RESULTS OF SEPT 22 SURVEY
by Bill Goichberg

        For a weekend Swiss tournament with an entry fee of about $90 or $100, how would you rate the following possible sections and prizes?
       Type A, 7 sections. Open Section $1400-700-500-300, top U2400 $800-400. U2200, U2000, U1800 Section each $1000-500-300-200. U1600 $900-500-300-200. U1400, U1200 Section each $800-400-200-100. 
       Type B, 4 sections. Open Section $1600-800-500-300, top U2400 $800-400, top U2200 $800-400. Under 2000 Section $1400-700-500-300, top U1800 $800-400. Under 1600 Section $1300-700-400-300, top U1400 $700-400. Under 1200 Section $800-400-200-100.
       Please rate each type of event either poor, fair, good, or excellent.

Previous surveys have shown that players prefer a section for each class (200 point range) to double-class sections (400 point range with 200 point class prizes), and either to a one section event.  This survey spelled out the likely consequences of 400 point range sections-larger prizes for some classes and smaller for others- to see if even the classes with larger prizes would prefer a 200 point range.

Scoring excellent=3, good=2, fair=1 and poor=0, the 200 point range scored 2.0, the 400 point range 1.8. 
Going from excellent to poor, the 200 point range had 85-154-78-10, the 400 point range 70-145-78-34.

Classes with larger prizes in the 400 point range example given did vote for that range, but rather narrowly: Master 1.8 to 1.7, Class A 2.1 to 1.9, Class C 2.0 to 1.9.  Classes with larger prizes in the 200 point range voted much more heavily for that range: Expert 2.0 to 1.5, Class B 2.1 to 1.6, Class D 1.9 to 1.5.  Class E, with the same prizes for both in the sample, was 2.0 to 1.7 in favor of the 200 point range, presumably because some of these players want to play up in a U1400 Section but not a U1600.

Overall, the support for 200 point sections was impressive.  We have been using 400 point sections with class prizes for most tournaments under 200 players, as this minimizes having occasional tiny sections, but we may now start doing 200 point sections for events expected to draw in the 150-200 player range.

Consider a 5 round Swiss tournament with 3-day and 2-day schedule options.  The 3-day plays 40/2, SD/1, with games Friday 7 pm, Saturday 11 am and 6 pm, Sunday 10 am and 4:30 pm.  The 2-day plays G/75 in rounds 1 and 2, Saturday at 11 am and 2:30 pm, and then merges with the 3-day so all players play for the same prizes.  For each statement below indicate whether you disagree, are not sure, or agree.

The statements offered combined two thoughts, making the reasons why players voted as they did on many questions unclear.  3-day schedules were most popular, most players who preferred them did so because of their slower time control, and most who like the 2-day do so because it is inconvenient to play Friday night.  Only 13% of the respondents expressed dislike for having two merging schedule options, giving as their main reason that this may distort the pairings.  The statements obtaining the most agreement, with 1.5 each out of a maximum 2.0, were "I prefer the 3-day, because the time control is slow all rounds" and "I prefer the 3-day, but it's OK to offer the 2-day for some players." A future survey will be worded differently.

What should USCF policy be regarding the rating of online play? For each statement below indicate whether you disagree, are not sure, or agree.  A Tournament Director would be present in all but the two saying "no TD." These abbreviations are used:        RR= Ratable in the regular rating system        QR= Ratable in the quick rating system        OR= Ratable in a new separate online rating system

On an 0-1-2 scale, scores were as follows:

Game/30 or slower with TD present should be RR 0.7
Game/60 or faster with TD present should be QR 0.9
Game/60 or faster with TD present should be OR 1.2
Online play should not be rated by USCF at all 1.1

Game/60 or faster with no TD should be QR 0.6
Game/60 or faster with no TD should be OR 0.9

I would be more likely to enter if the event is RR 0.9
I would be less likely to enter if the event is RR 0.8
I would be more likely to enter if the event is QR 0.8
I would be more likely to enter if the event is OR 1.0

Respondents voted strongly that online play at G/30 or slower with a TD present should not be regular rated, but paradoxically, if it is regular rated, more said this would make them more likely to play than said less likely.  Online play at G/30 or slower is currently ratable under the regular rating system, but few such rated events are held.

