uschess.org/msa/XtblMain.php … 1-11315844
There are two sections in this tournament. It seems like Section B was given bonus points and Section A received none. I’m not really complaining because I was in Section B. But I’m curious about the process. Can someone explain why this came out this way?
It looks like Section A did receive bonus points. If you put in the results for the 1804 into the ratings calculator it calculates his rating at 1831 (off by 1) including 6.42 bonus points. None of the other players appeared to have gained enough points for bonus points.
Remember, with variable k factors, you could end up with the sum of all the ratings stay the same and still have had bonus points awarded.
According to our records, one player earned bonus points in section 1 and 3 earned them in section 2.
So what was the trigger for bonus points? Take me for instance. I lost one game against a higher rated player and won against three lower rated players.
The details are at:
math.bu.edu/people/mg/ratings/rating.system.pdf
(though it appears that B has been changed from 10 to 6 as of 6/7/08)
So the players get an extra bonus point for each point earned over 13.4 (12 for the players who played 4 or less games). You can see an estimate of how many bonus points are earned via the ratings calculator at:
main.uschess.org/content/view/7875/400/
You earned 17.58 bonus points, so you must have earned 29.58 normally for a total of 47.16. (Of course, there are probably some round-off issues with the calculator.)
I’m not sure what you mean by a trigger. The bonus formula is:
K(S − E) − B * SQRT(m)
Where K is your K for the event (27.6)
S is your actual score (3)
E is your expected score (1.926703)
B is the bonus factor (changed to 6 as of June 7th)
m is the number of games you played (4, the square root of that is 2)
Plug those together and you get 17.623 in bonus points which, when added to the points you gained from the other portion of the ratings formula, took your rating from 1652 to 1700. (We carry all internal calculations in floating point and round away from your current rating at the end.)
BTW, the expected score I gave results in a slightly different computation of bonus points from what Tom came up with because I can access the actual expected score used in step 5 of the ratings formula, he cannot. Since we round up fractions of a point anyway, it probably didn’t make any different in your post-event rating.
I now realize that I wasn’t higher rated in all three of my wins. I’m sure that would affect my expected score.
It seems to mean like the variance between the two sections is that in Section A everyone played 5 rounds and so ended up closer to their expected score.
While in Section B due to unplayed games three players only played 4 rounds.
I submitted this as a Swiss event but in reality it was a RR. Should I have mentioned that somewhere in the rating report? Would it being a round robin affect the bonus points?
Was the expected score inaccurate due to the unplayed games?
We only compute the expected score for the games that were actually played.
well then I will just be happy with my Bonus points. At least for as long as I can hold onto them.
On average somewhere around 21% of players were earning bonuses when the
bonus factor was 10. It’s a bit early to tell what the long term trend will be but in the events
that have ended since the bonus factor was changed to 6 around 25% of players have earned bonuses.
And will be changing again (going lower) RealSoonNow.
Whaaaaaaa? Now it will be even easier to get bonus points? We already see enough cases of kids who gain 200+ points in some good tournaments (by beating a couple of higher rated players) and then spend the next six months giving most of those points back because they can’t compete at their much higher rating. Try explaining to the kid that “you did so well at that one tournament that you’re now overrated.” I’m surprised that not more of these kids (or their parents) quit chess because they are “getting worse over time.”
Sigh!
Michael Aigner
Yes.
How many cases is that, precisely?
How many cases are there now, precisely?
Is this really the sort of statistical training and analysis that you bring to the table? “Whaaaaaaa?” and “Sigh!”? Those are technical terms with which I am not familiar.
Please bring your data to the Ratings Workshop in Dallas, so that we can discuss it.
And, please be prepared to explain how to resolve your concern in this post with your concern expressed elsewhere:
I am not sure what kind of statistical evidence that Professor Sloan wishes to see. I can only relate the concerns of young players and parents who are so excited to gain a large number of points in a short period of time, only to become discouraged over the next several months as they lose them back.
Do I have to present evidence of players who jump 100+ points and then slip all the way back to their original rating? Speaking just from my own limited set of students, I have seen all too often where they get lucky in an action (G/45) time control and gain a ton of points. Last weekend, one of my 1700 rated kids won three up pairings where he was down at least a piece in each game, only to be stopped by an IM in the last round. Did he deserve to see his rating jump by 90 points? I can provide USCF ID numbers in a message if you wish.
Statistical evidence is nice. I have college degrees in engineering and math, so I like numbers. The problem is that numbers alone don’t show the human side of the story. The parents that I see here in California tend to be highly educated in the sciences, but if their children don’t achieve “progress” then they look for other activities. Unfortunately, that progress in chess is usually measured by rating.
Finally, to address the apparent discrepancy between my complaints with the rating system: My old complaint about the difficulty in achieving bonus applies to masters, players with a very low K factor and thus rarely if ever get bonus points (it now got a bit easier in June). My recent complaint applies to C and D players (and below) who already have a high K factor and can get bonus points simply by beating one player rated above them, whether by skill or by luck.
I thought the fixed K=32 solution (K=24 for 2100-2400, K=16 for 2400+) worked reasonably well. In fact, FIDE and ICC both still use slight variations on this method, apparently with success.
Michael Aigner
I thought you said 200+ points? Which is it?
Is this 90 point gain an example of a 100+ gain, or a 200+ gain?
You are welcome to your opinion. The data disagrees with you.
Are you going to answer my questions about HOW MANY players show the kind of 200+ gains followed by falling back to the previous level in 6 months that you described in your first post? Or, was that just hyperbole? Or, was your example of a single player who gained 90 points (but has not yet lost them) supposed to be your answer to that question?
You use phrases like “all too often” - but when asked HOW OFTEN, you change the subject. How are we supposed to judge whether it is too often, or not often enough, or just right?
Keeping in mind that the ratings system is statistically based and there are always going to be outliers on the curve, I looked at all the players with established ratings under 1500 who had a 100 or more point gain during an event in FY 2006-07. There were 6116 such players.
Of those players, 43.9% have a current rating today that is higher than their peak post-event rating during the 2006-07 fiscal year. 21.3% have a current rating within 25 points of their peak rating during that year, 9.0% have a current rating 26-50 points below that peak rating, 7.5% have a current rating 51-75 points below that peak rating, 5.6% have a current 76-100 points below that peak rating, 7.3% have a current rating 101-150 points points below that peak rating, 3.3% have a current rating 150-200 points of that peak rating, 1.4% have a current rating 201-250 points below that peak rating and 0.8% have a current rating more than 250 points below that peak rating.
The question was: how many players gain 200+ points in a single event and then lose those 200 points within 6 months.
Of the 1505 players with established ratings who had a 200 or more point gain in the 18 months from 6/1/2006 through 12/31/2007, it looks like 52 of them subsequently gave back at least 200 points within six months. That’s 3.45%.
Excellent! Now, the question is: is this “too many”? or, “unexpected”? Is this a problem that needs to be addressed? How many more such cases will there be when we lower the bonus point threshold?
Following up on Mr. Aigner’s post, I suppose a reasonable next question might be: how many of these 52 players quit chess in disgust because their parents thought they were “getting worse over time”?
And after that…if this is a problem, how do we fix it? (without breaking something else…)
Those with answers (and other questions) are invited to the Ratings Workshop in Dallas.