Recently, we played twice through the first mission. The first attempt featured too much doors (realized it only when we almost finished the mission), so we restarted the game and gave it another run. What keeps puzzling me is the AI behavior, though. Although the AI cards are pretty clear in stating what enemies should do, we had a lot of situations with either: a) several possible options or b) options that were plain stupid For example, we have two Cadmus located out of sight. Both have the remote Billy as well as Quiang within range of their hacking device. Both possible hacking targets feature the same amount of aggro (say 2). Additionally, there is Cadin, who has only 1 aggro, but if the Cadmus move they could attack him with their Combi Rifle. The Cadmus activate, and an AI card comes up stating: 1) Acquire target lock on target with lowest aggro 2) Attack target with lowest aggro 3) Attack target with lowest aggro 4) Attack target with lowest aggro and fall back Strictly following the rules, the Cadmus would move out of cover to shoot at Cadin once. Which would be an absolutely idiotic move, especially given the "Evil AI" statement in the rulebook which emphasizes that the AI always strives to achieve the result most detrimental to the players. Alternatively, we could forego Cadin, since - strictly speaking - both Cadmus already have Target lock with their hacking devices on both Billy and Quiang. With both possible targets now sporting the same amount of aggro, the rulebook states that enemies would prioritze Allies over Characters. Again resulting in a behavior that is far from an "Evil AI": Both enemies are now hacking an unimportant remote instead of the Heavy who is incredible hard to take out with traditional attack but highly vulnerable to hacking attacks. In the end, we played the situation like this: Both Cadmus were hacking Quiang twice, with their second action being "attack and fall back", as this resulted in the most efficient use of their tool set and in the most detrimental results for the players. It is, however, quite a stretch of the rules as written. How do you handle those contradictions between goal ("Evil AI", most detrimental result) and actual behavior arising from AI cards? Just stick to it and let the enemy do whatever stupid stuff they might do, or re-interpret the meaning of the AI card?
"Evil AI" is supposed to only be used in cases where it's ambiguous; I think that the rules for Hacking state that you can treat hacking like an attack, so they would have just hacked Billy twice (each), likely killing him and then choosing to target Qiang Gao. Enemies prioritize Allies, that's a choice the designers made. In in-universe terms I think it's reasonable for hackers to try to brick the Repeater that's close to them before they deal with something like Qiang Gao.
While I do agree with the reasoning based on "real" Infinity as to why a remote would be considered a priority threat and - in extension - even why a traditional ally like Agnes Ferreira or the Sygmaa is on the top of the hit list for the enemies, I just can't see why remotes would enjoy the same popularity among enemies in Infinity: Defiance. Ultimately, it makes live much easier for the players as the enemies do not exploit apparent weaknesses despite being perfectly able to do so, without any repercussions. Don't you think havign enemies that behave competently makes for a much more rewarding experience? Perhaps the ally preference was never meant to include remotes? Maybe just one more thing that got lost somewhere between development and printing...
I disagree that it makes things easier for the players, necessarily, because when you're trying to rescue ally x and get them back on the Stalker, Sheskiin going ham on them makes things hard.
No objection to that. I absolutely agree on the high priority on traditional allies, as stated. However, unless the role of remotes changes into something similarly crucial to mission success as Vaarah Zaandar (for example), I do think that remotes should not be treated with the same priority. Sheskiin coming in is actually quite a good example: If there are both Billy and Vaarah Zaandar heading for the Stalker, with Billy having zero aggro and Vaarah Zaandar having 1 aggro, Sheskiin would kill Billy. If both have the same amount and Billy was closer, she would (again) kill Billy. That's just plain stupid. Remove remotes from the list of priority targets.. Et voilà. The EI's hitwoman is actually doing her job. :)
Sheskiin preferentially targets whoever has the lowest aggro, though. That's kind of her gimmick. If Billie had 1 aggro and Vaarah had 0, Sheskiin would go after Vaarah.
I know. But don't you think that's a bit... gaming the game? If I know Sheskiin will show up, I'll just have my mech-shield remote ready with 0 aggro, so Sheskiin will go after him instead of the crucially important ally? Just doesn't feel right to me...
