Assuming this isn't done already, an interesting idea would be to have an adaptive variable that will weight the selection criteria. If the previous tactic was failed, the chance of a lower rated tactic will increase and if passed the chance of a higher rated one will increase.
Tactics Selection Algorithm
Sort:
That's way too logical and well researched Martin. This is the chess.com forums. Please start acting like an idiot immediately.
Thought Martin Stahl was a staff. He used to have the green pawn icon instead of the diamond icon
Jun 14, 2017
0
#7
I've noticed an improvement yesterday and today. Will keep tracking but things were much more balanced on my last couple of sessions (by feel) and objectively better on my last session. Feels like a change was made, to me, though I guess it still could be a random upswing 
|
% |
||||
| 200+ | 5 |
20 |
||
| 100-199 | 3 | 12 | ||
| 0-99 | 2 | 8 | ||
| -1-99 | 4 | 16 | ||
| -100 - 199 | 2 | 8 | ||
| -200 | 9 | 36 |
First off, I want to say I appreciate the new selection algorithm that pulls tactics from a much larger range of ratings. It was something I asked about in 2009 and I really like it
https://www.chess.com/forum/view/suggestions/tactics-trainer-option-for-lower-rated-puzzles
That said,it seems that the last set of changes may have caused a larger spike in selecting tactics on the lower end of the range rather than trying to spread it out and be more even. I guess it is very possible that the sampling I have been doing has randomly been on the low end but I don't think that is the case.
Here are some stats from the tail end of 5 of my more recent tactics sessions, over a few days.
46% of my tactics from my sampling have been more than 200 points below my rating and 76% have been lower than my rating. It would seem like the algorithm shouldn't be weighted that much on the lower end.
I didn't document during the earlier changes but it seemed like it was more like evenly distributed then.
I'm still getting a good mix of actual tactics, so from a quality standpoint, that isn't a big deal. This process seems to put a lot of downward pressure on ratings and other topics have brought that up. I understand trying to keep TT from getting inflated ratings but is seems like the attempt should be trying to balance them overall.