I stand by what I said, Paul. Winning a game from someone rated 150 pts less than you gets you more than two or three points. I think it gets you at least eight, probably more. Lots and lots of people have inflated ratings due to the low average-rated player they 'harvest.' It's all there in the player stats. A few months ago, someone named Sapan had a rating of 1931. He's now in the 1700s. A couple bad days? No. His rating got inflated by bottom fishing. Not necessarily intentionally, but then he could play up - in this not so great system - which he did, and wham! Hammered. Now it's possible that some of the tourney entrants were happy about that because they could then pad their own scores. That's not me, but I guess that's fine, too. To each his own.
Originally posted by joesheppeI don't think you understand the concept of the ratings formula and how it works, but I understand I won't change your thinking.
I stand by what I said, Paul. Winning a game from someone rated 150 pts less than you gets you more than two or three points. I think it gets you at least eight, probably more. Lots and lots of people have inflated ratings due to the low average-rated player they 'harvest.' It's all there in the player stats. A few months ago, someone named Sapan had a rat ...[text shortened]... then pad their own scores. That's not me, but I guess that's fine, too. To each his own.
I know for a fact that I personally have been a "conduit"- winning points from higher rated players and passing them to lower-rated ones when I get scalped.
I did not cite the 150 points reference, but you are half right- a win would actually yield the higher rated player approximately 10 points, more than you suggested.
However, if the higher rated player loses, he loses about 22 points. He has to win better than a 2-to-1 clip just to stay even in his rating. And 150 points is not that much of a gap.
If both players are accurately rated, and they play a large number of games, both of their ratings should stay about the same- the whole point is that the formula assigns an uneven number of points based on the ratings difference.
Even more to the point, if a higher rated player has an inflated rating, and he plays lower rated players, he will LOSE points, as the formula corrects the situation.
This is the point of the Elo system, and how it works.
It's not very often that the player rated 150 points higher loses. Too lopsided, and yet he will reap a good reward. Multiply that time ten, and voila! I have personally been crawling my way to where I am - with a very low won games to games played ratio - because I always challenged up. Going from this point in time, I'd bet that I could see a 2000 rating in just a few months by playing down. Not likely by playing up or even.
Originally posted by joesheppeThe ratings formula says otherwise, and I trust the mathematical minds behind it. The points gained or lost are proportional to the win/loss percentage. If higher and lower rated players play a series of games, and they win and lose at the percentages they are supposed to, their ratings stay the same.
It's not very often that the player rated 150 points higher loses. Too lopsided, and yet he will reap a good reward. Multiply that time ten, and voila! I have personally been crawling my way to where I am - with a very low won games to games played ratio - because I always challenged up. Going from this point in time, I'd bet that I could see a 2000 rating in just a few months by playing down. Not likely by playing up or even.
It's clear you think otherwise, but that means your issue is with the theorists who made the formula.
It also strikes me as funny that you think, as the lower rated player, your rating will go up, meaning that all the higher rated players you are playing will go down, which is the opposite of what you say should happen.
By your own logic, your rating should go down, and the higher rated players playing you are padding their ratings. Shame on them for playing you!
Of course, you could argue that you are underrated, but all the other lower rated players aren't, and I would not have an answer to that.😕
Paul, you wrote: "It also strikes me as funny that you think, as the lower rated player, your rating will go up, meaning that all the higher rated players you are playing will go down, which is the opposite of what you say should happen."
I don't think that, and I'm not sure where you got that. My main point is that I could gather rating points much easier - and probably faster - by beating up three 1500+ rated players v. winning one out of two games with a player rated 1900. Does that make me a better player now that I've gained the quick points? No.
Originally posted by joesheppeI evidently misread for the first part- sorry about that.
Paul, you wrote: "It also strikes me as funny that you think, as the lower rated player, your rating will go up, meaning that all the higher rated players you are playing will go down, which is the opposite of what you say should happen."
I don't think that, and I'm not sure where you got that. My main point is that I could gather rating points much eas ...[text shortened]... er rated 1900. Does that make me a better player now that I've gained the quick points? No.
For the second part, for you to win a given number of rating points, whether you play up or down does not matter, if all the players are rated accurately. The formula acounts for it.
If you are currently 1900, and you play a 1500 player, if you win 31 times and they win only once, it's roughly a wash- you won't gain anything. That's why most players don't like to play down.
I don't think you will raise your rating the way you think you would, but it's all speculation without the games being played.
I think this is simply a misunderstanding of the formula, and until that is resolved, I suppose we will simply see things differently.
It may well be that you have observed players with inaccurate ratings, which would give skewed results. For my own observations, I have noticed that, other things equal, a 1400-rated nonsub is usually much stronger than a 1400 subscriber, which I have attributed to the fact that the nonsubs have usually played far fewer games, so their rating isn't grounded as well- and often they are simply on the way up.
Paul
1 edit
Originally posted by joesheppeThis rating table is not guaranteed accurate above 2000, but it will do for this purpose. It was published in Thread 80000 some time ago.
Your last point is a good one. When you said earlier that one win by a 1500 player over a 1900 player would erase the results of 31 wins, was that grounded in fact??
If so, that surprises me and gives me pause!
