I had a thought as I was preparing for my next tournament, would it be possible to display the rating performance of the last 5-10 events to see how a player has done recently (without me doing the calculations myself). We could call this new rating a “past performance rating”. It would especially help for those who are sitting on their rating floor to give a more accurate gauge of our recent playing strength.
Um, I think that’s called the “current rating”. In most cases, nothing that happened more than about 10 tournaments ago has any significant effect. It’s almost as though you had started as unrated at that point.
Bill Smythe
Based on an admittedly limited set of examples, situations we have identified since 2005 where a player had an unusual increase or decrease in rating, usually due to an incorrect ID in an event, it generally takes somewhere between 75 and 100 games for that increase or decrease to be attenuated out.
We have also done some testing looking at someone’s performance rating (the old +/- 400 points formula) over 25 or 50 games. The unreliability of that formula (the reason it is no longer used except for players who have 8 or less games, when we have little else to use) was pretty obvious, especially for players who face a large number of opponents who are much weaker or much stronger than they are.
And if you look at the ratings charts for very active players who are not at or near their floor, you will see 75-100 point swings in their ratings are fairly commonplace.
Perhaps performance rating, instead of being calculated with the 400-point formula, could be defined as “that rating which, if used as the pre-event rating, also comes out as the post-event rating”, using the regular established-player formula.
Of course, due to rounding (to the nearest whole number), there would likely be an entire range of ratings that would satisfy the above definition. That range should, in “average” cases, be rather narrow, say 20 points or less. One could then define the performance rating as the rating smack dab in the middle of that range.
But then there’s the usual glitch – what happens when the player loses all his games? In that case, the “range” would be from minus infinity to several hundred points below the player’s lowest-rated opponent. Then one would probably have to use the rating at the top end of that range, or revert to age-based (etc) ratings.
Bill Smythe
So out of curiosity more than anything I manually calculated my last 10 game average, post rating.
1680
1615
1636
1642
1637
1582
1606
1604
1652
1702
for and average of 1635 Not exactly sure of what that tells me.
One thing that the MSA already does is show the individuals peak year rating.
My original intention for this would be for players who are “sitting” on their rating floors; so that their opponents can get a more accurate read on what their rating actually is, as opposed to what it was 200 points less than it was at their peak.
Why? Should opponents play to the level of their competition, or should they play the board as best they can?
I think his point is that if someone is sitting on his floor, his ‘performance rating’ might be well below that floor. Some people could find that interesting, though possibly depressing.
However, the old performance rating formula was and remains flawed, and that’s why the USCF no longer uses it except when there is nothing better to use.
Is the old formula not similar to the FIDE one?
Generally speaking, when someone refers to ‘performance rating’, I assume they mean the formula where you average the ratings of your opponents, adding 400 points to the ones you beat and subtracting 400 points from the ones you lost to.
So, someone who beats a 400 player and loses to an 1800 player has a performance rating of (400+400+1800-400)/2 or 1100.
But someone who beats an 1800 player and loses to a 400 player has the same performance rating. (400 - 400 + 1800+400)/2.
The USCF used to use that formula to determine someone’s rating for their first 20 games or so. Now we only use a version of that formula when a player has 8 or fewer games before the current event.
I don’t think FIDE uses that formula.
I think I got the point, and he said "
I can understand that there are arguments against floors - I’ve personally supported removing floors in the past, and while I haven’t given it a lot of recent thought, probably would continue to support that now.
But ratings are used for pairings and prizes. The argument that he actually presents above “so that their opponents can get a more accurate read on what their rating actually is, as opposed to what it was 200 points less than it was at their peak.” doesn’t seem to make sense. Should the opponent (that he mentions) be trying to play less well if the recent performance rating is below the floor (for example)? I don’t see the point to his comment.
I think Iragefree wants players to become depressed, by learning of their opponents’ “true” strength rather than their floored ratings.
Well, at least that could help combat the ego problem in tournament chess, I suppose.
Bill Smythe
Ego problem? What ego problem?
Most of the people I play have very healthy egos.
I’ve been on or near my floor so long that I’m depressed enough without finding out what my true playing strength is.
