tl;dr version: Anyone have any TBStrategy games that don’t fall into common AI tropes?
I’m a big fan of turn-based strategy games, typically referred to as 4X-style games. I’m also kind of good at them (like a world champion body builder is “kind of good” at lifting things).
I’ve played a lot of them, and sunk a very large number of hours into them, and while I love playing them, the AI is, shall we say, disappointing. It isn’t capable of figuring out when to hold them and when to fold them, as the case may be. It isn’t capable of realizing that certain strategies are good at different times in the game and not so good at other times. To borrow from chess, you have a beginning game, mid game, and end game. Using a beginning game strategy end-game doesn’t work.
This saddens me. It also makes me wonder if I can do better. Sadly, I’m not much of a coder and I’ve never tried making an AI (well, more like a decision algorithm), so I have no idea if I could do better or not. However, I can be an armchair developer/designer and pick at AI foibles.
Problem #1: Deciding when to attack.
Source: Master of Magic. The sole remaining AI is Ariel.
In Master of Magic, you’re playing a wizard/leader in control of powerful magic and mighty armies that can sweep the world. The AI, however, decides power purely on army size and ignores everything else. While it is true that, in the beginning of the game, army power is everything, even by mid game this falls flat on its face. Sure, the AI could have ten thousand halfling slingers, but it doesn’t do a damn bit of good against the reprobate spell-slinger burning the pure living crap out of the world (namely my wizard).
An AI would see that graph and assume that, since the AI has a significant army strength advantage, it should attack the player. It is ignoring a few key problems – such as the fact that it has huge numbers of crappy units (whereas I have a medium number of good units instead), or where those units are positioned (they have a bunch of units on their capital island and a few nearby islands. Coincidentally, so do I), or even that I have such a metric asston of magic built up that I could simply waltz into their capital with a single unit and that unit would never need to fight – I can destroy everything with magic alone.
I, on the other hand, look at that graph and know exactly what is going to happen – if the AI attacks me, they will be curb stomped. Just like the previous three AIs I wiped out within five turns of each other that had far more massive army strengths. This isn’t limited to older TBS games either; Civ4 is notorious (in my mind) for this exact same thing; they’ll compare army strengths and ignore economic strengths and locations of their units vs. the location of my units. These wars are often quick and painful for my foes. Just about the only way they’re able to handle themselves is because of AI cheating (even then it doesn’t help out enough once I get used to a game).
Problem #2: How to defend.
Source: Master of Magic. I’m only picking on this game because I’m addicted to it [again].
AIs can’t defend worth crap – or at least they can’t defend important things worth crap. Take the above screenshot as an example. They have an astronomical number of units (each of those blue icons represent a full stack (9) of units of some variety) on their home island. The purple icon is my single unit stack (not even full – 8 units) and immediately next to it is the AI capital.
If those units weren’t all weak sauce, they’d be doing a decent job at defending the island itself. Ignoring the weaksauce issue (see issue #3 below), there is still another major problem with this.
In Master of Magic, you lose if your capital is taken and you don’t have enough magic to cast the Spell of Return. Honestly, you’re pretty much screwed even if you do cast that spell, since it takes a large number of turns that your foes will use to just mop you up, so losing your capital is tantamount to losing the game.
The capital itself has nine units in it, just like the rest of those tiles. The problem is that they have their entire lower flank left empty and have crappy units (even crappier than what is around) defending the capital. All I needed to do was waltz on into the capital and fight a very easy battle to slay this AI. I didn’t need to take a single other city, as long as my stack-of-DOOM! was strong enough to get to said capital. Hell, with the way I’m set up, even a SINGLE UNIT walking in to the capital would have taken out this AI.
So, what happened? The AI was too busy making crappy units and casting offensive spells with her armies to bother defending herself. AIs in turn-based strategy games are utterly incompetent at defense, mostly because it isn’t easy. It is far easier to make an AI know how to zerg rush a player than it is to adequately defend itself against a player. In this case, for instance, all it would have taken would be to protect all sides of the capital. In MoM, you can’t attack more than once per unit per turn, so I would have either had to either split my stack up to attack multiple units to get to said capital, or take multiple turns in order to reach it – which, in that time, the AI could have thrown better units inside said capital. Or really much of anything beyond the crappy wolf riders (scout-style units) she had in there.