The vote was somewhat against allowing G/60 or faster with a TD present to be rated in the quick rating system, even though not only is this currently allowed, but the quick rating of online events without a director present is also allowed, and many such events have been held without significant problems resulting. For some players who live far from the nearest club or tournament, such events (and correspondence chess) are their only opportunities for USCF rated activity.  I wonder how many respondents considered that no TD is present for correspondence chess, and even though there is no doubt some cheating, the correspondence rating system has been successful in promoting chess.  And it's a lot easier to cheat at a slow time control than at quick chess!

The strongest vote was for having a separate rating system for online play, but it narrowly beat out the second choice, that online play should not be rated by USCF at all.  The latter vote is surprising as USCF membership has probably declined quite a bit due to online play, and a logical response to this problem would seem to be to better promote some sort of USCF rated online play.  Objection to the regular rating of online games makes sense, but voting against rating online games at all seems shortsighted.

With no Director present, opinion is heavily opposed to quick rating (even though this is now successfully done) and even mildly opposed to a separate online rating.  Of course there will be more cheating in games without a director present, but requiring a director could dramatically reduce the amount of potential play and resulting benefit to USCF and American chess.  Especially, there is great potential in online USCF rated school and club team matches, but if a neutral certified TD must be present, this could stop such a development from ever getting off the ground.

A problem with starting a separate online system is that our quick system is not very accurate, a main reason being a small sample of recent games (players improve dramatically in regular rated play while playing little or no quick chess).  To start an online system as well might lead to two flawed systems, while by combining them and rating fast online play in the quick system (as we now do) the sample of recent fast games is greater and we may have one relatively accurate system rather than two that are less accurate.

Future surveys may address these issues.  Also, it would be interesting to see how those who play online vote, as their opinions are probably more important on this issue than those of players who don't do so.

Should changes be made to the USCF rating system? For each statement below state whether you disagree, are not sure, or agree. Note: due to formula change, ratings of about 1800-2399 now change much more slowly than they did prior to 2001, and Class D and below ratings change more rapidly. 
HOW FAST SHOULD RATINGS CHANGE?
My class should use a faster changing formula 1.0
My class should use a slower changing formula 0.6

If I could gain more points for a good result, I would play more 0.9
If I could lose fewer points for a bad result, I would play more 0.7

These responses suggest that a rating formula that weights recent performances more heavily might encourage more play.  Technically the rate of change is determined by the "K factor" and a higher K gives more weight to recent performances.

Currently the K factor varies according to rating.  It is highest with very low ratings and lowest with Master ratings, which makes sense as low rated players are capable of very rapid improvement that is rarely seen at the Master or Expert level.

From about 1970-2000, players under 2100 were rated at K=32, also known as "full K." This produced similar results to the old linear approximation formula, New Rating= Old Rating plus (16 x (wins-losses) + 4% of the total rating difference, with the maximum single difference set at 350).  From 2100 to 2400, K was 24 so ratings changed 75% as fast, and for over 2400 K was 16 so those ratings changed half as fast as full K.  Today there is a different K factor for each rating, and the old K=32 is close to what USCF does at the 1500-1600 level.  Lower ratings use a higher K, with 3-digit ratings much higher so those ratings change rapidly.  In the Class A and Expert area, the ratings change almost half as quickly as they did prior to 2001.  The response breakdown by class is interesting:

The first number is "my class should use a faster changing formula," the second number "my class should use a slower changing formula."
              Faster Slower
Master     0.8      0.9 
Expert      1.1      0.6
Class A    1.1      0.5
Class B    0.9      0.6
Class C    0.9      0.8
Class D    1.0       0.7
Class E     0.9      0.7

