So whats the .org verdict on RTW?

**Slyspy** · 10-22-2004, 18:50

Hear him, hear him!

Quite right. Most people are fully aware that the AI is not going to be the advanced. However it should at least follow the basic logic of its own game. For example the scissors/paper/stone system. It should be able to react when it comes under overwhelming fire. It should know what unit is capable of doing what and, to take the above example, that while phalanxs can kill cavalry its still a bad idea to chase after them! It should know to wait for reinforcements when out numbered, or to find a defensive position and above all it should know that the General should not be the first one into the fray! Like Spino said, a sensible AI routine. Not a brilliant one, but one which follows RTWs own internal logic.

**hoom** · 10-22-2004, 23:53

IBM and all its resources are only just getting round to building computers that can beat humans at chess

Not true at all.
Making an AI that can beat mere humans at chess is not that hard.
Making an AI that can reliably beat Kasparov, probably the greatest chess player ever, that is an entirely different thing.

The Rome AI is definitely something that needs work though

**Didz** · 10-23-2004, 00:12

Originally Posted by hoom

Making an AI that can beat mere humans at chess is not that hard.

Producing an AI that can beat a human, even a master, at chess is easy compared to producing an AI that can play a strategy game well enough to beat a human.

By comparison with RTW chess is a simple game. The board is of fixed dimensions and featureless and the pieces are of fixed number and value with rigidly controlled patterns of movement. When two pieces collide the result is pre-determined and none of the elements are subject to unforseen risks such as morale or surprise. Everything is perfectly visible and quantifiable thus the value of every move can be mathematically calculated in advance.

The AI for RTW has no such certainties and that has been the problem that so far has defeated every attempt at producing a decent result. Unlike the chess AI which can calculate the mathematically best move from scratch every turn a wargame or strategy game AI needs the ability to learn from its mistakes in order to predict future outcomes. So, far nobody has managed to produce a commercial AI that actually imporves its game based upon its past errors. If they did then the computer would stop sending units to die needlessly under the walls of our cities and try different approaches until it found one that worked. Just as we do.

**Akka** · 10-23-2004, 00:44

Additionnally, the chess AI has only one movement a turn by each side, unlike the hundred of moving units in the strategic map for each side, and the tens of simultaneously moving units on the battlefield.

And remember that a chess AI also have libraries of games, and "standard" openings/moves, and that it takes quite a lot of time to play a single turn (involving a single movement of a single piece on a very limited area with very simple rules and no random effect), while it has to do all in real-time on RTW (with many sides, many movement on each sides, a HUGE area, and MANY random effects)....

**Kaiser of Arabia** · 10-23-2004, 00:51

Guilty

**DisruptorX** · 10-23-2004, 02:41

Originally Posted by Didz

So, far nobody has managed to produce a commercial AI that actually imporves its game based upon its past errors. If they did then the computer would stop sending units to die needlessly under the walls of our cities and try different approaches until it found one that worked. Just as we do.

Well, in starcraft if you slaughter the computer's zeolots with air units, they will send dragoons as support the next time they attack. You think that RTW could have at least done little things like that which make you feel like you are playing against more than a pre-scripted AI routine.

Starcraft came out what, 6 years ago? It has AI that learns at least a little from what you do. When you defeat the computer using a tactic in Starcraft, in the future it will be ready for it.

**Midnj** · 10-23-2004, 01:14

MP is atrocious and probably is not something that can be fixed to an acceptable level because many of the problems are integral to the single player game too and not really problems there, so the odds that CA will address them are about nil. They will cater to the money, can't really blame them for that, but I am dissapointed they put multiplayer in when it's so broken.

They should just take "multiplayer support" off the box, because it's a complete distortion. I felt so strongly misled that I returned my copy of RTW and downloaded a uh, "backup" copy. I'm not playing it of course, because that would be illegal.