I play computer strategy (RTS) games. Command and Conquer, that kind of thing. If you don't know I'm on about already this idea will make no sense so you should quit now with my apologies.
Right -- whenever a game is played on the internet, a central server should collect stats on who won, and what
weapons (tanks etc) and buildings they chose to build.
Now, simple. Inject market economics. Each night, the server thinks 'Wow, 70% of the winners built mainly [hover-mega tanks]. From tomorrow, they're not gonna cost 300 'resource', they're gonna cost '350'.
Or 'Wow hardly any of the winners ever bother with the 'Spider mine layer' -- tomorrow their cost gets reduced by 15%. If that doesn't work .. the next day .. another 15% -- UNTIL the results prove that unit is remotely useful (if cheap and cheerful!).
Advantages:
1) All the units that are programmed in get used over time (get cheaper and cheaper until they are). Currently most RTS's have units that are soon made as good as extinct, as the developers have simply got the price/usefulness to the army wrong (so only losers ever even build them)
2) The games at the moment are a pretty much a case of 'Just learn what the best weapon is/build order is and do that then storm the enemy'. This will require more dynamic thinking .. and .. heaven forbid .. actually using real time invented tactics and strategy, rather than the traditional 'learnt off by heart build order' then units!
3) I'll no longer get beat by some 13 year old kid not because he is brighter than me, but because he spends 13 hours a day just learning off by heart the fastest way of building the 'apocolypse bomb raider'. No 'real time thought' involved -- just a pre-practiced build order.