Training would be a down time activity, and includes things like training to get proficiency or expertise with a skill or tool.   Xanathars suggests something like 3+months of dedicated downtime to gain new skills, which is realistic but wont happen in game. I decided on a system for languages, but wanted to get your thoughts on how to treat training skills going forward.   James and I were also talking, and we thought that training should also apply to Ability Scores. This would add a house rule in that ASI is not an option (ability score improvements) at the appropriate levels, but it might make it so you improve ability scores more often.   I would (for more realism purposes) require it to be tied to downtime, and take real time and money investment. Want to up your strength? Join a gym and go to it two hours a day, 5 days a week. Want to up your intelligence, take a class at the university. That type of a thing.   I have two ideas on how to approach this. One is to record hours dedicated to improving that skill (and you can train more than one at a time) and at regular intervals (maybe every 20 hours? every 100 hours?) you get to roll to see if it improves. In 2nd edition, we roll a d20. We have to reach 100% in that ability before we improve it. When we reach the time to roll (so leveling in 2nd edition), we roll a d20. If we roll over that ability score, we get to improve the % by 2d10. If we roll under the score, we improve the % by 1d10. Here's an example:   I have Wisdom at 18 / 56%. My score is 18. When I reach a threshold to see if it improves, I roll a d20. I get a 19. I get to add 2d10 to that 56%. When that 56 reaches 100%, my wisdom becomes 19. This works in AD&D because % is used for most of the rolls. I think there is an easier mechanic we can use here.   But the basics are Ability Scores can be improved outside of the every 4 level option, and the more close to perfect that score is, the harder it is (and perhaps should be) to improve.