. Over time, what used to be a system where you could look back 5 years and say my rating has risen by 200 points since then, so I must be better, changes to a system where you need to find a benchmark. People would be able to say "my rating is 2201, so I must be a Master", whereas they have not improved at all since when they were an expert rated 2101 - the rating points are just worth less. And so someone desiring popularity would need to move the goalposts for being a master.

Note to self: Must post less.

Originally posted by gezza
Continued...

The problem is inflation. Just as in the high street your dollar, euro, or whatever no longer is worth what it was 5 years ago, so rating points become inflated. Over time, what used to be a system where you could look back 5 years and say my rating has risen by 200 points since then, so I must be better, changes to a system where you need to ...[text shortened]... pularity would need to move the goalposts for being a master.

Note to self: Must post less.

Anybody who thinks that minor changes in their RHP rating proves anything is nuts. Rating floors seem a reasonable solution to the tournament entry problem and the "sandbagging" problem. The effect on overall ratings would be trivial; you're making a mountain out of a molehill.

Originally posted by gezza
[b]Hint: Discouraging spite-resignations isn't the main point of rating floors. It is only an small side benefit. There is a more important consideration in play.

Really. What's that then?

Me:
You argue that the RHP rating system should be changed without understanding what it currently does? How amusing!

You:
This from the guy who quotes t The problem is inflation. Just as in the hig...[/b]

Really. What's that then?

Discouraging sandbagging, and avoiding damage to the ratings of the opponents of a player who is demonstrably, by past rating, hundreds of points underrated.

But you have not bothered to find out what is in use on RHP before saying it should be changed.

Shrug. It is not necessary to know the exact formula to realize that sandbagging and vastly underrated players are a serious problem. It is also easy to see that losing 1000 points due to mass resignation has zero correlation to chess skill.

Either he is sandbagging or not. The clan leader needs to decide, and decide what to do about it. It would be feasible to deny the person entry until their rating has stabilised again.

No, it is not feasible. The player is not necessarily at fault for the rating drop and thus should not be penalized by being denied entry into a clan. And it's so easy to fix. Add a rating floor to keep him in his proper rating class so that he can be fairly matched in clan challenges!

"Leave for a while" implies come back.

But it does not necessarily mean that the player will ruin clan challenges and banded tourneys when he comes back. He may just ruin some open invites. I'm not happy with that either, but I wouldn't call it sandbagging.

a TER floor solves the banded tourney problem.

I have not disputed this.

Edit: On second thought, I want to start disputing this now. A player rated 1000 points too low could be unfairly penalized by not being able to enter tournaments in his proper band [like 1800+]. Flooring the TER only does not go far enough.

After all, losing a game (against equally rated opposition) shows that your ability and your rating are not in sync. Therefore your rating should drop.

...until it hits the floor, then it should stop.

Why not? They reflect ability to manage time, which is, as you say, part of the game.

Nonsense. They reflect the player's need to stop playing chess entirely, at least for a time. You shouldn't claim that someone lacks an ability just because they stop using it for awhile.

The likelihood of winning is reflected in a rating as is, without a floor.

Let's say User 129166 returns to the site some day. We'll see just how well your as-is rating system predicts performance then, eh? How about a serious $$$ bet?

Just because you repeat it does not mean you are right.

However, it does refute your silly claim that I have provided no justification for rating floors.

However, if someone really, desperately wants to sandbag, they just have to convince an admin.

And the admin should not let them drop through the floor until they've been stuck on the floor several games. This is standard.

It is not inconsistent at all - I want the points when my opponent blunders - I have played better.

No, you had not in that example. You were at a disadvantage, but kept playing, and then your opponent made an odious blunder and then immediately resigned before you could even play another move. It would be pathetic for you to take credit for such a victory.

I want my rating to reflect pure chess ability only, so I make sure that time is not a problem.

Do you also refuse to claim timeouts?

I am aware that time can be a problem OTB, or in blitz, but to say that it is part of the game on RHP, when you can chose the number of games and the length of timeout/timebank you play here is stretching it.

Exactly my point, and I don't think it's a stretch at all. Time is still part of the game even if you never get timed out. Gameload, time controls, etc. all play a part in determining the quality of your moves. I have been surprised at some of the careless moves played by 1800-rated players here, but they seem to have a high gameload, so they end up playing it like blitz chess instead of correspondence! [I too have fallen into this same trap on occasion.]

He has NOT "actually lost the game" UNTIL he resigns.

