Seems like you've got pretty large errors in methodology when you are relying on users to remember to hit the buttons. Your results are probably going to be so skewed towards the actions where users remember to actually hit the buttons that they will be effectively worthless for measuring what they are actually doing, rather than what they hit the button to show they are doing.
Better choice is to actually get the telemetry data game developers actually scrape from their AAA games that show these exact things then analyse that.
Or better yet, put in a hook for when the combat music plays and analyse that "variation in the amount of time spent in combat across as number of titles" and then just ask your users to do a playthrough with a mod that maps the amount of time the combat music is playing. Then you can compare which titles have proportionally more time spent in combat.
The program/script includes an overlay which is displayed over the game which shows which key corresponds to which type, so you can always see which key to press. At the bottom of the screen there is a bar which displays "F1 Non-Objective · F2 Objective · F3 Interact · F4 Decision · F5 Consumption · F6 What? · F7 External", and the one selected lights up in green. You may occacionally make mistakes, but from tests I made with a couple of friends, pressing the right key when relevent went well almost every time.
Do you mean I should take telemetry data that are publicly available for the games, or do you mean I should ask the developers for telemetry data, or do you mean I should myself collect telemtry for the games? For the first one I would have to be lucky for there to be telemtry available by the developers, and for it to also be useful to what I'm doing. For the second one, I don't think the developers would want to give me telemetry data just because I asked them. Finally, as for collecting telemetry data myself, which your music example is a type of (as I understand telemtry), I have considered collecting telemtry before, but felt that it would be inadiquate for what I want. I want to preferably know what the players consider themselves to be doing, and not just what I can see them doing. For example, the player may be making a decision about what to do, but not through a in-game presented decision, it could be about deciding in which direction to go, so I wouldn't then know they're making a decision. Or it could for example be that they're presented with choices but don't actaully make decision and instead chooses one randomly. There's nothing ingame, game mechanics and such, that can let me know that. I also can't easily know, if I'm using my gameplay types, for example whether a player is randomly exlporing (non-objective) or if they're following a quest (objective). Another factor is while I might be able to do it with the music for combat, I don't want the amount of combat to be the only data I collect, I want data about the other aspects of gameplay as well and compare them, as I don't think I'd get any useful data otherwise. If I want to categorize every action a player does into one of my gameplay types, I would need triggers for almost every action a player makes eg. I'd need a trigger for combat, a trigger for picking up an item, a trigger for a cutscene, a trigger for loading screens, a trigger for opening the escape menu, a trigger for dialog, a trigger for opening doors, etc. And I would also do that for multiple games, which I think probably would be difficult to do within a viable timeframe, although I haven't looked into it. Also the combat music thing might not always work, as just because there's an enemy nearby doesn't mean you're currently fighting it, you may be able to ignore the enemy if it's not super close, or another scenario is that you ride your horse past the enemy and ignore it.
