AI War:AI Reinforcement
How Do Basic AI Reinforcements Work?
A: In addition to the incoming waves that the AI players send against the human players, the AI gets occasional reinforcements to its own planets. These are the only two ways in which the AI is able to get new ships after the game has started.
AI reinforcements come in cycles per AI player, like the offensive waves do. These reinforcement cycles are a bit randomized, and a bit based on the difficulty level of the AI player (higher difficulty level AIs get more frequent reinforcements, same as they launch more waves).
There are two basic determinants to how an AI player reinforces: how many planets does the AI player NOT control (so, unclaimed planets and player planets), and what the AI Progress is. During a reinforcement, the AI player can do a full reinforcement to however many planets they don't control, and the number of ships added to each planet scales upward based on the AI Progress.
An AI player can never do full reinforcements on more than around fifteen planets in one cycle (the highest difficulties -- 8, 9, and 10 have planet limits of sixteen, seventeen, and nineteen, respectively). It is possible that the AI will choose to do a double-reinforcement to certain high-value planets, in which case it reinforces fewer planets overall, but reinforces them more heavily. Either AI player can reinforce planets belonging to either of them. So if a very important planet is attacked, they might together put as many as 4x reinforcements into that planet in a single pair of cycles (which might be staggered, or might come nearly all at once).
At each planet that is reinforced, the AI player will get ships at every command post and also at the command station. The number of ships added at each post/station is based on the AI Progress. There is no cap as to how many ships an AI player can have at any given planet, although the AI will sometimes choose to turn excess defensive waves to offensive purposes (the "Attack Forming at Planet X" warning).
There is ship cap per planet with the AI: 250 x the number of command posts and stations, with a minimum cap of 250 and a maximum cap of 2500. After that number of ships per AI player is reached they won't keep reinforcing. They'll fill up all the planets on the map to that degree and then stop. They won't reinforce to any non-alerted planets unless all of the alerted planets are full up, in which case they will reinforce any other planet they control. It is possible for a single reinforcement to send the AI over the per-planet limit, but no further reinforcements will be made once the per-planet limit is passed per AI player.
Examples
Take the following map as an example. I own A, computer owns B, C, and D:
A - B - C
|
D
Q: If I turtle on A the computer will stack all his reinforcements on B?
A: Yes. Until B fills up, in which case he will start putting them elsewhere -- maybe on C and D, or maybe on E, F, Q, Z, A1, A2, etc.
Q: If I conquer B the computer will split the same number of reinforcements between C and D, assuming I do not attack any of the command stations on them, which will each receive roughly 1/2 the amount of reinforcements that would have gone on to B with the full number of command stations.
A: Yes, you can be assured that he will not place reinforcements on any planets that do not have an alert level on them, unless all of the planets with an alert level have filled up. "Filling up" depends on the number of guard posts and command stations at that planet.
Q: If I move on to B, kill everything but the main command and warp gate, the retreat back to A... what now? Will he still reinforce C and D, or will he still put all his reinforcements on B? Is a reinforcement a set # of ships or does it depend on the number of command centers on that planet? For example, will he reinforce 100 ships, but he is only able to put 10 on B because I killed all the command centers rather than the full 100, so the rest go elsewhere?
A: The number of ships that it takes to "fill up" B will then be vastly reduced, and so it will more quickly start spilling over to wherever else in the galaxy. Maybe C and D, but often not. Sometimes on their home planet or other planets they want to defend more for whatever reason. They can't do any more to short-term defend against the planets you directly threaten, so they move out and accomplish other stuff instead.
Q: I guess the question is does the AI get a set number of planets to reinforce (he can reinforce 100 ships to a planet with 5 command posts or 20 to a planet with 1 command post as 1 reinforcement) or a set number of ships to reinforce that he must spread out according to the available command centers?
A: Yes, a set number of planets per reinforcement event. This is dependent on how many planets do not belong to the AI, as discussed in the main body of the article above. The ships are a set number per planet, roughly speaking, so if there are fewer guard posts then it will reinforce more heavily at the remaining ones. But it's using rough percentiles, not fixed numbers, so the numbers won't line up exactly. It's in your favor to have fewer posts there, let's just say.
