diogens said: "This is not exact. If you got your 20 problems wrong, instead of earning 3 points you would have lost, let's say, between 150-200 points."
Well, that's not the comparison I made. Of course, if you miss 20 out of 21, you'll lose lots of points.
I was thinking about comparing to the "average player". This "average player" over these 21 problems, would answer 56.8% correctly, or about 12 problems. Since each of these would be done in "average time", each would generate a little over 0.8 "points" (the score put into the Glicko formula, which is given by a graph on the link above). (This is an approximation, since the function appears to be slightly concave down.) Altogether, this average player would earn roughly 10 points in this "21-game match" vs. the TT. Since the ratings of the TT problems are on par with the average player's rating, what do we expect? Well, what do you expect if two players with roughly similar ratings split a match (11.5-9.5, or 11-10, or 10.5-10.5)? You expect their ratings to remain practically unchanged.
Well, this is what happened with my results. So, according to the TT ratings, my performance is on par with the average player. In other words, taking an additional 27 seconds per problem is as bad as answering half the problems wrong. (Or equivalently, saving 27 seconds per problem is as good as getting all the problems right.)
"Earning 3 points will allow you to have a similar roster of problems next time until you get quicker.... I would say that chess.com TT ratings are focused to longer (than blitz) time controls. "
Maybe longer than blitz. Not longer. Certainly not OTB standard time, which gives 2-3 minutes per move. I think the reason chess.com TT ratings are focused to blitz (or rapid, at least) time controls is because that is the way the players are reacting to them. Without a penalty (or a significant penalty) for incorrect answers, there is little motivation to focus on accuracy, so "average times" (however computed) go down, and pass rates go down, and these both reinforce the same reaction.
The reason I go on and on about this, is that (besides wanting to offer constructive feedback to the site), I find it very interesting that two of the most fundamental and well-accepted principles of educational psychology are completely ignored in designing such a training system:
1. Students will alter their behavior to maximize rewards and minimize punishments. Whatever is rewarded, that's what you'll get more of from the student. This is the whole idea behind grading. Anyone who's taught for any length of time knows this.
2. "Practice makes permanent, not perfect". In every other endeavor of learning, stress is always put on accuracy first, then speed. Music students don't fly through their scales and exercises at top speed month after month, year after year; instead, they practice very slowly until their technique is mastered, then they gradually increase speed. The same applies to other games, from poker to basketball to academic subjects. Only people who design chess tactics trainers seem not to know this.
When you ignore both of these simultaneously, you create an environment where old habits and mistakes are reinforced and repeated, and true improvement stagnates.
Have you noticed how many pathetically easy tactics problems (hanging piece, obvious mates in 1, etc.) are rated 1400-1600 and yet still have 40% of the people miss them? What do you expect?
"They are also telling you that if you don't resolve faster, you mostly get into time trouble or just miss the tactics."
Every expert or master I've talked to, and everything I've read written by serious chess teachers, say the same thing: blitz is detrimental to your game. If you want to improve, don't do it. A local expert told me, "Unless you're a master, anything less than 20-minute time controls is at best a waste of your time. You're reacting completely on instinct and reflex. You can never progress tactically or strategically at blitz time controls."
iused said: "Here it says that it is the average time to solve the problem. That would mean that it does not include the times of those who failed."
That's an inference. However, they do not say if there is a penalty given for incorrect answers. Someone on another thread alleged there was such a penalty. So, I don't think it's clear. If this is the case, however, there's a real problem with that, for reasons I've said above, namely the measure is an artifact of player behavior, without some type of penalty for incorrect answers.
http://support.chess.com/Knowledgebase/Article/View/53/0/how-does-time-affect-my-tactics-rating
Here it says that it is the average time to solve the problem. That would mean that it does not include the times of those who failed.