My Ideal Tactic Trainer

In a recent thread, I made a point that I felt the tactics trainer here was pedagogically flawed. 

I thought it might be worthwhile to have a discussion about what features we would like to see in a tactics trainer that would make it "ideal." I make no pretense that these ideas are entirely my own. Nor do I think that my ideas would result in a perfect system. But I thought I'd outline the features that would be most important to me and see what others think. I'm sure other people have ideas that are equally good.

Maybe someone will take these ideas and put find something to do with them. Since I am not a programmer, that someone isn't me.
The ponit of this is a brain-storming session for an ideal tactical trainer. I'm not suggesting that chess.com, chesstempo.com or anyone else should "fix" anything. As I've pointed out repeatedly, while I have issues with the TT here, I don't see those issues as a flaw with chess.com as that isn't the core business here. Rather, I'm just trying to have a positive discussion rather thant he recent thread where I found myself in the role of critic a bit more than I like.

1.  All the basics that most tactics-only sites do: 

a. Graded, checked problems covering every tactical theme

b. Problems are sortable based on theme, difficulty, etc.

c. Problems come from opening, endgame and middle game with emphasis on middle-game

d. Problems selection can be customized based on theme, difficulty, etc.

2. Default scoring is based on correctness of solution, but with time mattering to relative score. 
a.  Correct problem solutions should never be punished no matter how long it takes, but quickly finding the right solution should be rewarded. 

b. Quickly finding the wrong solution should be punished more severely than taking a long time to look at a problem but then making a "less than ideal" move. 

c. Punishing the "wrong" move should be relative to the badness of the move relative to the best available move. Making a -9 eval swing move should cost more than making a -1 eval swing move and that should cost more than making a +1 eval swing move if the best move is a +3 eval swing. Postive eval swing moves should never punish the player but should not reward the player as much as the best move.

3. Problems are presented in a mix of "user to play" and "opponent to play" format, so that the student will sometimes be looking for the move for the other side.

4. A seperately scored set of problems (probably which can be toggled off) to present "is it safe" problems where the question is not to find the tactic but to answer the question of if a particular move will lead to a tactical refutation.

5. Have problems interspersed within the problem set where there is no tactic on the board! The inclusion of "null" problems would help train people to real game situations. This shouldn't be frequent, but should be possible. Finding problems that look like they contain common tactical themes, but which don't, certainly would be a challenge. 

6.  Problems which the user got wrong should be recycled into the problem set faster than mere random selection to so that they are seen again. 

7. Full statistical reporting

8. Problem selection based on user history, not just rating range once the sample size is large enough. If the player is scoring 85% on hanging pieces but is scoring 50% on removal of the guard, then removal of the guard should start to appear more frequently. Some level of randomness should always be present. 

What about you guys - what feature(s) do you want to see in your ideal tactics training system?

That would be a nice tool indeed. :) 

I would like to see the problems/solutions played out to obvious conclusions. Many times, when I make the wrong move, I can't see the rationale for the "solution" and I am told to "view analysis and source" which also, many times, doesn't make sense. I'm sure that to higher rated players, most of this is obvious, but why not carry them all out so that us patzers can enjoy and always understand.

Thanks.

That's another great point.