My own capital is defended by four units, total. There are no units immediately outside, all of my flanks are wide open, and I’ve had a lot of attacks (this is the first time the AI has been attack at her capital, to my knowledge). So why haven’t I lost? For one, I actually put four good units in my capital, I have magic held in reserve to defend said capital, and I have enough people scouting around to see if an AI would approach so I can shore up my defenses as they arrive. 4X games often reward Just-In-Time defensive systems over constant ones, as you need to do maintenance on units that you have just sitting around; it is more economical to provide great defense to areas that you know will be hit and token defenses to the rest than it is to provide mediocre defense everywhere. You can always rush good defenses to a location if you’re set up to do so.
In other games where losing your capital is just doubleplusungood rather than ‘you lose’ level of bad, AIs often allow their economic areas to be weakly defended, or have other big gaping holes in their so-called defense that makes felling them simply an exercise in playing Where’s Waldo (or Where’s Wally, for the rest of the world). Same concept, although admittedly a smidge trickier to defend. That brings me to..
Problem #3: Choosing the right units for the right situation.
Source: Sid Meier’s Alpha Centauri.
Back when I was younger, I used to always use the same military units for everything. In SMAC, I’d try to build a bunch of Skybases (flying units that didn’t need refueling, but ridiculously expensive) to attack everything, defend everything, and scout around everywhere. I’d lose, sometimes before I even got to said unit.
Why? I wasn’t effectively using units right. I was trying to shoehorn in my favored unit into every situation possible, even when it wasn’t the correct one; I’d ignore everything that came before as merely a means to using my favorite thing.
Eventually, I became better at the game and realized that combined arms style of strategy (either multiple styles of unit in the same stack or multiple stacks depending on the game and situation) was far superior to what I was doing before. Generally, that’s when I started winning a lot more in SMAC, becoming the SMAC veteran I am today.
Problem is, the AI never learned these things. Either they use gobs and gobs of the identical units, gobs and gobs of older units, or they spend an astronomical amount of resources on building new units perpetually instead of using old units in ways that are still useful.
For an example: In SMAC, you often see AIs have a huge number of needlejets. Sometimes, they use them for economic warfare – which is fine, that’s actually a role that they’re useful for. They also use them for garrison though, which is a terrible idea. They use them for patrolling, which is an okay idea but not all that great. They use them for a lot of things, and hardly ever for an actual attack (their primary purpose). They’re okayish at picking out good garrison (they’ll adapt to your strategies over time), but they often fall by the wayside so the AI can build more needlejets to harass you with.
Another example: In MoM, the AI will have a metric ton of units that are high in mobility. Calvary, wolf riders, paladins, hell hounds, things that can move more than one square per turn.. and then use them as garrison. They’ll practically never upgrade their units (so if they have a bunch of wolf riders toward the early parts of the game, they’ll still have a bunch of wolf riders by the end instead of loading up on unique units from varying races and summons. That earlier picture of my stack next to their capital? Their capital had nine wolf riders in it with nothing else. My stack had four heroes (two melee/speed, two magic/ranged), a priest (healing/ranged), a wolf rider (speed), slingers (ranged), and pikemen (melee/anti-speed) with a bunch of offensive magic to back them up – and this was just because my stack was so far away from my own cities that I didn’t have a chance to upgrade them to better units.
The AI was more than capable of having a similar stack; in fact, they had several of the same heroes I did in random spots of their empire. They just didn’t, as they were using them for random other stupid reasons. Those wolf riders would have been great if they weren’t stuck on a relatively small island – load up the bajillion transports they have and start invading me with them at least! They were the ones to declare war on me, after all.
Final example: GalCiv2. Ye gods, what is with this game’s obsession with a huge number of really crappy ships? Sure, you can add them to a fleet, which allows the stack to attack as one, but that makes it… worse than a single modern ship. It is like the AI never wants to make new ships because it sees a bunch of perfectly usable ships, even though they’re ridiculously crappy, and continues making the same ships.
I could keep going on, but I’m already well beyond tl;dr range. I wouldn’t mind playing a new TBStrategy game that didn’t have these issues… but I don’t think they exist.