@TkzonKy , I forgot to add the image link. There you go: imgur.com/a/QFlJl
Green line indicates the +/- 200 range.
Green line indicates the +/- 200 range.
@mdinnerspace , yes as I illustrated in that last image, there is a large range of variation, actually 400 points.
Do you agree with the statements below ?
- if cheating is to affect overall rating statistics, then the "size" of this negative effect depends on the *percentage* of games in that category that are affected.
- most cheaters are not software developers and do it manually, the lame way, copying the moves from a different board back and forth into Lichess board.
- the lame cheating method requires extra time. Probably not doable in 1+0 games, for instance.
@dudeski_robinson , yes Fide players play more Blitz it means that we have more Blitz data, but it doesn't mean that Blitz ratings are a better predictor. Actually, standard deviation in Blitz chart seems much wider than standard deviation in classical chart. Doesn't that indicate that the classical chart provides a better estimate ?
Do you agree with the statements below ?
- if cheating is to affect overall rating statistics, then the "size" of this negative effect depends on the *percentage* of games in that category that are affected.
- most cheaters are not software developers and do it manually, the lame way, copying the moves from a different board back and forth into Lichess board.
- the lame cheating method requires extra time. Probably not doable in 1+0 games, for instance.
@dudeski_robinson , yes Fide players play more Blitz it means that we have more Blitz data, but it doesn't mean that Blitz ratings are a better predictor. Actually, standard deviation in Blitz chart seems much wider than standard deviation in classical chart. Doesn't that indicate that the classical chart provides a better estimate ?
No, I do not agree. Bots are easily downloaded and used for bullet by those so inclined.
Besides... I'm of the opinion computer assisted play makes up for no more than 5%. Has little influence on the stats.
Your arguments regarding classical vs blitz ratings have been gone over to no end. The OP is stead fast, will never waver in his hypothesis. He did all that work, not about to reverse rudder. The formula has been shown to be faulty, making predictions well outside expected results that were established by the original premise.
Besides... I'm of the opinion computer assisted play makes up for no more than 5%. Has little influence on the stats.
Your arguments regarding classical vs blitz ratings have been gone over to no end. The OP is stead fast, will never waver in his hypothesis. He did all that work, not about to reverse rudder. The formula has been shown to be faulty, making predictions well outside expected results that were established by the original premise.
@EvilChess I'm not sure. The correlations between (a) blitz and fide, and (b) classical and fide are basically identical.
The wider spread may be an optical illusion, due to the fact that I used a low opacity for individual dots. Since there are more observations in blitz, the extremes look "darker" (the mid-range doesn't change; once it's black, it's black).
But yeah, not sure.
The wider spread may be an optical illusion, due to the fact that I used a low opacity for individual dots. Since there are more observations in blitz, the extremes look "darker" (the mid-range doesn't change; once it's black, it's black).
But yeah, not sure.
@mdinnerspace , so you think lame cheaters are "easily downloading bots from internet" ? I would not think so.
From my experience on playing longer games against 2100+ opponents, actual percentage of cheaters in classical category is likely way above the 5% number you suggest. Within the games I have played this month, 2 of 29 were marked as cheaters and I'm quite sure a couple more were also cheating. Players rated below 2000 will hardly notice any cheating, but I notice it at about 10% of my classical games (25 min +). I don't know if that makes it relevant. Probably not, but makes it very boring :-).
By the way, you seem a bit angry and I don't know why.
From my experience on playing longer games against 2100+ opponents, actual percentage of cheaters in classical category is likely way above the 5% number you suggest. Within the games I have played this month, 2 of 29 were marked as cheaters and I'm quite sure a couple more were also cheating. Players rated below 2000 will hardly notice any cheating, but I notice it at about 10% of my classical games (25 min +). I don't know if that makes it relevant. Probably not, but makes it very boring :-).
By the way, you seem a bit angry and I don't know why.
@dudeski_robinson Really interesting stuff. That scatterplot is stunning. I never would have guessed the correlation between online blitz and FIDE (self-reported even!) would be so strong.
I would have thought that the difference in rating between the blitz time control and 'classical' would also be a useful signal. Someone who gains 200 rating points moving from bullet to blitz and again from blitz to rapid would seem to have the capacity to do even better in a 90 minute OTB game. Am I correct in understanding that your data indicates this is not the case?
I would have thought that the difference in rating between the blitz time control and 'classical' would also be a useful signal. Someone who gains 200 rating points moving from bullet to blitz and again from blitz to rapid would seem to have the capacity to do even better in a 90 minute OTB game. Am I correct in understanding that your data indicates this is not the case?
If I hit the formula exactly than this single event will prove the formula right as every stickler can see obviously. What are averages; the single example counts!
I will never play again and the formula stands for eternity. :D
I will never play again and the formula stands for eternity. :D
@agarden Thanks!
You might be right, but the effect of those gaps would have to be modelled explicitly (I haven't done that). My general feeling is that that might be asking too much from what is still a pretty limited dataset.
You might be right, but the effect of those gaps would have to be modelled explicitly (I haven't done that). My general feeling is that that might be asking too much from what is still a pretty limited dataset.
@EvilChess : Please do us all a favor and ignore mdinnerspace like everybody else does. He sounds somewhat coherent at first glance, i give you that (like you, i fell into the same trap, see some 10 pages ago), but in fact he is just "against everything". (I suppose if there ever would be a thread about being against everything he would be against that too.) We (the collective "we") spent several pages of discussion just because he thinks standard deviation is a hoax - as well as other statistical procedures and/or terms.
Regarding the correlation about blitz rating and OTB rating: it is perhaps less strong as supposed because the win expectancy is already 100% for ~350 points difference. That means: if two players play a ELO-rated game and the difference between their ratings is 350 points (or more) then the higher rated player has an expectancy of 100%. Of course will a GM (say, ~2500 in OTB) blitz better than a 2000-rated player, regardless of the 2000-player being good or bad in blitz and the GM being particularly good or bad in blitz. But it might well be that from two players with both 2500 in OTB one will win 8 out of 10 blitz games consistently, jsut because the one is (relatively) good in blitz and the other is not.
krasnaya
Regarding the correlation about blitz rating and OTB rating: it is perhaps less strong as supposed because the win expectancy is already 100% for ~350 points difference. That means: if two players play a ELO-rated game and the difference between their ratings is 350 points (or more) then the higher rated player has an expectancy of 100%. Of course will a GM (say, ~2500 in OTB) blitz better than a 2000-rated player, regardless of the 2000-player being good or bad in blitz and the GM being particularly good or bad in blitz. But it might well be that from two players with both 2500 in OTB one will win 8 out of 10 blitz games consistently, jsut because the one is (relatively) good in blitz and the other is not.
krasnaya
The inconsistencies in the formula that I presented can not be refuted. The formula makes a prediction based on data presented of a players online blitz and classical ratings. It predicts the FIDE rating to be "x". It has been conclusively proven that after substituting both ratings into the formula, the results differ greatly from the original premise. Furthermore, the higher the rating or the lower the rating, the predicted Fide rating becomes even more skewed away from the prediction.
Anybody can insert the numbers into the formula and observe the results.
Anybody can insert the numbers into the formula and observe the results.
This topic has been archived and can no longer be replied to.