AlexChurchill's experience of RTSs is somewhat limited. but having been recently playing WarCraft 3 and mostly very impressed by its UI, I'd suggest the following:
Advantages from landscape
Land / Sea / Undersea / Air movement and units
Unit formations and orders - 'advance as line' 'advance as column' 'wedge', etc. Customisable (see Homeworld) but with sensible defaults (see WarCraft 3): ranged units stay behind + out of the way of melee units, etc.
Vitenka: There shall be no, none zero, nicht and nada 'cheese' strategies that everyone has to deal with. Tank rushes are the obvious current cheese, but 'build many walls' is equally boring.
As in, every simple strategy should have a reasonably obvious counter? So the only repeatably usable strategy is a non-simple one? --AC
That would be a start. I'd go one stronger and assert that these counters should be possible after you become aware of the threat, rather than being something you have to do in advance. Any 'must do' stuff should be done for you unless you explicitly turn it off. --Vitenka
How much is this a reflection on the problem of inadaquete recon on your part rather than slowness of deploying counters? What I mean is how late are you allowed to become aware of the threat? I would argue that good recon is an important part of RTS, and should be duely rewarded --NR
Following the discussion WhatIsCheese, I don't think the requirement here is about whether the opponent can surprise you, but about gameplay balance, that there shouldn't be certain strategies which become widespreadly successful while also being rather irritating to play against. While this is subjective and there will always be *some* people who'll get irritated at anything, it is a fair general rule to make the more powerful strategies also the ones that are more powerful to play against. This is why recent MtG land destruction hasn't been the strength it once was, because the makers realised it's very non-fun to play against, which I think is a large part of what's meant by "cheesy". --AlexChurchill
There's also a variety of shades of "must do": must do in order to protect against this unlikely strategy X, and vaguely useful otherwise? Must do either A or B within 5 minutes, but which and when is up to you? Must do in order to protect against this unlikely strategy Y, but is actively harmful otherwise? Being a good player consists of deciding among options such as these. --AlexChurchill
Allow MetaOrders to as many depths as your programming language can support. Certainly allow "auto-repair" mode (toggleable) on repair units, "auto-build tanks" or "build 3 tanks then 2 wraiths then tanks until I tell you otherwise" on buildings.
KeyboardShortcuts? for every command are good. Minimap pings (switchable-off) are good. History of recent messages is good. Notification of what "Research is completed", what unit has just built, etc, is very good.
Let's start simple, shall we? My army should do what I want it to do, not what it thinks it should do or what I tell it to do inadvertantly due to a horrible UI. Ideally, I should be able to tell my army to 'win' and then go and play something fun instead ;) --Vitenka
Yes, I was thinking we're going to end up at an EndersGame-style abstraction here. We want truly intelligent line managers and individual fighters - say humans. They get to play a FPS? game, we get to play the RTS directing them. And fortunately there seem to be lots more FPS? fans than RTS :)
Such games exist. An obvious current one is NaturalSelection. Ever since TeamFortressTwo? was announced, people have been beating it to its punch (and solving all of the obvious problems, such as llama-commanders) --Vitenka