Rating Chart
Difference ____ H -- D -- L
0-10 __________ 16 - 0 - 16
11-32 _________ 15 - 1 - 17
33-54 _________ 14 - 2 - 18
55-77 _________ 13 - 3 - 19
78-100 ________ 12 - 4 - 20
101-124 _______ 11 - 5 - 21
125-149 _______ 10 - 6 - 22
150-176 _______ 9 - 7 - 23
177-205 _______ 8 - 8 - 24
206-237 _______ 7 - 9 - 25
238-273 _______ 6 - 10 - 26
274-314 _______ 5 - 11 - 27
315-364 _______ 4 - 12 - 28
365-428 _______ 3 - 13 - 29
429-523 _______ 2 - 14 - 30
524-719 _______ 1 - 15 - 31
720+ __________ 0 - 16 - 32
With a 400-point rating difference, lower-rated player earns 29 points for a win, loses 3 points for a loss. (31 may have been an exaggeration, but it seems to be about right for a 550-point difference.)
In practice it doesn't work quite that way. The higher player may start 400 points above, but after 10 wins he'd be 430 points above and in the next bracket; after 20 more wins he'd be 470 points above. One loss would cost him 30 points and bring him back to 440; a second loss would make him 410.
Not going to happen like this of course, but as someone already said the rating tends to regress to the mean.
Originally posted by KewpieI was using the rough "16 points even, +/- 4 points per 100 rating points difference" rule of thumb, but Kewpies numbers are exact, and her qualifications are also right on. She's more accurate than me, in other words!
This rating table is not guaranteed accurate above 2000, but it will do for this purpose. It was published in Thread 80000 some time ago.
Rating Chart
Difference ____ H -- D -- L
0-10 __________ 16 - 0 - 16
11-32 _________ 15 - 1 - 17
33-54 _________ 14 - 2 - 18
55-77 _________ 13 - 3 - 19
78-100 ________ 12 - 4 - 20
101-124 _______ 1 ...[text shortened]... pen like this of course, but as someone already said the rating tends to regress to the mean.
It is for this reason that many higher rated players woh are concerned about ratings avoid playing down, because the get very little for winning, but one underrated player whacks them mightily.
Objectively, playing up or down should not affect one's rating if all parties are rated accurately, but subjectively many prefer to "play up" because the risk/reward ratio is so much more favorable.
Kewpie, thanks for clarifying in one post what I failed to explain very well in several!
2 edits
Originally posted by OdinsonI am fortunate that my rating is high enough to largely avoid the problem of which Odinson speaks. On the rare occasions when I play up 400+ points, I accept it as an opportunity for a lesson.
In the past there have been numerous ratings-based public tournaments instantiated by RHP. It seems that there are far fewer now.
The issue is that certain VERY-highly-ranked individuals are entering many generic public tournaments - there is nothing wrong with this and it's in violation of no rules, morals, or ethics. Unfortunately, however, when the la ents were created with a rating "ceiling"...
Interested in others' thoughts.
Dave
Nonetheless I can empathize with Odinson's frustration. It seems like banded tournaments were much more prevalent in the past, and they seem to have been reduced in number. Maybe I'm wrong...as someone else pointed out in this thread, the 1800+ bands are so slow to fill as to be silly.
In any event, I've had the pleasure of playing Odinson many times and have found him to be friendly and gracious in spite of our rating differential.
-Will
Originally posted by joesheppeBy beat do you mean win by checkmate or resignation or are timeouts to be included?
So something needs to be done. One seemingly easy solution would be to require that the player beat ten 1800+ rated players before he qualifies for the 1800+ tournament, regardless of current rating. I don't know; that's just one idea. Something has to be done. I don't think the July 1800+ quartets is an accident. Do you see any players there rated higher than 1900? How many above even 1800??
If timeouts are to be excluded then is that fair on the player who has outplayed say a 2,000 rated opponent only for the opponent to stop playing?
Is it fair when a "genuine" 1800+ player decides to resign all their games in a tournment after 1 or 2 moves, knowing it will not impact on their rating, because they don't have the time and their opponents then get 'two wins' towards their 10 wins needed to qualify?
Are the players to be "genuine" 1800+ players or 1800+ players who have reached that rating as a result of playing opponents with an average rating of say 1500/1600?
Are wins against "banned" players to be included or excluded from the calculation?
Are the wins to only count if the players beaten was rated at 1800+ at the time of the win?
Are the wins to be counted regardless of when earned or only if obtained within say 90 days of the tournament starting?
I think what you are looking for creates more problems than it solves.
Originally posted by joesheppeYou will find very few people like yourself where your average rating, never mind your highest rating is lower that your opponents average rating.
It's not very often that the player rated 150 points higher loses. Too lopsided, and yet he will reap a good reward. Multiply that time ten, and voila! I have personally been crawling my way to where I am - with a very low won games to games played ratio - because I always challenged up. Going from this point in time, I'd bet that I could see a 2000 rating in just a few months by playing down. Not likely by playing up or even.
However, all it would take for your opponents average rating to plummet is to enter is for you to enter 1 or two "open tournament" such as say the August 2010 Grouped Random I and the August 2010 Sprint Duel 64 I. By the time you won or were knocked out your opponents average rating will have dropped to closer to 1600.
Winning game after game from opponents two-hundred or more points below the rating I enter a tourney with - and I imagine that would be at least 80% of the time, with 15% draws and 5% losses, would not be fun for me. My personal taste. I like the game a lot, and I appreciate a challenge. That's why I have so many losses - mostly from the past - and a relatively high avg-opponent rating. Unless I slip up, that metric will slowly improve, as it has, because the number of players on the site rated higher than I gets thinner as you climb up. But I'm fairly consistent, as the graph shows. I can hold my own against anyone below 2000, and above that I'm in for difficulty. To date I have a very poor record against anyone above 2100. Funny how much a hundred points can mean.