Why no provisional ratings?

So I'm in a tournament and as is my luck (this sort of thing has happened before) I draw the guy with a 1585 rating, who has never lost a game -- making me think: gee I wonder what his real strength is because until he loses a game or two I have no idea...  and, of course,  by move 30 of our game his rating has shot up to +1840... and now his record is something like 20-0. 

A) I was supposed to draw a weaker opponent in the first round...  but not actually a big deal... I don't mind playing a stronger opponent...

However>>>B) I really dislike playing people whose ratings should be provisional. I want _my_ rating to reflect my actual performance... and getting blind-sided by  players whose ratings do not yet accurately reflect their performance screws up the system. 

Why doesn't chess.com handle ratings like the USCF or FIDE or any other rated chess organization and have a provisional period for the ratings? 

Apologies if this question has been asked before... and I'm sure it has... just couldn't find it.  

I agree with your point. I play on multiple sites and one prominent site uses a provisional rating. It's a difficult problem to solve, but should be addressed.

 Brian

I think this happens in very few cases, but it does occur.

Previous discussions in this area have generally agreed that Tournament Directors need to start making better use of the functionality available to them to avoid this sort of problem.  

For instance they could set a 20 game minimum limit for participants, less than 5% timeouts etc.  If these functionalities were used more often, the problem you describe would die out.

Is it really that big of a problem? After 6 months on a site, you will have played roughly a 100 games and the effect of 1 game or 3 games for that matter would be low.

After 20 games your rating can't possibly be accurate anyway. Relax, learn from the game and the rating will adjust in time. 

Tournaments can restrict the field to control this. I am always leary of opponents with a 1200 rating. It could be the best player in the world or the worst, or anywhere in between.

It's a tournament issue secondarily IMO.  Mostly it's a rating thing, and it's not important, just aggravating. 

Fluffy Rabbit: Relax, learn from the game and the rating will adjust in time.

I am relaxed, and the game is a rather difficult closed KID... I'm learning I don't like closed KIDs -- LOL. and as for the rating adjusting -- yes it will adjust (in an ideal world my rating based on 20 games with a healthy spread of wins loses and draws would be leaving the provisional realm about now), but why shouldn't the rating adjust as intelligently as possible with as few randomizers as possible? That's all I'm suggesting. You seem to be saying -- Ok the situation is mediocre, but why fix it? ... I'm saying, the situation is mediocre, why NOT fix it?   

 Fluffy Rabbit: you will have played roughly a 100 games and the effect of 1 game or 3 games for that matter would be low ...

Perhaps chess.com does ratings in a way where this is true, but the system used by USCF and FIDE does not work using a simple formula averaging all games... if there are two players, and one has a thousand games under his belt and one has 50 games under his belt, and both are rated 2100, and both lose to a player with a rating of 1800, I'm under the impression both will lose (the same?) or a very similar amount of rating -- am I wrong about this?

Perhaps chess.com does ratings in a way where this is true, but the system used by USCF and FIDE does not work using a simple formula averaging all games... if there are two players, and one has a thousand games under his belt and one has 50 games under his belt, and both are rated 2100, and both lose to a player with a rating of 1800, I'm under the impression both will lose (the same?) or a very similar amount of rating -- am I wrong about this?

=======

USCF formula handles this issue precisely.  There will be a rating difference between the two players based on the total games they played (both were defeated by the same 1800 player).