So all classes would like the ratings to change more rapidly, but this feeling appears especially strong in Expert and Class A, the classes which now are much slower changing than they were before 2001.  Here are the votes by class for agree-not sure-disagree: 

                 Faster       Slower
Master    3-10-7        3-11-6
Expert    19-18-13    7-15-28
Class A  29-25-20    2-32-40
Class B  16-36-20    4-37-31
Class C   7-35-13     4-34-17
Class D    5-20-5       1-20-9
Class E    4-13-6        1-13-9

A/Exp     48-43-33     9-47-68
B/below 32-104-44  10-104-66 

ALL      83-157-84   22-162-140

These results suggest that an increase in K factors Expert and A area, with a lesser increase in B, may help to promote tournament and club activity.  The Ratings Committee feels that this would decrease accuracy, but it should increase accuracy for juniors (who will be less underrated), and create taller peaks and deeper valleys for many players whose ratings fluctuate while their basic strength remains fairly constant.  The taller peaks and deeper valleys may be good from a promotional standpoint if players are motivated more by the hope of a big rating gain on a good result than by the assurance of a small rating loss on a bad one.  The latter issue was addressed by two questions on the survey:

If I could gain more points on a good result, I would play more 0.9 (92-97-135)
If I could lose fewer points on a bad result, I would play more 0.7 (71-97-156)

The breakdown by class on playing more:

               Gain more on good    Lose fewer on bad 
Master               0.6                           0.5
Expert               1.1                            0.9
Class A             1.0                            0.8
Class B             0.7                            0.6
Class C             0.7                            0.7
Class D             1.0                            0.9
Class E              1.0                           0.7

JUNIORS UNDERRATED?

Another interesting statement: "The rating system underrates young players too much."  Few disagreed with this as overall results were agree 126, not sure 146, disagree 52.  The overall average was 1.2 and the high vote of 1.3 was in classes Expert, A, B and C.

SIMPLE OR COMPLEX?

A simple rating system is best, if reasonably accurate 1.3
A complex rating system is OK, I don't need to understand it 1.1
The system should be simpler so I know points I may gain or lose 1.0

A bit of a split decision, with the "simple" statements finishing both first and last among the three.  The class breakdowns:

                 Simple best    Complex OK    Want to know points at stake
Master           1.2                   1.5                     0.6
Expert            1.5                   1.1                     1.2       
Class A          1.3                   1.1                     0.9
Class B          1.2                   1.1                     1.1
Class C          1.2                   1.1                     1.2
Class D          1.5                   1.0                     1.2
Class E           1.2                   0.9                     0.9
ALL               1.3                   1.1                     1.0

Apparently most players below 2200 would prefer a simpler system, but for many of these the reason is not so they will know how many points they may gain or lose in a game.  What their other reasons are might be explored in a future survey.

PRECISE OR CREDIBLE ENOUGH?

Ratings are rewards, and must be credible enough to promote chess 1.5
Ratings must be theoretically precise in order to promote chess 1.5

                 Credible    Precise
Master          0.9          1.2
Expert          1.6           1.6
Class A        1.7           1.6
Class B        1.3           1.5
Class C        1.6           1.6 
Class D        1.7           1.5
Class E         1.7           1.7
ALL             1.5           1.5

Both statements were strongly supported by players under 2200, and obtained about equal backing overall.  The lower numbers for Masters were accompanied by some comments to the effect that for them to be active they need money, not rating points.

Two philosophies are represented here.  One is that ratings are rewards, their main function is to promote activity, and absolute precision isn't necessary, just reasonable credibility.  Supporters of this view point to the cumulative Master Point system in duplicate bridge, which has been very successful at promoting play despite its inaccuracy.

The second philosophy is that ratings are a predictive tool and virtually any improvement in accuracy is justified, no matter how small, as precision is necessary to promote activity. 

Chesstour.com homepage           Tournament Schedule           Survey of August 26