AI, hidden and nearest

What happens when the AI model activates, needs to go to the nearest model and the nearest one is hidden and with LoS blocked?
AI should test INT to detect their enemy when they are going to engage or attack, and they need LoS to perform the test, so it would need to move to get LoS, but they could not try to detect if they are not going to engage or attack the hidden target in their 2nd action, is that right? (not the case in the image because AI has ranged attack)
Should they go to the next nearest model? Or do they take into account a model like the one in the image as the nearest?

Hey @DK-dark

They’d behave just like normal models when detecting Hidden models. If we check ‘Detecting Hidden Models’ on page 27, only Hidden models within Clear or Obscured Line of Sight can be Detected.

If the nearest model cannot be Detected due to Line of Sight issues, the Adversary would instead turn its attentions to the next nearest enemy model that meets its Target Priority.

so the adversary would not take into account its enemy if the enemy has blocked LoS, even if the response is M, so moving to the nearest (wich does not need to detect any model to move) would move the adversary to the next nearest one, not the blocked LoS one, right?

To my understanding, yes; the adversary would ignore that hidden model because you’d determine the ‘target’ of the adversary action prior to the Move.

So in that instance, because the hidden model isn’t within clear or obscured line of sight, the adversary wouldn’t be able to try and detect it so it effectively doesn’t register that the hidden model is there at all, therefore it can’t ‘target’ it. As a result it would move on to the next potential target according to its priority matrix (usually the next nearest).

You’re both right, aye.

Ok, thanks.
So, in the same case, if my model is with Obscured LoS, when the adversary would take the INT test? If the adversary is just moving with their first action, and the test is performed when they are going to engage or attack, when is the adversary take into account my model? I mean, should they roll to detect my model after rolling the response? Even if the response is M? How would be possible to roll the INT test if they are not announcing an engagement or attack? I don’t find anything about that in the ‘detecting models’ paragraph or in the adversary section.
If the adversary could attack, they would roll the INT test, and then they would roll the attack, or change the action and the target, but for moving without engaging… how would they take into account my model (being hidden) to perform a first movement (without engaging) and why?

If the model is obscured (and therefore could potentially be detected) then that test to detect should be after the Response type is determined as target priority is checked after the response is rolled for (pg.59 of the core rulebook for reference). There is still a ‘target’ for an adversary response, even if it isn’t an attack; it doesn’t matter if they’re ‘just’ moving, they still have a ‘target’ to determine what they’re moving towards/away from etc. So to break down that instance:

  1. Adversary activates and rolls to determine it’s Response (which is, let’s say a Move on an aggressive model with melee as its preferred attack for this example).
  2. Target priority in this example is for the Nearest enemy (which is the Obscured Hidden model). Because the model is in obscured line of sight, there’s a chance they can be detected so the Adversary rolls against its INT (with the bonuses if they’re within 6" etc.)
  3. If successful; the adversary has detected the hidden model and would therefore move towards it and potentially engage/ranged attack/whatever according to the Move response.
  4. If the INT test is failed then they cannot target the hidden model and so their Target Priority is switched to the next Nearest enemy.

yes, it is logical, thanks. Anyway, I think this should appear in next review, in the ‘detecting models’ paragraph, because it does not say anything about rolling if the hidden target is being target without any attack

I agree that it could do with a bit more of a definitive statement regarding when the detection check would occur yes. It is sort of there in the rulebook as written but it takes a fair bit of ‘working out’ to get to the answer when a very simple statement could be much clearer.

There not being an attack is irrelevant though. A ‘target’, in the context of adversary Responses, doesn’t just mean the target of an attack; it’s a target of a Response in general.

yeah, right, but the thing is in the next paragraph, it says ‘when the detection test is failed, the enemy model can choose a different target for that Action’, or change the engagement/movement, so…:thinking::sweat_smile:

…which is in line with everything else? I’m not sure why that would be a contradiction.

If the original Hidden target of the Response can potentially be detected (i.e. if it is obscured or in clear line of site) then you make the detection roll; if it’s successful then that model is the target, if it’s failed then you move to the next eligible target for the response (as per page 69. when discussing how Adversaries operate slightly differently when detecting hidden models).