The Glicko system used here at chess.com does not use "provisional" ratings. But the number of game you've played DOES effect how your rating is determined. This is done via another number called RD or Ratings Deviation. I can't possibly get into all the specifics of this, but basically, the fewer games your opponent has played (and/or, the longer it has been since their last game), the less your rating is effected. You can read a more detailed explination here:

http://math.bu.edu/people/mg/glicko/glicko2.doc/example.html

Example -- what the hell? A lot of impressive forumlas (wish I hadn't cut _every_ statistics class now) and yet nowhere does he actually tell us what the hell the player's rating result would be -- and I'm not certain that this glicko business actually impacts the player rating at all. (He tells us what the change in the RD will be, but who cares about the RD, we want to see the impact of the RD on ratings -- ) So, in the "example" he gives, what the hell rating does our 1500 rated guinea pig end up with? He doesn't say. And from what I gather this method doesn't help us calculate ratings more precisely at all, it let's us quantify the accuracy of a rating.  In his example our new RD is 151.something or other, and our new volatility is, get this: 0.05999 -- wow! incredible!... So, uh, our 1500 player is now... uh... yeah, uh... *skritchskritch* uhmmm.... his rating is... well at any rate we do know his volatilty to the 100 thousandth of a... of a... umm... what are the units again? 0.05999 on the Glicko scale?

And here's what the document ends with:

Note that the resulting rating for this computation does not differ much from the original Glicko computation because the game outcomes do not provide any evidence of inconsistent performance.

So the example wasn't very well chosen then? -- Errrrm except, he chose it himself, with numbers he made up himself!?

Secondly -- and rather more directly to the point -- in this document Prof. Glickman says this time with admirable clarity: The Glicko-2 system works best when the number of games in a rating period is moderate to large, say an average of at least 10-15 games per player in a rating period.

Aloha! At last the professor and I have found common ground, for it sounds like he thinks a provisional period is necessary too, eh? 

this is certainly a system that reduces this problem to at least minimum

I am very new here, only 3 completed games so far for the time being; thus, my opinion will be viewed as biased and perhaps it is - for now.  But my tune will never change even after I've played 1000 games, you can accept that or not it really doesn't matter to me. 

Years ago I liked playing YahooChess, and was vexed by the attitudes there especially toward anybody whose rating was 'provisional'.... it took me over a month to get that label cleared, and not for the lack of trying!  My biggest problem was people could not only see my 'prov' label, but could still access my record as well - I seldom lost, so other provs avoided me but even worse, the rated players refused to play me simply because they were fearful for their precious rating!

Here's my take:  if you really 'think' that you are as good as your rating would suggest, what is the problem?!  If you have 2100, likely you will not play somebody <1600, and that is fine.  But why does the opponent's game total matter so much?  My understanding is that a 2100 losing to a 'rookie' 1800 will suffer less in ratedrop than that same 2100 losing to a 'veteran' 1800, and this is fair enough.  As well, the 'rookie' does not benefit the same gain enjoyed by the 'veteran' in this example.  So other than the 'label' of prov, it seems to me that the rating adjustments ARE taking provisionality into consideration.  So what's the problem?

Obviously, some people take things far too seriously.  I've only played 3 games here so far, and my last 2 were against just such people:  When it became clear to them that they were about to lose, all of a sudden they are too busy to finish the game.  I could understand that in tight situations, you may want to take far more time to try a comeback or maybe gain a stalemate; but in both of those games, the pathetic losers were in 'end-game unavoidable rundowns' and indeed had NOTHING to think about since their final few moves were forced (only one legal move to be made).  Yet both tried to tell me they needed time given their adverse positions, no need to hurry, and other nonsense.

It's just a game.  I'm not saying be careless or reckless, and I do in fact take my own game and rating with a HEALTHY degree of seriousness, but also have integrity and gamesmanship.  I am currently in a fourth game where the opponent is promptly playing when he has a 'plan' or some momentum, but each and every time that I have thwarted him, he disappears again..,., comes back with a good move or two but again I turn the tables and he flees.  I will reserve judgement on that one for the time being, as it is pretty tight; but I just KNOW that should I gain more of an upperhand, or take it to an endgame scenario, likely he will do the same.  It's ridiculous and childish and I really hope this is not rampant on this site!!

My point is PROV labels are not necessary and would only cause problems for new people trying to participate.  I am not the least impressed with my experience on this site so far, given the 3 of my first 4 games, as I would expect some class from people who claim to be intelligent.  Now, can you imagine the nonsense what would have been, if I were to have had a PROVISIONAL label attached?  Good gawd!

