16 May 08
Originally posted by gezzaI think this is by far the lesser evil. Players won't be able to maintain an inflated rating for long if their play is not up to it.
Overall, more points get added to the system - giving inflation.
With the system as it is, the 2400 drops to 1800 or so, and then deflates his opponents' ratings on his climb back up to 2400.
17 May 08
The point about rating inflation is valid, so why don't we just put a floor on the TER?
The only minor point is that it sounds like the guy in the original post was talking about games he'd started from open invites. In this case, if you get someone with a rating below the TER, can an automatic message be sent warning the player of this, so that they can delete the game if they feel they don't want to play him/her.
17 May 08
Originally posted by Humph3Fiddling with the TER only fixes the problem for banded tournaments. The problem is much bigger in scope than that. A player who is underrated by 800 points causes trouble in ALL rated games, including clan challenges, open invites, and open tournaments. The worst case is open tournaments - there is no way to avoid playing the guy without hurting your tourney standing. [You could resign on move 1 to save rating points.]
The point about rating inflation is valid, so why don't we just put a floor on the TER?
The only minor point is that it sounds like the guy in the original post was talking about games he'd started from open invites. In this case, if you get someone with a rating below the TER, can an automatic message be sent warning the player of this, so that they can delete the game if they feel they don't want to play him/her.
And I cannot concede that the point about rating inflation is valid. For example, say your 1800 leaves the site and resigns all his games. Under the current system, the first people he resigns to get more points than the last. The guy can basically chose who gets more points for beating him. The poor saps who are among the last-resigned games may get nothing for their victory at all!
This even allows a spiteful player to deliberately tank his rating so that certain opponents don't get any points for beating him. It has happened here before.
With the floor system, the opponents in the above scenario at least get credit for beating at least a 1600 [or 1700 if the floor is 100 points below class]. Since they all have a victory against a player who is actually 1800 strength, this is the more just result.
18 May 08
Yeah, I was thinking about this earlier, and have now changed my mind. The rating system is not a zero-sum game anyway, with people joining and leaving all the time, and I think the number of players this affects should minimal (especially when there is no point as you won't be able to sandbag).
So, it would be much simpler simply to put a floor on your rating.
At what level should it be set? I'd think 100 or 150 pts below your maximum (non-provisional) rating would be about right.
18 May 08
1 edit
Originally posted by Humph3I would say round down the rating to the nearest hundred, then subtract 100 more to get the floor. For example, an 1857 would have a floor of 1700.
Yeah, I was thinking about this earlier, and have now changed my mind. The rating system is not a zero-sum game anyway, with people joining and leaving all the time, and I think the number of players this affects should minimal (especially when there is no point as you won't be able to sandbag).
So, it would be much simpler simply to put a floor on your ra ...[text shortened]... et? I'd think 100 or 150 pts below your maximum (non-provisional) rating would be about right.
Also, it may be prudent to wait until someone plays X games over 1800 to lock in the floor. The idea is to make sure they establish themselves in that rating class before assigning the floor.
- Joined
- 07 Jun 05
- Moves
- 5301
18 May 08
Originally posted by SwissGambitAnd I cannot concede that the point about rating inflation is valid.
Fiddling with the TER only fixes the problem for banded tournaments. The problem is much bigger in scope than that. A player who is underrated by 800 points causes trouble in ALL rated games, including clan challenges, open invites, and open tournaments. The worst case is open tournaments - there is no way to avoid playing the guy without hurting your to ...[text shortened]... ll have a victory against a player who is actually 1800 strength, this is the more just result.
Why not? I gave the extreme example. Are you trying to say that it wrong?
A player who is underrated by 800 points causes trouble in ...
There are issues with the current system, including when you first join. But if someone, for lifestyle or whatever reason, regularly resigns bunches of games, then they are likely to do it again, and you are as likely to get free rating points as to lose them. Someone who is trying to be considerate could resign the games first where he is in a losing position, and last those where his opponent is playing on to the checkmate despite being in a losing position - the opponent gets a win, but no rating points - and why should they? - if he had played on they would have lost.
Rating is only a way of keeping the score. In an open tournament, you get a stronger opponent than the number says. But in an open tournament, you play a variety of strengths anyway, so there should not be a problem.
... get credit for beating at least a 1600...
The extreme case again to illustrate: So in the open tournament, the 1000 rated player who was just hanging on to the bitter end gets a rating boost for beating a much stronger player twice? To me that is not any better, and certainly no more just. If anything it would encourage playing on in totally lost positions.
Without care, a rating floor could also impact when people change their playing style. I have seen one player drop 200 points just by putting less effort into his games. How can a rating floor cope with that? They would just "bounce" along the floor all the time.
I am aware that people arrive and leave all the time. But given that we all start at 1200 points, the stronger players take points as they rise, and give them back as they resign everything and leave. There is a big difference between the 300 points someone stopping playing and leaving with 900 points puts into the system, and the effect of a rating floor.
