Warzone 2100 Forums

I'd like to start a debate here where we talk about all kinds of strategies that may be employed in multiplayer games, and how to best counter them. While this thread may seem to be mostly useful for competitive gamers, I believe that it's also useful for AI developers.

Let me start things off:

Say that you're playing on a map with lots of open space and no chokepoints, meaning that both you and your opponent are vulnerable to flanking attacks by ground forces. Say that both you and your opponent have equally-massive forces (over 100 units each), but your opponent launches an all-out attack on your flank, away from the bulk of your forces. How would you prevent or deal with this?

Personally, in a wide-open map, I'd say that turtling's out of the question. The costs of protecting such a wide-open stretch of territory just don't seem to be worth it imo. Rather, I'd consider my territories to be pretty much doomed, though I do have a plan to stay in the game. I'd not only launch an all-out assault on my opponent's territories, expecting (and hoping for) minimal resistance (assuming that we both had access to equal amounts of resources, and assuming that the opponent neglected in developing defenses for the same reasons I would), but I'd also bring along a few trucks to rebuild within my opponent's former territory, cancel all factory production, and have what trucks I leave behind within my doomed territory demolish whatever base structures they can salvage before they get destroyed by the imminent onslaught of enemy units. That way, even if my base gets destroyed, then not only would I most likely destroy their base as well, but I'd probably still have a few trucks in hand, along with a little power from salvage and cancelled production orders, to potentially stay in the game after suffering such a devastating loss.

Shadow Wolf TJC wrote: Say that you're playing on a map with lots of open space and no chokepoints,

I wont play that. lol. But say if I play that anyway, it will be just for fun and hence, strategy talk is useless.

Shadow Wolf TJC wrote: Say that both you and your opponent have equally-massive forces (over 100 units each)

It's much more impossible that we both stack that much as I will attack a lot before we reach that (as soon as 4 units probably)

Shadow Wolf TJC wrote: but your opponent launches an all-out attack on your flank, away from the bulk of your forces.

Since he moved from his front, he do not "launch" anything since I am already in his base. Because I constantly pushed from the very beginning.

Answer:
I already stopped my research (since the game end here), increased my production, sent most of my trucks spread around the map to hopefully save one from the massacre and have a chance to rebuild and the remaining of the built some defence in the middle of my base to defend the factories a bit longer... Since I reached his base before he reached mine, I'll probably win this game easily...

I want to say that it is difficult to plan strategies and counter-strategies without at least a picture of the map preview...

I believe that it's also useful for AI developers.

Uhm unfortunately no. Map geography is completely unavailable for AI developers, so things like "where to move units", "where to put defense", "put short-range tanks in the front", "attack the closest target" etc. are pretty much out of question.

@Iluvalar: Sounds quite similar to the situation that I set up (though with fewer units), to the point where I'd pretty much counter your strategy in the same way that I'd counter mine, by launching my own raids into the opponent's territory (though I wouldn't expect to storm my opponent's base), escorted by some trucks to rebuild in said territory, though given that fewer units are in play, the chances of me losing my main base are reduced (or delayed considerably) if I turtle a small area around my starting derricks (assuming that the base was as open-ended and lacking in chokepoints as the rest of the map, though if there existed a chokepoint close to each player's starting base, then turtling there would work even better). I'd likewise stop or greatly slow down my research in order to spend more of my power on producing enough military units to fend off yours, that is, until we reach a stalemate.

Speaking of, given the wide-open area, I'd probably expect far more tanks and Cyborgs to come in play than defenses, so Flamers would likely be very useful on such a map. Given how flamer users would focus more on mobility over defense, I'd likewise focus on more mobility over defense, along with focusing on either the Machinegun or Rockets line, if I saw that my opponent was indeed focusing on Flamers.

@NoQ: Are you sure? I'm thinking about starting work on my own AI, and I'd like for it to execute many kinds of strategies, some tried-and-true, others more unorthodox, daring, or even experimental, and this thread seemed like a nice place to talk about said unorthodox, daring, or experimental strategies.

While AIs may not be able to identify their surroundings by themselves, there exists a global called "mapName" that AI developers could use to have their AIs automatically re-tailor themselves for particular maps, either by using a specialized script for each map (which would likely take up lots of space), or by using a more all-purpose script for specific types of maps (which, although less effective performance-wise, would be more cost-effective). For example, in a hover-only or VTOL-only map, developers could program their AIs to try to develop hover or VTOL propulsion ASAP.

