It just happened that I became a power user of CPT 3.3 myself again. I partly re-designed my opening repertoire and at the same time realized I forgotten most of my old repertoire.
I have summarized my main observations below.
Editing repertoire
- It’s difficult to keep track which parts of the repertoire are already completely entered and which not. My methodology is to read an opening book and then analyzing which variation lead to my preferred middle game positions. I have the book in front of me and I manually enter all moves and replies. I always enter all possible opponent replies and then each resulting variation step by step. Now, it’s not very likely that I complete the whole opening or even variation in one session. The other day I have to find where I stopped entering variations. Sometimes I only want to focus on the main lines first and variations later. The only way to keep track is manually marking each not complete variation as “not complete”. Not bad, but still too much extra-work and time is money in such a busy world.
- Managing a complex opening with many sub-variations becomes quickly pretty confusing. Scrolling through all tabs is not very comfortable either.
Training
I always thought the applied flash-card concept is almost close to perfection to study chess material. I changed my mind.
- The algorithm to study a new opening has some flaws. It would be best to thoroughly test the user in a whole variation with all sub-variations instead of “jumping” from one variation to another which might belong to totally different main lines. The reason behind this is: Once you played through a whole line most of the positions will have an improved training score. So, if a different variation starts at move 10 the first 9 moves have been already trained. However, CPT will always focus on your worst line as a whole. Thus it will skip variations which happen at a later stage until you have trained the other main lines.
- CPT stores your last 10 training results. If you want to have a 100% score you need to play the right move for ten times in a row. Even if you had 100% and made one mistake which decreases your score to 90% you still need to play the right move for 9 more times to get back to 100%. This is not really logical, but at least “by design”
- The training score calculation is in general not reflecting your knowledge for the position, because it does not take into account that you forget moves over time. At the same time there is no science behind declaring a position 100% only after you played the right move for 10 times.
- Every position is rated equal y difficult.
- CPT will through at you tons of positions before it will repeat one (except those which are part of several variations). Instead it would be smarter to focus on a bunch of positions and only once the player has mastered them CPT should add “new” positions for training.
- Most important: There is no science behind when a position is repeated except the training score. It would be much smarted to only test the user in positions just before he is about to forget the right move. Sounds like magic? It is not!
Now, the bad part of the story is that I “lost” quite some time for CPT 4 development due to my power user sessions with CPT 3.3. The good thing is that all these issues will be resolved in CPT 4. For a very long time I’m thinking already about a new methodology for the training module. Originally, I didn’t want to implement it for CPT 4 yet, but the last two months changed my mind. I’m in the middle of implementing a new sophisticated training strategy. So far I didn't have a chance to test it in practice as the implementation is not yet complete, but I can't wait anymore to finally give it a try and to see if it is really such a great improvement as I believe.
Stay tuned!