Unfortunately for you, you specifically stated "your opponent resigns a bunch of games, with yours among them" (emphasis added) right before you said that he hadn't actually lost the game!

You managed to beat someone who resigns a lot of games, and his rating WHEN he resigned reflects that.

I'd prefer that his rating reflect something resembling chess strength, instead of spiteful, deliberate manipulation. 🙄

So does it mean he loses, or not?

Yes, a player who resigns a game has lost the game.

The core is that a loss (except to a much higher rated opponent) should mean that you drop rating points - you are on average less likely to win a game than before.

This is still too simplistic.

A drop of hundreds of points is warranted if your opponent loses hundreds of games. The consequence of losing a game (against a similarly rated or weaker opponent) is to lose rating points.

Nothing new here. I still think this position is too rigid, short-sighted, and unhelpful to victims of sandbagging.

The problem is inflation. Just as in the high street your dollar, euro, or whatever no longer is worth what it was 5 years ago, so rating points become inflated. Over time, what used to be a system where you could look back 5 years and say my rating has risen by 200 points since then, so I must be better, changes to a system where you need to find a benchmark. People would be able to say "my rating is 2201, so I must be a Master", whereas they have not improved at all since when they were an expert rated 2101 - the rating points are just worth less. And so someone desiring popularity would need to move the goalposts for being a master.

That's really what this is about, eh? Desiring popularity. Bragging rights. All meaningless egoism which is irrelevant to the point of chess ratings: to compare the strength of players, and predict the likely result of a game between them.

Originally posted by SwissGambit
That's really what this is about, eh? Desiring popularity. Bragging rights. All meaningless egoism which is irrelevant to the point of chess ratings: to compare the strength of players, and predict the likely result of a game between them.

While I couldn't agree more with this sentiment, I do feel that there is some small amount of pride in knowing you've reached a certain level. Having others ruin that for you for their own selfish reasons is ridiculous.

Originally posted by zakkwylder
While I couldn't agree more with this sentiment, I do feel that there is some small amount of pride in knowing you've reached a certain level. Having others ruin that for you for their own selfish reasons is ridiculous.

I don't think that rating pride would actually be ruined by inflation. So what if "master" becomes 2300 instead of 2200? Once you reach 2300, you can still claim to be a master.

Originally posted by SwissGambit
I don't think that rating pride would actually be ruined by inflation. So what if "master" becomes 2300 instead of 2200? Once you reach 2300, you can still claim to be a master.

With ratings floors, the average RHP player's ratings won't go up 100 points. 15 perhaps... but no way will everything go up 100 because a fat fist full of users tank 1 or 200 games.

P-

Originally posted by gezza
But they may do so again. Having a rating of 1900 would be wrong because the use in question is likely to resign the game. A lower rating is appropriate, because the resigner has lost a bunch of games. A 1900 rated player is someone who plays well, and consistently - if they do not, they are not rated 1900.

I am not sure what you are trying to say with your second comment line. Ratings change all the time based on whether you win or lose games.

Ratings are the skill of a given chess player, not the likelihood they will tank out any given number of games.

If a player is rated around 2000, chances are they will beat me. I'm not going to depend I'll be part of a batch of 100 games they time or resign in hopes of sandbagging or just taking on too many games.

Skill is skill, rating floors make sense.

Do you think everyone's ratings are 'deflated' by 100 points because of these few users? I don't think so, but by your thoughts this should be true in your mind. Shouldn't that be fixed?

P-

Originally posted by Phlabibit
Ratings are the skill of a given chess player, not the likelihood they will tank out any given number of games.

If a player is rated around 2000, chances are they will beat me. I'm not going to depend I'll be part of a batch of 100 games they time or resign in hopes of sandbagging or just taking on too many games.

Skill is skill, rating floors make ...[text shortened]... o, but by your thoughts this should be true in your mind. Shouldn't that be fixed?

P-

'fraid not. Ratings reflect past performance in games. As such, they are the best available indicator of future performance. Normally skill and ratings coincide, but there are cases where they do not.

When someone resigns or allows to timeout a bunch of games, then their rating ought to drop, as they did not win those games. Again, the past behaviour is the best indication of future performance.

Rating floors do not help that indication of future performance, and have the nasty side effect of causing rating inflation. When I look at ItsYouThatIAdore's graph (picked up by SwissGamibt above), there are a couple of small V's, where he won a few games, but the general trend was and should have been down. A rating floor would only hinder his rating indicating his resignation of games.