Fiddling with the TER is to my mind the lesser evil.
- Joined
- 07 Jun 05
- Moves
- 5301
18 May 08
Originally posted by Humph3Each individual game is zero sum - the total of your and your opponent's rating before and after the conclusion is the same. And if more points are pumped into the system, it effects everyone.
Yeah, I was thinking about this earlier, and have now changed my mind. The rating system is not a zero-sum game anyway, with people joining and leaving all the time, and I think the number of players this affects should minimal (especially when there is no point as you won't be able to sandbag).
So, it would be much simpler simply to put a floor on your ra ...[text shortened]... et? I'd think 100 or 150 pts below your maximum (non-provisional) rating would be about right.
Sure, you cannot sandbag. But there is also not much penalty for starting 400 games, resigning them all, and then coming back 2 months later to do the same thing. If anything, a floor should only affect TER, then at least it is visible that you have resigned a load of games.
18 May 08
1 edit
Originally posted by gezzaWhy not? I gave the extreme example. Are you trying to say that it wrong?
[b]And I cannot concede that the point about rating inflation is valid.
Why not? I gave the extreme example. Are you trying to say that it wrong?
A player who is underrated by 800 points causes trouble in ...
There are issues with the current system, including when you first join. But if someone, for lifestyle or whatever reason, regular tem, and the effect of a rating floor.
Fiddling with the TER is to my mind the lesser evil.[/b]
No, I am pointing out that mass-timeouts or mass-resignations lead to inflation no matter which system is used.
But if someone, for lifestyle or whatever reason, regularly resigns bunches of games, then they are likely to do it again, and you are as likely to get free rating points as to lose them.
I don't want a rating system that is like a points lottery. The point of ratings is to measure chess strength, and mass-timeouts are NOT an indication of chess strength at all.
Someone who is trying to be considerate could resign the games first where he is in a losing position
I don't want a rating system that relies on people to be courteous. History has shown that certain spiteful people take advantage of such a system to screw over people who are beating them.
The extreme case again to illustrate: So in the open tournament, the 1000 rated player who was just hanging on to the bitter end gets a rating boost for beating a much stronger player twice?
Yes. So what? If that player hasn't actually improved, he will not be able to hold the points for long. And the most he could get is 64 points. Compare this to someone who is 800 points underrated, and tell me, who is more harmful to the system?!
To me that is not any better, and certainly no more just.
I would much much rather a few players get 64 extra points than one player drop 800. Tell me, do you think a legitimate 800-point drop in playing strength occurs often?
If anything it would encourage playing on in totally lost positions.
Players already have the right to do that. It's nobody's business to tell someone else when to resign.
I have seen one player drop 200 points just by putting less effort into his games. How can a rating floor cope with that?
It can cope the same way that USCF does. If the drop in playing strength is legitimate, and someone is floored for lots of games, they can petition the site owner [or delegated volunteer] to lower the floor.
There is a big difference between the 300 points someone stopping playing and leaving with 900 points puts into the system, and the effect of a rating floor.
I would guess that a player leaving the site permanently causes more inflation than rating floors, but I'm no statistician.
Fiddling with the TER is to my mind the lesser evil.
Are banded tourneys all you care about? Do you have an answer for those who are playing clan challenges against 1000-rated players who are really 2000 strength?
19 May 08
1 edit
Originally posted by gezzaSure, but the point I made in my post was that as a whole system ratings are not zero sum due to people joining or becoming inactive. Compared to the number of people joining or leaving, the number who'll be at their rating floor will be minimal, so the effect of this new rule will be minimal.
Each individual game is zero sum - the total of your and your opponent's rating before and after the conclusion is the same. And if more points are pumped into the system, it effects everyone.
.
- Joined
- 07 Jun 05
- Moves
- 5301
19 May 08
Originally posted by CrowleyI have yet to see a comment either way. I would certainly prefer what you are saying - Tournament Entry Rating (TER) impact only.
Guys! Guys!
Surely these rating floors will only affect the banded tournaments this person could enter and will have no bearing on actual rating calculations...
19 May 08
Originally posted by gezzaI would prefer it as an all around rule, there is no need for having 100 timed games drop you to 800 when you're a 1900 player. Now that player is messing up the ratings of anyone they play as they climb back where they belong at 1900.
I have yet to see a comment either way. I would certainly prefer what you are saying - Tournament Entry Rating (TER) impact only.
The impact on ratings either way isn't that big a deal, so why drop a rating down to 800 when a floor would keep you closer to your true rating?
P-
- Joined
- 07 Jun 05
- Moves
- 5301
19 May 08
2 edits
Originally posted by SwissGambitWe have a difference of opinion. Cool.
Why not? I gave the extreme example. Are you trying to say that it wrong?
