Request for new time regime.

[henkvj]

Hi Richard,

Congrats with this great site. Imho its the best thing that happened to chess since the invention of the back rank mate!

I immediately switched from emrald when I noticed I could do rated non-blitz tactics here. One suggestion: For me blitz is to fast, but do think some time constraint helps to improve faster. Could it be possible to introduce a second time regime, where you have about 3 minutes to solve?

keep on the good work!
Henk

[richard]

Thanks Henk,

I'm glad you're enjoying the site. I've considered adding one more time control but there are a few (solvable but annoying) technical issues involved as well as the matter of choosing the new time control.  Some users want a slightly slower blitz, others want an even faster one :-)  I'm probably going to continue with the current time controls for a while and will revisit this a bit further down the track.  One of the reasons is that spreading users out over more time controls means that the time for new problems to gain reasonable ratings goes up a bit too much.  As more users start to use the site this becomes less of a problem so new problems can be absorbed more quickly, at that point I'll consider another time control (or two :-)  ).

Regards,
Richard.

[drahacikfm]

A suggestion to address this issue without adding another time control:

In standard-rated, award only half as many points for solving a problem correctly, if you spend more than 5 minutes to solve it.

Reasons:

1)  In a real tournament game, you can't spend forever looking for a move.  You have a clock and if you use too much time, you get into trouble later in the game.

2) I find I get a little lazy in standard-rated because there is no time control.  I'll watch TV for a couple minutes, or talk to someone, in the middle of doing a problem.  But if I knew that after 5 minutes I wouldn't get as many rating points, I would focus better and probably improve the way I think during tournament games too.

[newlook]

Glicko ratings can also account for a draw. That might be the best way to score a solution that is correct but takes over 5 minutes in standard mode.

[boybawang]

A suggestion to address this issue without adding another time control:

In standard-rated, award only half as many points for solving a problem correctly, if you spend more than 5 minutes to solve it.

It's a very good suggestion but sadly, others don't want time constrains because they come here just for the sake of solving with leisure.

But I have a perfect solution to that in Standard-rated

Just Double the Rating gained to those who can solve the problem quicker (in less than a minute). Beyond one minute, normal rating will apply.

advantages:
1. Solvers can still disregard time if they wish because normal rating system will apply.

2. Solvers will have the option to solve quicker and they will be rewarded.

conclusion:
The above suggestion will result to more accurate rating in standard mode. Currently, the rating is unfair because it all depends on how much time the solver spent.

[tmr]

Some of these are good ideas.  In standard mode ratings are really only comparable when the users spend roughly equivalent amounts of time or at least don't get a bunch of problems wrong by not spending enough time.

Another idea to account for this would be gradually reducing the rating increase for getting the problem right when taking more time (never negative or zero though) and gradually increasing the rating decrease for getting the problem wrong with more time.  The time scale could be linked to the problem difficulty.  This is probably harder to impliment than say a flat 5 minute limit or cutoff but is probably better overall considering different problems and skill levels.

I know I take pretty much as much time as I need to solve a problem, something I bet many don't do.  Allowing this makes the standard mode ratings less comparable to those that say never spend more than a minute on a problem.  For example I've seen mcuh lower rated users solving problems in standard mode as if they were in blitz mode and getting many problems wrong.  Their rating thus suffers.  Because of this, ratings under the blitz mode are probably a more true indicator of performance between players than in standard mode.

[newlook]

Richard has the numbers. Maybe he has calculated correlations between people's OTB ratings and ratings here, Blitz and Standard. My bet is on the Standard ratings.

Edit: I cut and pasted records for people who are active (defined as RD 120 or less) in both Standard and Blitz into a spreadsheet. The correlation between the two is .8 (which is very high for behavioral data). Contrary to popular belief and my own expectations, Blitz ratings are about 190 points higher than Standard ratings.

[boybawang]

Allowing this makes the standard mode ratings less comparable to those that say never spend more than a minute on a problem.  For example I've seen mcuh lower rated users solving problems in standard mode as if they were in blitz mode and getting many problems wrong.  Their rating thus suffers.  Because of this, ratings under the blitz mode are probably a more true indicator of performance between players than in standard mode.

I agree tmr.  The standard mode is just good for testing your accuracy But is an unfair rating comparison.  Without time, Everybody is allowed to make the correct solution which corrupts the rating of each problem.

IMO giving rating bonus to fast solvers should remedy this...

[drahacikfm]

Without time, Everybody is allowed to make the correct solution which corrupts the rating of each problem.

IMO giving rating bonus to fast solvers should remedy this...

I don't agree:  I sometimes spend 15 minutes on a problem and still get it wrong!

It would not be good to turn standard into another version of blitz by rewarding fast solvers.  It should either be kept the way it is now with no time limit, or have some very long time limit, such as 5 or 10 minutes, after which you don't get as many rating points for a correct solution.

[newlook]

I agree tmr.  The standard mode is just good for testing your accuracy But is an unfair rating comparison.  Without time, Everybody is allowed to make the correct solution which corrupts the rating of each problem.

