Guys, i made some data collection to see if a linear connection beetwen player's value and training speed actually exists.
Before making this experience my test for checking if a 3/4* player was set the player to train a SA or a new role and make a 6x Fast Counter Attack train (by my experience, when you train a SA or new role doen't matter if the drill you use train some grey skills). If your player gain 2 "skills" than he's a FT, otherwise he's not. Then i found that also players that weren't FT by my test, but very similar in value to a FT, could train at the same pace of a FT. So I decided to track some players to check if my thesys was correct.
I trained 3 and 4 stars players with full white skills drill for an entire season (level 29), and below you can see the results. The red marked players are 2 "false" FT (only gain 1 SA skill at my test); as you can see when the drill used was in the 60%/80% average range they trained as FT but as the drill get to upper ranges, the pace reduced drastically. As for the other players (the green one with a little bit higher value/quality, the yellow ones all with the same value/quality, and the orange one with a little lower value/quality), it seems that does not exist a uniform model. Some train at a better pace at a range and others trains better at others range, but the average (last column) respects the value/quality criteria, except Mayo (he's spanish, and since i paly this game i never found a spanish FT, did you?) that seems not to be a fast trainer even if he passed the SA test and he has a value/quality of a FT.

Click image for larger version. 

Name:	Capturar.JPG 
Views:	291 
Size:	76.3 KB 
ID:	131216