Cyclonus --

I seldom lost, so other provs avoided me but even worse, the rated players refused to play me simply because they were fearful for their precious rating!

This isn't how a provisional rating works, or is supposed to work (I don't know how Yahoo does things).

A provisional rating means that because there isn't enough data behind the number to make it meaningful... so your opponenst don't have to "wager" rating points against you.  A provisional rating means that game is "rated" for you but unrated, or at a greatly reduced points hazard, for your opponent.

It's very simple -- when I play a 2100 rated player, darnit I'd like credit for it... I'd like a nice rating bump if I can win! And if I lose, don't rape my rating, please...a slight nudge downward please the guy is 250 points higher than me I was never expected to win.  

Right now, when Mr. 2100 or Ms.2300 for that matter, joins chess.com -- If I'm the unlucky chump who has to play against her first or second game -- if I lose to this "1200" level player my rating gets creamed. Glicko be damned the rating adjustment is harsh! That isn't fair.  And then I want to join a tournament that I belong in.... like an 1800+ tournament... I'm ineligible. Too low rated now, sorry. So these ratings do have some consequences.  Why should losing to a freaking titled player ever cost someone the rating penalty of losing to a beginner?  I don't have the patience or the time (nor do I enjoy) to beat up a steady diet of players 200-300 points worse than me (which is the best way to pump up a rating)... I prefer to play against opponents who challenge me.  That costs me rating because good opponents means I lose and draw a fair amount. I also play very few games at once, againt to keep up the quality. So, I don't have 150 points to give away to every expert who joins Chess.com -- I love the challenge of facing excellent players --but I'll take the risk and the reward, not just the risk please. 

JG27 - thank you for your reply.  I think you slightly misunderstood or I wasn't clear enough, given your 3rd paragraph after the quote.  lol and it's a funny read - the tone I mean - you have a great writing style!  But I am aware of how 'very simple' it really is, fear not ;) 

I understand your point, in fact I made it myself;  2100 wouldn't want to play <1600 which is fine, because the risk isn't worth the reward.  I don't think the Yahoo system is any different, calculation-wise, just that the label was on the player; but that didn't mean if a newbie beat the grandmaster that the newb wouldn't be rewarded for such.  Indeed, the newb's reward was comparable, but less than an equally-rated vet's; but the grandmaster's rating would NOT be 'raped' when losing to the newb, yet rightfully  WOULD be when losing to the vet.  The newb would not suffer any more or less than an equal-rate vet in the case of losing to the grandmaster however; and the hit was, like you said, a few points down.  The underdog discount, if you will.

It was mentioned that this site does the same in terms of factoring in number of games played, but if I understand your opening line, this factor is not adequate enough?  Right now, if a 2200-vet loses to you, how much will his rating suffer?  If instead that player played  JoeBlow - rating 1815, 450games played - would 2200's rating not suffer more severely?  Both of you as winners will gain substantially but won't JoeBlow's be slightly more lucrative?   

 Your wording is interesting to me, in particular when referring to playing a titlist who just joined; you sound as though you are a 'vet' of chess.com who would not like to see his rate get 'raped' by somebody he'd greatly underestimated due to their having just joined chess.com.... I see only 20 games played on your own profile; how does, as you claim, your own rating 'get creamed' losing to a newb when you in fact are also a newb?  Or is this a new ID and you are referring to frustrations of IDs past?  I can't see how you losing to 'an expert who just joined' could be any more detrimental to your rating than if you lose to a 1200 who has played 300games.  In fact, as I understand the experience/seniority factor here, you'd lose MORE playing the vet. 

If chess.com does not regard the newb-1200 any differently from the vet-1200 when it comes to  rating adjustments, then that is indeed wrong and should be fixed; but GP has been said to be factored in as well so where's the beef?  If you yourself, have recently lost to a 'newbie-expert' and it 'creamed' your rating, I would imagine that your rating would have suffered moreso had you lost THAT very game to a 1200vet.  If this is not so then I would agree that they should address it. 