As floors have not yet been implemented, there is no rating point inflation *yet*. I don't recall mentioning deflation, so I have to interpret your comment - If the interpretation is not what you intend, please make the comment clearer. How much inflation or devaluing of a rating point there is depends on how many people resign lots of games, and hit their rating floor. I do not know how widespread mass resignations are, but you mention V graphs, and I know that I have also seen more than one. Depending on the number of people who actively play, there could be a noticeable effect. SwissGamibt picked up on ItsYouThatIAdore as an example of someone who had resigned a lot of games. Some of those where against players rated higher than I am, a group of a mere 500 people. If suddenly there are a lot more rating points being sent the way of a small group of people, then the effect is likely to be noticeable. It is hard to say precisely how much inflation there will be, without having access to all the data.

Originally posted by Phlabibit
With ratings floors, the average RHP player's ratings won't go up 100 points. 15 perhaps... but no way will everything go up 100 because a fat fist full of users tank 1 or 200 games.

P-

Where do you get this 15 from?

I said I was worried about 50 points per year. The truth is that I do not know what the effect will be. It could just as easily be more than 100 points per year, but I have no data on which to base an estimate.

Originally posted by SwissGambit
I don't think that rating pride would actually be ruined by inflation. So what if "master" becomes 2300 instead of 2200? Once you reach 2300, you can still claim to be a master.

The trouble is that it is now a continually moving goalpost. You won't know what level a "master" is - it moves year on year. Someone who clearly is that good (i.e. reached the level before inflation), could find themselves surrounded by weaker players who have the same "title", and who in essence just devalue it.

Originally posted by no1marauder
Anybody who thinks that minor changes in their RHP rating proves anything is nuts. Rating floors seem a reasonable solution to the tournament entry problem and the "sandbagging" problem. The effect on overall ratings would be trivial; you're making a mountain out of a molehill.

"Proves" is hardly the correct word. But a 200 point improvement in rating *is* an indication of some progress.

Rating floors may *seem* a reasonable solution, but there is a disadvantage.

How do you know that the overall effect would be trivial?

Originally posted by gezza
'fraid not. Ratings reflect past performance in games. As such, they are the best available indicator of future performance. Normally skill and ratings coincide, but there are cases where they do not.

When someone resigns or allows to timeout a bunch of games, then their rating ought to drop, as they did not win those games. Again, the past behaviour is rd to say precisely how much inflation there will be, without having access to all the data.

If someone is rated 2000, and they lose 100+ games their rating will tank. The next game they start, do you REALLY think they are now playing as an 800 player?

USCF has a rating floor for just this reason... to keep people from doing what they do here at RHP. Is the USCF wrong for doing this?

P-

Originally posted by SwissGambit
snip ....

Shrug. It is not necessary to know the exact formula to realize that sandbagging and vastly underrated players are a serious problem. It is also easy to see that losing 1000 points due to mass resignation has zero correlation to chess skill.

If you don't know whether the system in use is Glicko, Glicko II, or ELO, then you have little basis on which to comment. It is necessary as soon as you start criticising the detail of whether rating has to drop 400 or 600 points for the opponent to get nothing. Losing 1000 points due to losing 200 games is perfectly reasonable.

No, it is not feasible. The player is not necessarily at fault for the rating drop and thus should not be penalized by being denied entry into a clan.
Is too. So I say the clan leader decides, and takes the flack either way.

And it's so easy to fix. Add a rating floor to keep him in his proper rating class so that he can be fairly matched in clan challenges!

The fix is broken. It has knock on effects on the rest of the system.

But it does not necessarily mean that the player will ruin clan challenges and banded tourneys when he comes back. He may just ruin some open invites. I'm not happy with that either, but I wouldn't call it sandbagging.

"ruin" open invites? How? He is back and playing better than his current rating - it should rise. Players against whom he plays get a game against a strong player, helping them improve, just like the chap rated 1200 who has currently got an open invite open for >2000 points.

I have not disputed this.

Did too - you repeat it as a problem.

Edit: On second thought, I want to start disputing this now. A player rated 1000 points too low could be unfairly penalized by not being able to enter tournaments in his proper band [like 1800+]. Flooring the TER only does not go far enough.

Eh? Unless that player is new to the site, then the Tournament Entry Rating says which tourney band he gets to play in. Putting in a TER floor does exactly what it say on the tin - it puts a lower limit on the tourney bands you can enter.

...until it hits the floor, then it should stop.

So why should it stop? You still lose a game.


Nonsense. They reflect the player's need to stop playing chess entirely, at least for a time. You shouldn't claim that someone lacks an ability just because they stop using it for awhile.