From my understanding the game is balanced around being able to game the AI. Quote from the 5 player rule: "Infinity Defiance is a game designed and tested to provide the best game experience for 4 players facing an Artificial Intelligence system. ... The Artificial Intelligence is designed in such a way the the Enemies, especially Regular units, will seek obvious targets: the closest, the one with the most Aggro etc. " If you want the the Enemies to make the "best" move you can always have a player control them instead but as stated this will result in a noticeable spike in difficulty. If you are playing like that you can also choose not reveal triggers and events early in a mission which again will increase difficulty.
At that point you're giving up the ability for Billie to do things like interact with consoles, which can be very crucial for the game. Moreover, the game does plenty of gamey things in terms of cranking alert or making things disappear after a certain number of turns, you have to be gamey right back. Keep in mind, as well, that the rules say that both the highest *and* lowest aggro are targets, so that's part of aggro management. Sheskiin's AI behavior is "prey on the weak first" and if you make Billie look weak she'll smash the shit out of him.
In our experience Billie has a very low life expectancy anyway. Given the original situation, working the instructions as written, i would expect Billie is out of the picture after two or three attacks anyway, so the next ones will be fed to Qiang Gao. Cadin wouldn't make the target list, since Hackers prefer to use their hacking attack as per the rules.
I'm kind of giving up on Billie to interact with consoles. All I need to do is put some aggro on my mission-relevant allies, too, and Billie is back. But I do agree: there are plenty of gamey things going on anyway, so I'll stick with rules-as-written for now (with respect to enemies targeting allies) and see how it develops over the course of the game. Actually, strictly following the rules Cadin would be their priority target. The Cadmus' preference for hacking things sets in once it attacks. In my initial example (Cadin 1 aggro and out of sight but target lock can be acquired with 1 move action; Qiang and Billie 2 aggro and within reach of the Cadmus' hacking device), strictly following the rules would result in both Cadmus attacking Cadin. As per the rules, enemies follow the AI instructions from top to bottom, acting on each they can fulfill until the acted on maximum 2 instructions. With my board setup, the Cadmus would move towards the target with the lowest aggro (first AI instruction) and shoot at it (second AI instruction). The whole problem with Billie vs Quiang doesn't even come up, if I strictly follow the rules. The Cadmus do their tactically rather questionable move with a bad final position, and that's it. What bothers me is the massive contradiction between actual enemy behavior and what seems to be the idea of enemy behavior, with "Evil AI" explicitly stating (just looked it up): "If multiple options arise when performing the Instructions of the AI cards or applying the Effects of Enemy Actions and Special Skills, the players must always choose the option that is more detrimental to the Characters." The big question here - to me - is: When does this statement "If multiple options arise" trigger? Let's construct another situation, even more ambigious. We have to Cadmus with LoS and within range to Cadin, Quiang, and Billie. Both characters as well as Billie have 1 aggro. AI instructions are still the same: Now, I could either read "If multiple options arise" as "triggers when multiple options arise from a single instruction" and work my way through the instructions, top to bottom. Both Cadmus attack Billie twice with their hacking device. Or, I could read it as "triggers when multiple options arise from the AI card's instructions in general". Given that the "Evil AI" rule uses the plural ("... instructions of the AI cards..."), the Cadmus might also hack Billie twice, but as their second instruction use instruction #4, attack and fall back. Making life much harder for the characters as they have to re-establish contact with the enemy. Pushing the example even further, let's assume the same situation, but now Quiang, Cadin and Billie have 3 aggro each. Uma, located on the other side of the large room behind a corner, has only 1 aggro, though, and neither Cadmus can establish target lock with a move action. Strictly following the AI card now would result in both Cadmus running into the midst of the enemy, then hacking Billie once. Or, again interpreting "if multiple options arise" in the spirit of "Evil AI", they attack twice and retreat to a out-of-sight position, if possible. That's really the main issue I find myself regularly stumbling over... There seems to be so much interpretation involved in working out what the AI really does. It can perform actions that no human opponent would ever really consider because they are so obviously bad, leaving you with a game that was never a challenge from the beginning. Or it can play intelligently, putting up quite a challenge and force you to think and cooperate...
@Ygethmir The main issue is that the tiebreakers don't need to be there if the "Evil AI" rule is used in the way you suggest.