The distinction in that passage you refer to is that a player controlled model can indeed choose a different target or carry on it’s movement and engage a different model etc. Adversaries inherently don’t have that ‘choice’ to make as their targets are selected via their target priority list and how they go about this is reasonably explicit in their rules (the pg. 69 section) so I’m not sure how that particular bit is problematic?

because they need to be the target of an action to be detected (I understand that when I read it), not a target of a response or choosing them as objective
for a possible next action; being the target of a response (from the target priority list) is different from being the target of an action from a model, thats the difference wich confuses me

The specific wording is that they need to be detected in order to be the target of an action in the first place.

Hey @DK-dark

This comment from @JimmyW from is correct. Their step by step process is pretty spot on.

I believe that Detecting Hidden Models section on page 27 does cover both targeting a model prior to Detecting and moving into base to base -

" … enemy models may not target it, nor may they move into contact with it, unless they first roll to Detect it …"

Off the top of my head, I think targeting a model and moving into base contact with it are the only two ways two models can interact with each other but if you can think of any other situations not covered by this, let me know!

1 Like

yes, my confusion was the fact of not treating the target priority list as a fact of making target on a model, I thought targeting from a response was not eligible to roll for that because of the source (the response instead of any action), I thought the source for targeting needed to be an action

yes, my confusion was the fact of not treating the target priority list as a fact of making target on a model, I thought targeting from a response was not eligible to roll for that because of the source (the response instead of any action), I thought the source of this ‘targeting’ needed to be an action.
Also, the other paragraph about failing the test confused me, because it says you can change the target for the action (it seems, under my understanding, like you need to announce, for example, a ranged attack to try to detect the hidden model, so you need to spend an action on the hidden enemy to test INT, and, if you don’t detect him, you can change your target, and if there are no target available, change the action).
So could you try to detect the hidden model from 34", even if you don’t have an action to spend on the enemy, only LoS? Just for removing the hidden token?

Hey @DK-dark

A Response contains Actions though. You are not performing the Response to attack an enemy model, you are performing the Actions containing within it. Therefore, the normal rolls for Detecting apply.

You would declare your intent for an Action and then do the Detect. If you fail, you would get to pick another valid target for the Action. If there is no valid Target after the failed Detect, you would get to change the Action - this is on page 27.

You do need a valid Action in order to generate a Detect though, which requires a legal target. If you don’t have something that interacts with a model 34 inches away, you cannot Detect them.

I can see why the current version could be misleading as the text on Pg.69 does suggest that the detection roll being made is a result of an attack specifically (so yes @DK-dark I think I know what the issue is now and agree that, if this is the case then the reference to an ‘attack’ should probably be altered).

Unless this is entirely intentional and the Adversary would indeed activate with the Hidden model as the ‘target’ of the Response but only then try and detect them if that response causes them to attempt an attack of some kind (and instead attacking someone else if they fail the detection test and there’s another viable target).

Personally I think that would complicate the situation too much; it works ok if the adversary is attacking at range but if the whole purpose of their response is to engage in melee then it creates a catch 22 as they can’t engage without detecting but they wouldn’t attempt the detection until after they’ve moved… at which point it’s too late to then engage and they’d presumably just end their activation/block (which is a waste). I’d say it’s much cleaner for it to operate as @Modiphius-Dom has clarified and the detection attempt is made (if able) at the point of determining the original target priority for the response.

Yes, I understood that, so I needed to engage or attack, but that would make the engagement harder for a melee model…
but being the target of the AI for doing their actions due the response and their target priority list would not be enough to roll, because, as @Modiphius-Dom says, you need a valid Action to interact with the hidden model, so using a Move Action from point A to point B, how could be possible to roll because of that to detect a hidden model? In that case, only an adversary could do that, not a player’s model (they don’t have a target priority list and cannot interact with another model in a normal movement, when they not engage), and that does not seem fair.
According to @Modiphius-Dom , a melee adversary could not roll for detect a model (+9", so they cannot engage even sprinting), because they would not have any actions in their response to interact with the hidden model and have a chance for rolling, so they would need to look for the 2nd nearest model

well, the melee adversary could roll for detect the hidden model if, within their movement, would reach him to engage, but not at 34" just for being the nearest one…