Floored players are funny: their UNDERSTANDING of the game is typically at least (floor + 250), so anything can happen. If they don’t blunder, they crush you.
I speak from bitter experience (glares menacingly thru intertubes at the player who has beaten me more times than anyone other than a local FM) 
Now, now. In the Illinois Games database, your score against me is 8.5-3.5 when you have white, 7-5 when you have black.
By the way, take a look at game 2399 in that database (against a different opponent). I must say, to this day I’m proud of that one – even though I missed a simpler win along the way (27…Qa6+).
Bill Smythe
Just blogged it. 
The French had a way of making strong players look not-so-strong in the 1970s. (Well, it still does, but folks are better prepared today…)
Yes, of course you should objectively be playing the best moves in any situation, of course this doesn’t always happen (for some of us it RARELY happens). I suppose that’s what makes chess exciting.
Perhaps I should have titled this post “Rating floors: The Lies we’ve been told” or something as dramatic. Here’s the thing: if I see your rating has been sitting on its floor for a long time I’m a lot less inclined to take that draw you offer when we reach a position that is equal and a lot more likely to play on. It’s a respect thing; if I think you are going to outplay me in such a position and you offer said draw I’m accepting it. Of course “Any given Sunday” is in effect in every game we play. Sometimes you play a great game; a million monkeys with typewriters and Shakespeare and all that jazz.
I have no intention of making other players “depressed” (not here anyway…otb is another issue). I just (and this is getting at a larger point in a different direction) think that these floors should be evaluated from time to time…perhaps a floor committee or something that could do their evaluation based on performance or something similar. I think the original intention of floors was a bit superfluous and attacked the elusive and mysterious problem of “sandbagging”. Show me the “sandbagger” and I’ll show you 20 players whose ratings are terribly inflated and should be (in some cases) significantly lower than their floor.
I guess I just have a different view. I try to play well. I try to understand what is good and bad about what I am doing and what my opponent is doing. Ratings are just a reflection of that.
If I don’t worry about my opponent’s rating, and just try to play well, I play like this: cavemanchess.com/Games/Bachl … s/base.htm - a game where Soltis finds a wonderful double sacrifice to hold a draw; But if I get overly concerned about my opponent’s rating, I hold back and play like this:
[Event “US Masters, Oak Brook IL”]
[Date “2000.??.??”]
[Round “2”]
[White “Bachler, NM Kevin”]
[Black “Baburin, GM Alexander”]
[Result “0-1”]
[ECO “A07”]
[PlyCount “66”]
[EventDate “2000.03.??”]
[EventRounds “7”]
- Nf3 d5 2. g3 g6 3. Bg2 Bg7 4. O-O e5 5. c4 d4 6. e3 Ne7 7. exd4 exd4 8. d3 O-O 9. a3 a5 10. b3 c5 11. Re1 Nbc6 12. Nbd2 h6 13. Ne4 b6 14. Bf4 g5
{The key position - if now either Knight takes on g5, (15 N/either x g5) the position is = or +/=. I spent a lot of time looking at this. Instead I chicken-out, simply because Baburin was Baburin, and am immediately worse.}
- Bd6 Ra7 16. Bxe7 Rxe7 17. h3 Rfe8 18. g4 f5 19. gxf5 Bxf5 20. Ng3 Bg6 21. Qd2 Ne5 22. Nxe5 Bxe5 23. Ne4 Qc8 24. Qd1 Kg7 25. Ng3 Bxg3 26. Rxe7+ Rxe7 27. fxg3 Re3 28. Qf1 Bxd3 29. Qf2 Qf5 30. Qxf5 Bxf5 31. Rf1 Bg6 32. g4 Rxb3 33. Bd5 Rxh3 0-1
This is precisely the situation I was referring to. I’m of the opinion that if I’m going to most likely lose anyway (statistically speaking) I might as well go for things like Nxg5. Why not? If i was playing some weaker player I would probably not risk it and try to play more solid (probably close to what you did in the game) and hence the rating comes into effect (although admittedly this is purely a matter of taste).
Also would it kill the web admins here to create some sort of chess viewer in the forum? I shouldn’t have to run to some third party to see a game someone posts on the USCF website.