The rating system document, glicko.net/ratings/rating.system.pdf, states “An individual’s personal absolute rating floor is calculated as AF = min(100 + 4NW + 2ND + NR, 150)”. However, based on crosstables I’ve seen of events rated in the online rating systems, this formula isn’t being applied to the online rating systems. Is there a reason why this formula isn’t applied to the online rating systems or is this a mistake in the programming?
What formula is apparently being applied?
Bill Smythe
Just the 100 minimum.
Not sure, will look at the code when I get a chance, probably not for several days, we’re still kind of busy.
I may have to do some research, I’m not sure if the floating floors as approved were intended to apply to all ratings systems or just the OTB regular one.
In the formula you quoted, 100 + 4NW + 2ND + NR, does NR stand for number of losses, or number of rated games? I had always thought it was the former, but NR makes it look more like the latter.
I must confess to having had a hand in this formula. On the forums several years ago, I suggested 100 + 2NW + ND. My idea was to reduce the number of players rated exactly 100, to spread them out into the range 100-120 or so. It would look better on wall charts, I felt, not to have so many players stacked up at the bottom with identical ratings.
But the ratings committee apparently ran with my idea, although they doubled 2NW + ND to 4NW + 2ND so they could also add NR and have three different coefficients (4, 2, 1) for the three addends.
I quickly realized, though, that my idea was dumb to begin with. Any player rated 100 who wins (or draws) even a single game will already gain far more points than that, anyway, so who needs the 2NW + ND (or its doubled form) to be added on as well?
The NR part, though, might still have some point, as it would reduce the pileup at 100. Awarding rating points for losses, though, distorts the ratings and compromises the rating system. But the damage is negligible, since it is limited to players rated below 150.
In any case, the whole matter is insignificant to begin with. Perhaps the absolute floor should just be 100, period, in all the rating systems. Alternatively, perhaps any player rated below 150 who gains rating points should simply be given a new floor equal to his new rating or 150, whichever is less.
Bill Smythe
Thanks Mike.
You’re almost certainly correct with respect to wins—the absolute floor is applied at each pass in the rating algorithm (unlike other floors, which aren’t applied until the end), so someone with a win has a win over (at least) a 100 rated player—you need to combine that with several losses to players rated (well) under 400 in order for the result to be <100.
Draws, however, are different. If you have two beginners in a K-3 section who lose four games and draw each other, they will almost certainly end up with a rating calculated at 100.
As implemented, a player below 150 gets a higher floor based on 4 points for a win, 2 points for a draw and 1 point for a loss. As someone noted upthread, in most cases a win or even a draw will result in a larger ratings gain so the low floor will probably no longer be a factor.
Although Bill may have intended the low floor process to just spread out the ratings a bit at the very bottom of the curve, several within the scholastic community embraced it as a means to give players some sense of progress until their chess skills begin to increase. Our experience has been that for most players, this learned expertise starts to kick in around a player’s 3rd event. A large percentage of new young players don’t stick around for that many events, though.
It does not appear that the low-floor code has been applied to any online ratings, I will have to check with the office/ED/Board to see what they want done there. I do not expect that to be resolved before next week’s creation of the August rating list.
Is it one point for a loss? When did this change? Last time I ran a tournament with a significant number of players rated between 100 and 150, I was pretty sure that the one point was for each tournament where a player played at least three games.
Alex Relyea
I think it’s been that way all along, the proposal was tweaked several times by the Board before it was passed. The Ratings Committee’s opinion was that these ratings are low enough that it would have little or no impact on ratings.
Whether all those tweaks got reported back to the RC for inclusion in the documentation is a separate matter, it wouldn’t be the first time they weren’t informed of some policy shift.
Sometimes I wonder with really low rated players if we’re doing them any favors by publishing those ratings. Then again, I’ve also wondered if a blood glucose reading might not be at least as accurate as their pre-event rating.
This is what I have from the board action in 2008.
Aaah, I’m getting too old for this, good thing I’m being put on idle soon.
The code does count wins as 4 points, draws as 2 points, and 1 point for any event with 3 or more played games in it.
Any update here?
Neither the EB nor the RC has chosen to make a recommendation yet.
Unlike regular OTB ratings, we aren’t seeing a lot of players stuck at 100 in online ratings systems.
On our list of priorities, thus is not on the list
I do know of one case where a player was floored at 110 after starting with 46 losses, won one game in a three round event (getting a 144 rating with a 115 floor) and also lost a game to a 181 in the extra games section to drop back down to the 115 floor. Two tournaments later he beat an 832 and since then has not dropped to his floor (he came close by dropping to 134 with a 127 floor before rising back up again).
The 110 floor must have come from playing in 10 tournaments averaging 4.6 rounds each. Then his next 115 floor resulted from one more tournament (1 point) plus one win (4 points).
That he would later, after winning another game, still eventually almost fall back to his floor, I guess (almost) justifies the existence of these “participation floors” (for players rated below 150) after all. Hmm.
Bill Smythe
The new rules for online play state: “Except as specified below, player rating calculations for online play follow the rules of Chapter 8, The US Chess Rating System.” and “For players who already have an online regular, quick or blitz rating US Chess shall apply the same rating formulas to calculate online ratings that are used to generate over-the-board ratings.”
Due to these statements, I would say the minimum rating floor formula does apply to the online rating systems.
Well, I asked for guidance, but haven’t gotten any specific instructions yet. There aren’t many players who would qualify for the floating minimum floors, but the 2200 or so players who were in the Online Elementary might change that.
Does anyone (except you know whom) actually care?