As I said, I'm very new here and I was not even questioning any of the math; just made a (very long lol) point that it might not serve well the interests of the site if it's staunch gamesmen avoided newbies like the plague.

Cyclonus: If chess.com does not regard the newb-1200 any differently from the vet-1200 when it comes to  rating adjustments, then that is indeed wrong and should be fixed; but GP has been said to be factored in as well so where's the beef?

There is some difference I think -- but what it is I can't tell you.  I think that the primary adjuster affecting how fast ratings move is in how many games one has played oneself. A player with 200 games will have his rating swayed less by a new result than a player with just a few games, who would see his rating move a lot, up or down, with each new result. 

My issue is that when you play a brand new player, you have a mystery -- the player has some strength and you have no idea what it really is. Whatever the strength of that player, your rating will be affected _as if_ the player is playing at his correct rating + (perhaps) some allowance that isn't IMO adequate.

Also, it isn't just the number of games that matters... a player needs to have some losses and/or draws for his rating to be meaningful... I am currently rated 1821 and playing a game with an opponent rated 1884 (and rising). We both have completed 20 games. My record is 12-5-3, my average opponent is 1653. My won lost record against players within 100 points of my current rating is 1-2-2 and for players +200 points it's 0-3-1 -- That tells you that my rating is roughly where it belongs. My opponent's 20 game record is 20-0-0, his average opponent is 1235. His record against players within 100 points of his current rating is 0-0.  How strong is my opponent? -- after 20 games the answer is: who knows? 

Cyclonus: As I said, I'm very new here and I was not even questioning any of the math; just made a (very long lol) point that it might not serve well the interests of the site if it's staunch gamesmen avoided newbies like the plague.

Of course I agree with this and I think anyone who wants the site to succeed agrees with this.  But a good provisional rating system would encourage, not discourage, people from playing newcomers.  The USCF uses a well-established, fair, and uncomplicated process for bringing unrateds into the system. I don't understand why chess.com doesn't use it. 

I get what you're saying, but I don't know the standards used by USCF etc but from what I gather, perhaps you are figuring there ought to be some kind of threshold to protect rating-established players BETTER than the current system which I think uses the totals from each side regardless to their volumes.  If so, I'm figuring you would set the threshold to, lemme guess, twenty?  ;)

I wouldn't have a problem with such a standard I suppose; but what I fail to understand is, if one is freightened by the unknown, why venture there?  If you enter a tournament wherein you will be a higher seed, you should expect the possibility of playing a virtual unknown if the lower seeds go as far as 1200.  But what of the high seeded newbs such as yourself?  If you play an 'unknown' who is 200pts higher than yourself with fewer games played, is this not as beneficial or risk-worthy than if they weren't newbs? 

One funny thing for me here is, out of your 12 wins against 1600 averages, how many of these were 'blindsided' by the 'unknown' you?  We all have to start somewhere, and just as in any other competitive game, top seeds will be 'upset' by unknowns who eventually become known high-seeds themselves. 

The opponent as you described him/her is 'known' just as well as yourself.  While I can appreciate your analyses and criteria for such, I don't think it has to be that difficult.  And I disagree that there has to be losses or draws for the rating to be authentic.  Your opponent, though not weak, is clearly not nearly as strong as one would expect from a 20-0-0 when his opponents have averaged 400pts fewer to your own. 

You could always 'scout' his completed games to see not only the types of strategies he employs but also get a read on his opp's aptitudes.  That would paint the clearest picture.  But you cannot simply "decide" that a player is strong or weak by reading the stats page, and it should not matter anyhow - do you play a weaker game against a 'known' inferior opponent?  Of course not.

But all said, if there is a superior system sworn by most that would know, then add me to the petition.  Meantime, play newbs the same way you play vets - the higher rated ones.