Blitz rating system

[drahacikfm]

Some questions that didn't seem to be answered in the FAQ, concerning how blitz problems are rated:

1) When you fail a problem, do you lose the same number of rating points, no matter how long you took to make the mistake?

2) When you take much longer than the average time to get a problem correct, you can actually lose rating points.  Can you ever lose more rating points than if you failed the problem?  In other words, is it better to just guess than to take a huge amount of time?  Or is it always better to get the problem right, no matter how much time you have to take?

And a suggestion:
If I understand correctly, as you take longer than the average time to solve, the rating change you will receive continuously changes.  But if you solve faster than the average time, you get one of two fixed amounts:  either with the bonus or without it.  Why not make this continuous also, increasing your rating gain the faster you solve, all the way from the average time down to some very short time, such as 2 deviations faster?

[richard]

Hi,

1) Yes, when you fail you always lose the same amount of points.

2) It is always better to get the problem right (unless you are a very lucky guesser :-)  ). You can never lose more points on time than you would lose for an incorrect problem.  The longer you take the closer you get to losing as much as a wrong answer, theoretically you never get the point where you lose as much as a wrong answer as the time penalty should asymptotically approach the wrong answer penalty so unless you have infinite time spent you'll never be as bad as getting it wrong (although eventually floating point rounding errors will have you reach the wrong answer penalty before the end of the universe - but even then you'll never do worse than getting the answer wrong).

You do understand the time bonus correctly, it is currently a fixed amount and I agree that having continuity across the whole scale would be nicer.  I'm planning to tweak the bonus a little in the near future to try and deflate the blitz ratings a bit so I'll look at moving to a continuous approach when I examine this.

On a related note, The CTS guys apparently stated in the past (I only read this second hand) that the main reason they have such insane time limits was that rating inflation was a rampant problem without the time constraints.  I believe that looking at the standard ratings here is evidence that if your problem set is hard enough , harsh timings are not required. Of course there are still a few artificially hard problems with alternative lines which makes it a bit hard to reach solid conclusions yet, but I think I don't need to worry too much about that for now.  Rating inflation due to too large a bonus seems to be the problem for CT blitz, rather than allowing too long for the problem as such (otherwise standard ratings would be through the roof).

Regards,
Richard.