So resignations should be a special case then?? Not like a standard loss? Laughs.


Let's say User 129166 returns to the site some day. We'll see just how well your as-is rating system predicts performance then, eh? How about a serious $$$ bet?


Given that he last moved a few minutes ago, he may wish to comment himself. Maybe it is a co-incidence that he moved when you mentioned his name. We will indeed see. But note that as his rating was falling, he also won a few blocks of games in a row. I contend that he is just as likely to stop playing again, as he has done in the past.


No, you had not in that example.

Oh. Your theoretical, has not happened, example.

Do you also refuse to claim timeouts?

I give more detail in my profile. I claim timeouts if I feel like it, but rarely in an interesting game.

Gameload, time controls, etc. all play a part in determining the quality of your moves.

So why do you insist that mass resignation because you cannot cope with the current game load for whatever reason should not affect your rating if you hit some "floor".

Unfortunately for you, you specifically stated "your opponent resigns a bunch of games, [b]with yours among them" (emphasis added) right before you said that he hadn't actually lost the game![/b]

I did not state the order. It was implied in the comment and then clarified for the hard of understanding. I notice a slight "clutching at straws" in your repeating the previous comment, rather than responding to it.

What about that example of "spiteful resigning"? Is that just another theoretical possibility to attempt to strengthen your arguments.

This is still too simplistic.

It is a fairly simple system - weighted difference of ratings gives a probability of winning. Past performance is used to give an indication of future performance.

Nothing new here. I still think this position is too rigid, short-sighted, and unhelpful to victims of sandbagging.

Nothing new here either. You want to "fix" something without understanding what it is. And I think your fix is broken.

That's really what this is about, eh? Desiring popularity. Bragging rights. All meaningless egoism which is irrelevant to the point of chess ratings: to [b]compare the strength of players, and predict the likely result of a game between them.[/b]

This is sure what it sounds like - you want rating points if your opponent resigns, even if he resigned 200 games before resigning yours. The only thing you care about is getting more rating points. Rating points which the "fix" you want to add to RHP will devalue - but it does not matter - you get more points and more bragging rights.

I want to know if I progress from one year to the next - I care about how I progress. Rating points are a way of keeping the score. I am not keen for them to be broken or devalued.

Originally posted by Phlabibit
If someone is rated 2000, and they lose 100+ games their rating will tank. The next game they start, do you REALLY think they are now playing as an 800 player?

USCF has a rating floor for just this reason... to keep people from doing what they do here at RHP. Is the USCF wrong for doing this?

P-

From what I gather, the US had a sandbagging problem. The US implemented a rating floor system. It fixed this problem for them.

But the fix appears to have led to rating inflation compared to the rest of the world.

Would it not have been better to say that you *cannot* win a prize if your rating has been established above the rating band for that prize? Would that not have had the same effect of discouraging sandbagging, without the inflation? Sure, you get to enter and win the low rated tourney, but the judge says that although you won, you get nothing apart from the game wins, because you're a sandbagger, and he gives the prize to someone else.

It is very easy on this site to carry 2 numbers for rating - current, and best ever - it is done already with current and TER. Making the TER depend on the best ever rating does not look too difficult. It is possible to improve on what the USCF has done, rather than repeat the same mistake of implementing a broken fix.

Originally posted by gezza
From what I gather, the US had a sandbagging problem. The US implemented a rating floor system. It fixed this problem for them.

But the fix appears to have led to rating inflation compared to the rest of the world.

Would it not have been better to say that you *cannot* win a prize if your rating has been established above the rating band for that prize ve on what the USCF has done, rather than repeat the same mistake of implementing a broken fix.

Check this out!

RHP subtracts 200 points from EVERY USER at RHP, and new users start at 1000 rather than 1200.

Now me, you, and everyone else is rated 200 points lower. Does that make us any less or worse a chess player?

A rating is a number to compare, I don't see the problem here or with USCF 'inflation' when all the number is telling you how you compare to other players.

If I'm rated 1450 or 1250, I'm still the same player. The number is there to show others how I play, and they'll know what the 1450 is or 1250 if the numbers were changed.

Inflation-shmation!

I don't mind your idea that high rated players could join, but not win any given event... but I still don't see a problem with inflation. We're not going to run out of numbers at the top, numbers are infinite and ratings will find their level.

Back when RHP started, the top rated players were rated 1700 to 2000 I understand, that's gone up too. Ratings will always go up and up, and there is no way to compare a number from now to a number from 10 years ago in any system. By nature, ratings will always go up.

P-