If this were true, people who are active in both Blitz and Standard would have higher success percentages in Standard. They don't. In fact, on average, Blitz percentages are slightly higher. Standard percentages are also closer to 50%, which is the target for a rating system. Also, there is a .8 correlation between Standard and Blitz ratings, which means that rank order doesn't change much.

I created a custom problem set of all problems below 1200 Standard rating. It's funny (and common) to see an 1100 Standard problem have a 1400-1500 Blitz rating. These are really easy problems -- obvious mates and hanging pieces and forks without any elaborate setup. So I'd say that they are a better match for an 1100 Elo or USCF than a 1500 Elo or USCF. Obviously there is more to an OTB rating than tactics, but I bet that a person who doesn't have a CT rating and does have an Elo or USCF rating would be better off using the Standard rating of problems as a guide to finding problems suited to his/her ability.

[richard]

Hi all,

At the moment I'd probably prefer look at an alternative rating mode over modifying standard.  While it is true that the longer you spend on a problem , the more likely you are to get it correct, there is a point of diminishing returns on how long you spend on analysis.  Generally users who have high ratings in standard are highly rated because they have better tactical vision than other players, not because they spend a lot longer looking at problems.  Newlook's fairly high .8 correlation value between blitz and standard ratings indicates this. I could spend 3 days looking at problems rated over 2100 standard and still get most of them wrong :-)

As long as users are not using computers to assist them solving problems, I'm not really that bothered by how long they spend, and I don't really see it as unfair. Each user is free to choose how long they spend attempting to solve a problem. If someone thinks others are getting an advantage by taking longer, then they can always take longer themselves :-)

I'll probably have a look at adding another rating type in the future but I have a few other features to roll out before I can consider spending time on that.

Regards,
Richard.

[slacker00]

I also think it would be cool to have more different rating categories.  But I agree with Richard, that it would be more  useful to implement this type of feature  after we've got more traffic, to establish better ratings under any new rating categories.  In essence, focussing on the "quality" rather than "quantity" of rating categories is the best option right now.

But, for future considerations, it would be cool to not only have blitz, standard, and "intermediate" time controls for problems.  But it would be cool to have a rating category for "mate in x" problems, for example.  Maybe  another rating category for "endgame" problems, etc.  I know this is way way down the priority list, but offereing a few extra ideas out there.


Otherwise, I like the standard mode the way it is.  I can really just relax and even take a break to check my email or answer the phone or whatever, then come back and sometimes I'll come back with a new mindset. lol.  I know when I solve problems in this manner, it's not in being really "true" with standard OTB chess, but I enjoy it for now,  and I know that  I need to get  much more hardcore  if I  genuinely want to improve my  OTB strength.  Currently, I am just trying to enjoy chess for it's own sake, if that makes any sense.  Standard is the perfect option for that type of approach.

[tmr]

I don't think it is so much an issue of folks feeling taking longer to solve in standard mode gives an unfair advantage as that the ratings of folks who take vastly different amounts of time in standard mode aren't really comparable.  This isn't a problem but just a fact of the standard mode that one should keep in mind when reviewing ratings and progress.

For me, I'm guessing that there are a lot of lower rated users that are better tactitions than me but their rating reflects they way they've decided to use standard mode.  Since this demographic can change over time, my own rating changes in comparison don't necessarily reflect my true progress or decline but the change in relation to other users during that period of time.  This can change as users come and go, or use the mode differently.

This also makes it is a bit hard to compare the statistics of blitz and standard modes as folks treat the standard mode differently.  This is pretty obvious, just look at the history of some low ranked users (with a fair number of standard mode attempts) and you'll likely see fairly fast moves regardless of the problem difficulty.  I'm guessing that this accounts for many of the occasions with blitz mode having higher success rates than standard mode.  That is in blitz mode the user does problems in about the average time and gets an average number correct but in standard mode they take much less than average and thus get a higher percentage wrong.

Interestingly there are only 109 users active (RD<80) in both blitz and standard as of this writing.  Obviously things can change over time.

[richard]

I agree that there is more room for variability in approaches on standard, but the same is true to a lesser extent in blitz and even OTB play.  In blitz there are users who prefer to spend a bit longer or those who treat it like bullet.  You can see the extreme versions of this on CTS where some users completely ignore ratings and treat CTS essentially like CT standard mode.  The 0.8 correlation between blitz and standard ratings shown by newlook is an indication that despite the potentially different approaches both ratings seem to be measuring a common underlying tactical ability.  I'm planning to have a more in depth statistical analysis of the relationships between blitz and standard and user's other ratings (FIDE/USCF).  I plan to offer a "rating estimator" based on the regression model that comes out of this analysis, however my real interest is looking at FIDE/USCF rating changes over time and seeing how they correlate with different CT tactics training behaviours.

Regards,
Richard.

[henkvj]

Richard,

Forget about my first request. By now I have discovered that your blitz time control is much more relaxed than emralds. So blitz+standard together do a fair job of covering my needs.

Regards,
Henk