What is the chance that anyone from Bethesda will ever actually hear about any of the ideas in this thread? Just curious?
Assuming there is some chance I'll just say the following:
1. Designers should check out http://www.taleworlds.com/main.aspx game "Mount&Blade Warband" and see if it gives them any ideas for practical ways to improve the combat system. Horse mounted combat in particular would be neat. [ADDIT: Just wanted to clarify, I'm _NOT_ suggesting something like the multi-player part of Warband. Just suggesting that the some of the combat dynamics in single-player Warband might spark ideas for improving combat for SINGLEPLAYER play in TES V ]
2. Basic needs (hunger, thirst, sleep, tiredness) as in some of the mods, but seamlessly integrated into the game and balanced with magic, etc.
3. Alchemy should be "nerfed" by making it more realistic. Making potions should not be possible in the dark in a sewer tunnel. Instead, you should have to take the gear out and set it up, get the bunson burner lit, a little mini-game as in the NPC negotiations would be cool. (toggleable for those who just want alchemy to be easy). Making potions should take time.
4. More detailed training (will come back and add a link to a discusion I contributed to in the Gen Discussion forums).
ADDIT Here is the link to the thread http://www.gamesas.com/index.php?/topic/1108023-real-life-leveling-system/page__p__16231286__fromsearch__1entry16231286
and I quote myself from there to make it simple
Spoiler I believe all skills should have at least three axes/domains of development, possible four. At minimum this would be (1) Theory + (2) Practice = (3) Manifest Skill Level.
A more complex form of this would be: (1) Theory + (2) Practice + (3) Realworld experience = (4) Manifest Skill Level.
Theory would come from instruction by Trainers. Practice would come from, practicing the skill, either alone, through sparring, and similar mini-quests with Trainers, guild members, or others that you recruit to willingly train with you.
In the simple version, realworld experience would also count as practice. In the complex version, it would count as having a special value, and the algorithms would be tweaked to make realworld experience crucial for actually advancing skills. The basic idea for both the simpler and more complex system I'm conceiving here is similar to your point that training/instruction should come before practice. Before you can gain anything from practicing a skill you need to amass training points. Training points should cost both money and time, and coolest part would be if there were mini-games similar to the NPC negotiating that you had to beat and which modified the rate of training gain.
Lets say you train up 5 points in Block, costing you 1000 gold and 3 or 4 days spent with the trainer. Now you are ready to go practice that skill and actually start amassing Manifest Skill Level. If you've trained 5 points, then lets say you can go practice up 10 points for a total Manifest Skill Level gain of 10 points.
"Realworld" experience in the more complex version would mean using the skill in a context where you are at risk of harm/loss. In the more complex version, there would be a limit to how much you could gain from practicing (say 66% of each Manifest Point?) and to top off each Manifest Point you'd need to get 2 or 3 points of Realworld Experience.
I tend to think that a complex system like this will not emerge simply because it would appeal to diehard TES fans, but not really to the general market.
5. Fix the Levelling/Major skills system so that the game does not reward out-of-character builds (e.g., Mages with all but one or two Major Skill that they never use, such as: Athlethics, Block, Blunt, H to H, Armorer, Hv Armor, Marksman).
6. Change the training system so that there is less impetus to just do tedious gamey things like jump repeatedly under a dock (see #4 above). Not sure how to do this, but I think TES 5 would be improved if they can do it.
7. Sparring, non-lethal combat, and interactive training with trainers (also see #4 above).
8. More randomnicity to location spawn creatures, less predictability to dungeon populations.