Q: Also, if I neuter a planet will he ever rebuild the guard posts? Do the guard posts on metal and crystal harvest points do anything for the AI reinforcement wise?
A: "Neutering" a planet refers to leaving the command station, but destroying all the guard posts -- so the planet still belongs to the AI. The AI will never add more guard posts. They gain no resource benefit from any structures, as the AI does not use resources. It's productivity is based on: AI Progress, # Planets it does not control, # guard posts at each planet. So when you cut down on the guard posts at a planet, you majorly hamper its productivity locally, but that does nothing to affect the overall productivity of the AI elsewhere.
AI Alert Level
How Does The AI Alert Level Affect Where The AI Reinforces?
AI players will generally try to do more reinforcements to highly-contested planets, planets they feel are more threatened by the players, or their home or core planets. The galaxy map contains information on which planets are likely to be reinforced based on actions the human players are taking -- these planets are shown with a red border in the galaxy map, and show "AI Alerted To Your Presence - Reinforcement Is Likely." This indicator is not a guarantee that an AI will reinforce any given planet -- far from it -- but it's an indicator that the AI is at least considering it. Keeping aware of this indicator as you expand into the galaxy is a good idea, since if you alert the AI too much to your presence, the going might get too rough.
In particular, it is a VERY good idea to not get too close to the core or home AI planets until you are ready to take them. Since they both will reinforce with Mark V/Core ships, it can make things very difficult on you if you tip them off to your presence too soon. Deep raids are always a good idea when it comes to that point in the game! When the AI sees your raid, it will then start trying to reinforce, but if your raid fails or withdraws, the reinforcements won't keep ticking upward. This can make all the difference between success and a stalemate...
How Do AI Reinforcements Work At An Individual Planet?
Q: Is the AI gifted with a kind of reinforcement value, that it expends to place new ships or lost turrets? In this hypothesis, is a starship worth 1 reinforcement as any other ship, or you set also a kind of budget here (based on the worth in resources)?
A: It's a bit more complex than that, and isn't a real linear resource system. Instead, it's more of a random population system (like the zombies in Left 4 Dead, for instance), using procedural generation algorithms that you might liken to Perlin Noise or similar, but heavily layered.
What the AI does get is "reinforcement points" in general, which it can spend on reinforcing any given planets it deems as needing a reinforcement. When it does the reinforcement, then that planet receives various ships that depend on:
- The type of the AI (turtles get a huge number of reinforcements, for instance, but no offensive waves)
- How many guard posts are at that planet (the fewer, the lower the AI's ship cap for that planet)
- Any special reinforcement-affecting ships at that planet (AI Troop Accelerators, etc)
- A random factor
- The tech level of the planet and the tech level of the AI doing the reinforcement (it uses whichever is higher).
That algorithm then goes through and populates ships at the command station (if it exists), at the wormholes (if the command station exists), and at each guard post. If there are fewer guard posts, that will "squeeze" more reinforcements into the remaining guard posts and the command station, so you'll see those reinforce faster.
There are separate sub-algorithms for things like starships, turrets, and mines, and for the other main mobile military ships. So different types of AIs have a different chance to built turrets of an appropriate level during each reinforcement, or to add more starships or mines, etc. The AI does not care about "rebuilding" per se, as it does not keep track of past events. But rather, it looks at the current scenario and builds more based on the needs and caps at the current time.
In general, the goal of the reinforcement portion of the AI logic at a given planet is not to be intelligent, but rather is to provide variety. This is the core of the "scenario building" part of the AI, and having variances is much more important than the AI acting intelligently at that time. Once the ships are down, the AI's goal is then to act as intelligently as possible with the ships it has been given, following the rules of guards/special forces/planetary defenders/etc.
The AI in AI War is doing double-duty, in many senses: it is both your opponent, and it is also the map-creation/scenario-builder tool. In the case of which planets to reinforce, it is acting as your opponent. And in the case of how it then attacks you or defends against your attacks, the same as true. In the case of how a specific planet is reinforced once it is targeted for reinforcement, however, it is acting purely as a map-creator/scenario-builder.