They all have 1 aggro. First target is Billie due to Allies before Characters. Second target is Quiang due to Evil AI, as he can be hacked, and hacking is WAAAAAAAAY better.
And remember what Acquire Target means. Acquire target. If at least one Character is in Range and Line of Sight of one of the Attacks of the Enemy, they will skip this Instruction. Otherwise, the Enemy performs the Move Action, and spends as many Movement Points as needed to place themselves on a space from which they gain Line of Sight to the target, and that is as far away from the target as possible, while still being in Range. If, even with movement, acquiring a target is impossible, the Enemy will move in order to reduce their distance from the closest Character. In your first scenario, the Cadmus already has both Billie and Quiang in Range but not in LoS. So they are not allowed to skip the instruction. They are required to Move to get LoS on Cadin, who is the one with lowest Aggro, and then attack him. If they cannot gain LoS on Cadin, they ignore him and apply the Acquire target on Billie (Allies before Characters). If they cannot gain LoS on Cadin nor Billie, they ignore them and apply the Acquire target on Quingao. If they cannot gain LoS on anyone, they reduce distance from the closest Ally/Character.
Following the tie-breakers, Billie is the target, I agree. However, I do not get why Quiang would be the second target? The second attack is a new instruction, so firstly I would assume I have to start tie-breaking anew, meaning I'd be stuck with Billie again - because, again, it's allies before characters. Secondly, Billie is also hackable. And I can't find a rule that says every tie-breaker can only be used once - but I may have missed it? With respect to Acquire target: Does this mean that establishing Line of Sight takes precedence under all circumstances, even though an enemy has weapons available that explicitly do not require line of sight to attack? In other words: A weapon that does not need line of sight to attack needs line of sight to acquire a target? @Hecaton Well put. Yes, I guess that's actually my main issue. What's taking precedence: "Evil AI" or "rules, step by step, no matter what".
Right, I didn't consider this aspect. The Cadmus with Hacking attacks (No Lof) should ignore the Acquire Target instruction as they do not need LoS
Ok, thanks for the clarification. That means the "Evil AI" rule only triggers on a "per instruction" basis, and then only if the tie-breaker cannot resolve the multiple options arising from that instruction? In other words, enemies will always follow the rule sequence when working out AI behavior (top to bottom AI card, then tie-breaker per instruction if necessary, then Evil AI ruling if still unclear) even if the resulting behavior puts them at a significant disadvantage and may even significantly reduce their effectiveness.
Let's remove the No LoS part here, suppose a standard Nox. Instruction: Acquire Target to less Aggro. The poor Nox cannot see any Ally or Character, so he cannot skip the Instruction. So, who is the "less Aggro" on the table? Qiang and the Peripheral are unconscious, so they are ignored. The lowest one is Trisha who is 5 tiles away. As there is no way to gain LoS on her, she is skipped. Who is now the "less Aggro"? We have Cadin and Uma on the same value. How do you resolve it? This is what the rules say: When there are two or more possible targets in equal conditions, one will be selected following these criteria: 1. Select an Ally before a Character. 2. Select the Character with the highest Aggro. 3. Select a Character or Ally that hasn’t been activated yet. 4. Select the Character or Ally with the most Damage tokens. 5. Select the Character or Ally with the lowest Wound value. 6. Select a Character or Ally randomly. Step 1. There are no Ally, so Characters. Step 2. They have the same Aggro. Step 3. If one of them has not already been activated, that's the target. Otherwise go on... Step 4. Check Damage tokens and find your target. If it is the same go on... Step 5. Check Wound value (and remember Consequences here!!!) to find your target. If it is the same so on... Step 6. RANDOMPARTY! Now you have your target, let's suppose it is Cadin. Can the headspinning Nox gain LoS and Range to Cadin? If yes, proceed. It not, skip to Uma. Can the depressed Nox gain LoS and Range to Uma? If yes, proceed. If not, the rules of Acquire Target say: "the Enemy will move in order to reduce their distance from the closest Character". The now reinvigorated Nox move all ahead to the closest one (good luck to split again if they are at the same distance!) Edit: FFS, i wrote Qiang in a different way each time!