No, I am pointing out that mass-timeouts or mass-resignations lead to inflation no matter which system is used.
But if someone, for lifestyle or whatever reason, regularly resigns bunches of games, then they are likely to do it again, and you are as likely to get e who are playing clan challenges against 1000-rated players who are really 2000 strength?
Are banded tourneys all you care about? Do you have an answer for those who are playing clan challenges against 1000-rated players who are really 2000 strength?
I have seen other comments that clan matches are the responsibility of the clan leader. I would expect clan leaders to be irritated enough when someone resigns a bunch of games to boot that person from the clan/not accept them due to the risk of repetition. That said, I don't play clans, so this is only hearsay.
Some people seem really keen to play stronger players - it was a problem with open invites for a while, and right now I see two invites asking for opponents more than 300 points above the player's current grade. If ability is all that matters, these people should be equally keen to play someone currently rated 1000 but strength 2000.
Banded tourney are my main concern for this issue - AFAIK there is nowhere else where someone who is underrated really has a significant impact - winning a tourney he should not be in.
If the drop in playing strength is legitimate, and someone is floored for lots of games, they can petition the site owner [or delegated volunteer] to lower the floor.
I'm not sure I would give the site admins any more work. This also means adding a manual tweak to what could otherwise be an automatic system.
I would guess that a player leaving the site permanently causes [b]more inflation than rating floors, but I'm no statistician.[/b]
Guess again. My exaggerated example could put 100 times 64 points into the system with a rating floor. The maximum a player leaving can impact is 1200 less whatever his rating ends up at.
Players already have the right to do that. It's nobody's business to tell someone else when to resign.
Absolutely not. But if I consider I am losing against a strong opponent I consider it good manners to resign - why waste my and his/her time playing on when s/he has already shown me that I have made a mistake. I have to keep thinking and playing to my best ability to win the "won" game. Yes, I can do this, and do so without complaint, not even a whining in-game message - I could be wrong after all. But occasionally someone drops a piece for no compensation - a clear blunder - and yes, I've done it too. To play on in such a situation, hanging on for six months for the inevitable checkmate, or the hope of a timeout win, to me is bad manners, and should not be encouraged.
I would much much rather a few players get 64 extra points than one player drop 800. Tell me, do you think a legitimate 800-point drop in playing strength occurs often?
I would rather have a few players drop 800 rating points, rather than devalue everyone's rating. I doubt that playing strength ever drops so much, but rating can, and should if the player plays in such a way as to merit it.
Yes. So what? If that player hasn't actually improved, he will not be able to hold the points for long.
But now he has points he does not deserve. That can push his maximum up, and only allow entry to tournament bands which are too strong for him.
No, I am pointing out that mass-timeouts or mass-resignations lead to inflation no matter which system is used.
Really? With the current system, if someone resigns a bunch of games, comes back and then does the same again, he gives away points by resigning, takes them back by winning, gives them away be losing - repeat as often as you like with no inflation. Try the same thing with rating floors and you get inflation.
I don't want a rating system that is like a points lottery. The point of ratings is to measure chess strength, and mass-timeouts are NOT an indication of chess strength at all.
I don't want the rating system devalued. Ratings are actually a measure of likelihood of winning the game. As such, they can cope with mass timeouts - you lost a bunch of games, so your rating tanks. Historical values of rating give you an indication of strength if you want it, as would a TER limit.
As for spiteful people, I'll ignore them where possible, and encourage people to be considerate, rather than break the system.
When responding to Humph3 I said the following:
Sure, you cannot sandbag. But there is also not much penalty for starting 400 games, resigning them all, and then coming back 2 months later to do the same thing. If anything, a floor should only affect TER, then at least it is visible that you have resigned a load of games.
Care to answer that one? At least with the present system, or a TER impact only, you have to regain all those points you dropped again.
Cheers,
Gezza
ps. Edits, as bold, and inserts got messed up.
- Joined
- 07 Jun 05
- Moves
- 5301
19 May 08
Originally posted by PhlabibitAt least to make the penalty for running away and coming back again a lot of hard work to get your rating back up, if that is what you care about.
... so why drop a rating down to 800 when a floor would keep you closer to your true rating?
P-
- Joined
- 07 Jun 05
- Moves
- 5301
19 May 08
Originally posted by Humph3This is where I am not so certain. There are a number of people who go away and come back. I also know that one of my current opponents has approx 400 games in progress. I have nothing against this. However, if he were, through no fault of his own, not to have internet for a month without the notice to put up a vacation flag, he would most likely lose a bunch of these as timeouts. He would probably want to come back, and if he can cope, may well want to have the same number of games again.
Sure, but the point I made in my post was that as a whole system ratings are not zero sum due to people joining or becoming inactive. Compared to the number of people joining or leaving, the number who'll be at their rating floor will be minimal, so the effect of this new rule will be minimal.