For example, in a hover-only or VTOL-only map, developers could program their AIs to try to develop hover or VTOL propulsion ASAP.

NullBot does that for hover-only maps. Not for VTOL-only maps yet. The only way to identify a hover-only map is making droidCanReach() calls, but in this call it's very hard to identify whether the droid that "can reach" is a hover or a ground tank. Thus, we have a problem even identifying VTOL-only maps.

Making an AI that works on only a few maps is possible, of course (: But this doesn't solve the combat tactics problem. A problem that i completely fail to even start solving. The convex and concave formations, surrounding the enemy, the advantage of pursuing over retreating, highground advantage, etc. The AI just can't understand this sort of thing (well, quite a lot of humans don't understand it either).

Shadow Wolf TJC wrote: While AIs may not be able to identify their surroundings by themselves, there exists a global called "mapName" that AI developers could use to have their AIs automatically re-tailor themselves for particular maps, either by using a specialized script for each map (which would likely take up lots of space), or by using a more all-purpose script for specific types of maps (which, although less effective performance-wise, would be more cost-effective). For example, in a hover-only or VTOL-only map, developers could program their AIs to try to develop hover or VTOL propulsion ASAP.

Successful maps don't have a static strategy but rather mixed strategies.

In a balanced mixed strategies nash equilibrium, there is no need to instruct the AI about what strategy to use. Because players are free to take whatever strategy they feel like. It's also the only situation where players need to react to the opponent strategy which is also know as a "game".

If you find a map that favor a strategy over the weapon modifiers margin (and armor layer margin if it's fixed in your favorite mod ). It's a lost cause. Fix the map, not the AI !

Maps are made to destroy strategies and AI. And the favored MP rush is not a strategy. Its boredom. It would be nice if there was an RTS where rush tactics could be tossed out the window.

Lord Apocalypse wrote:And the favored MP rush is not a strategy. Its boredom. It would be nice if there was an RTS where rush tactics could be tossed out the window.

WZ is one of those strategies. In it, every rush is countered by a little less rush (by technological advantage developed during the time you managed to acquire by rushing less). The exact details are depending on the map size.

So i have no idea what are you talking about.

@Lord Apocalypse: That is true, but what maps destroy what strategies? That's a question that AI developers would no doubt need to tackle, and that's something worth exploring here in this thread.

Speaking of, what maps ARE ideal for foiling rushes? I'd imagine that a really closed-off map full of chokepoints, bottlenecks, and either Scavengers or some really durable features (akin to StarCraft 2's destructible debris) along the middle could do the job against traditional ground-based rushes, though it wouldn't prevent artillery or VTOL rushing.

Edit: Perhaps I should consider creating some challenge maps for players to be able to explore these kinds of situations?

NoQ wrote:

I believe that it's also useful for AI developers.

Uhm unfortunately no. Map geography is completely unavailable for AI developers, so things like "where to move units", "where to put defense", "put short-range tanks in the front", "attack the closest target" etc. are pretty much out of question.

Kinda.
That is exactly what gateways were used for before...(among other things)

Shadow Wolf TJC wrote: Speaking of, what maps ARE ideal for foiling rushes?

hmmm.
Tokenring. most of the minis, garden

Those are the one that still fun occasionnaly. Now the one that for me are broke for the same reason :

Squared, cockpit, cockate, highground, many more...

oh thinktofin !!

Iluvalar wrote: ...Tokenring...

In that case, I had better look for better opponents when I play that map... haven't played garden so I don't know about that.

Regards, Searge

Iluvalar wrote: ...
Those are the one that still fun occasionnaly. Now the one that for me are broke for the same reason :

Squared, cockpit, cockate, highground, many more...

oh thinktofin !!

What is wrong whit cockate?

The distribution of oil. With ~50% of the oil straight in the centre. It cause the players to rush madly to the centre to a point were the building orders and research choices are screwed. The first player to get the control of the centre correctly is unbeatable. Generally it's around 5 minutes in the game.

From that point, unless the dominating player do a mistake, he is about sure to win. However, unlike some good rushing map, cockate offer a disturbing turtling position that forbid the winning player to attack on the front. He is forced to sit on his position and research another 20 minutes before giving the final assault. which is just a waste of time for everyone.