spoofing the AI setups

[wfzimmerman]

Peskpesk is doing an amazing job with these AI setups, but I wonder if a player could exploit the if ... then logic of the setup choices. Forex, a lot of the setups check to see if there is a transportable amph or para within range of the target; if not, they defend something else. What's to stop a player from "tricking" the AI by setting up with the key unit one impulse away, then moving it into position on the next impulse?

[composer99]

Not much... aside from the fact that they do not get the benefit of that key unit during the surprise impulse, which is when the defending country is at its most vulnerable.

[Shannon V. OKeets]

ORIGINAL: wfzimmerman

Peskpesk is doing an amazing job with these AI setups, but I wonder if a player could exploit the if ... then logic of the setup choices. Forex, a lot of the setups check to see if there is a transportable amph or para within range of the target; if not, they defend something else. What's to stop a player from "tricking" the AI by setting up with the key unit one impulse away, then moving it into position on the next impulse?

I don't see this as any different than when a human is playing. If the enemy has a threat, then you need to defend against it, whether the enemy follows through on the threat or not.

I think it was Alekhine (Russian chess grandmaster) who said (paraphrasing): "the threat is more powerful than the act".

[peskpesk]

ORIGINAL: wfzimmerman

Peskpesk is doing an amazing job with these AI setups, but I wonder if a player could exploit the if ... then logic of the setup choices. Forex, a lot of the setups check to see if there is a transportable amph or para within range of the target; if not, they defend something else. What's to stop a player from "tricking" the AI by setting up with the key unit one impulse away, then moving it into position on the next impulse?

If something is not detected by the minor country AI setup script, for sure a smart player can exploit it. The risk is reduced by the fact that the AI have always the option to choose the basic setup – Normally capital defence – instead of an extreme setup – like invasion defence.

So a player might think by threatening an invasion with for example a division and thus forcing the AI script to setup according to the invasion defence since the player has carefully examined the script and concluded the AI only thinks land units in neighbouring country are a threat and not those “hidden” at the border in the next neighbouring country. But since some degree of randomness will be included, the player can’t be sure that his feint will succeed. The AI might pick another defence strategy.
Also I try to have at least two variants for all setups so a player can’t exactly foresee where the AI will put units in a specific setup. The plan is for a MinMax evaluation to help to conclude on which setup is best used against the current enemy setup.

Any suggestion on how to improve the AI scripts are more than welcome.

[brian brian]

I know one thing I learned from the Finland set-ups...as the Russians I might just put a PARA in range of Helsinki and watch the Finns play turtle for a turn or longer, just because it is in range, never minding that the Finns would have FTR cover and the Russians probably wouldn't, and many other reasons why actually dropping in Helsinki is pretty unlikely to actually work. Some of the threat detection routine should look at the % chance of the threat actually working. And also, countries set up after alignment are probably joining a side on the offensive at the time, which should create priorities in set-up decisions over how to play defense.

[Shannon V. OKeets]

ORIGINAL: brian brian

I know one thing I learned from the Finland set-ups...as the Russians I might just put a PARA in range of Helsinki and watch the Finns play turtle for a turn or longer, just because it is in range, never minding that the Finns would have FTR cover and the Russians probably wouldn't, and many other reasons why actually dropping in Helsinki is pretty unlikely to actually work. Some of the threat detection routine should look at the % chance of the threat actually working. And also, countries set up after alignment are probably joining a side on the offensive at the time, which should create priorities in set-up decisions over how to play defense.

If the Finns are surprised, they won't have fighter cover.

[paulderynck]

ORIGINAL: wfzimmerman

Peskpesk is doing an amazing job with these AI setups, but I wonder if a player could exploit the if ... then logic of the setup choices. Forex, a lot of the setups check to see if there is a transportable amph or para within range of the target; if not, they defend something else. What's to stop a player from "tricking" the AI by setting up with the key unit one impulse away, then moving it into position on the next impulse?

I mentioned something like this in another of the AI threads. Fact is though I could have the Para on the other side of the world and get him in range to drop with the right chaining of transports, land movement and then maybe an air transport mission.

The AI check for paradrops should be:
1. Does the enemy have a paratroop unit anywhere?
2. If yes, does he have an ATR in range of the target you are worried about?
3. If yes, then the threat exists (although you could also add in analyzing likely action limits as long as the enemy doesn't have an Offensice Chit in the bank - but must take into account co-operative action limits among units that can co-operate).

[brian brian]

But when Germany aligns Finland, they are not surprised....so the set-up routine needs to keep that in mind. And just because a paradrop is possible, you have to consider how likely it is. A single PARA corps against a city is not very likely to succeed...is there also enough ground support factors in range to double the strength? Even if so, might it be better to dare the enemy to drop, then walk back and crush their lonely PARA?

And as Paul mentions, enemy action limits should be considered. Going back to Finland, one of the checks is for a threat of a DIV invasion. In 1940, this is important. In 1941 at the beginning of Barbarossa, this is irrelevant as the USSR won't be able to take a combined impulse.

[peskpesk]

ORIGINAL: brian brian

But when Germany aligns Finland, they are not surprised....so the set-up routine needs to keep that in mind. And just because a paradrop is possible, you have to consider how likely it is. A single PARA corps against a city is not very likely to succeed...is there also enough ground support factors in range to double the strength? Even if so, might it be better to dare the enemy to drop, then walk back and crush their lonely PARA?

And as Paul mentions, enemy action limits should be considered. Going back to Finland, one of the checks is for a threat of a DIV invasion. In 1940, this is important. In 1941 at the beginning of Barbarossa, this is irrelevant as the USSR won't be able to take a combined impulse.

I plan revises the AI for Finland, since it was the first minor setup and is not in the same format as the other. Many setups there are not shown with an image, also they lake offensive setups due to the scope of the setups has changed slightly since the start. And true it might be a big difference if country is aligned or declared war on, when it comes to choosing setup.

[micheljq]

ORIGINAL: brian brian

But when Germany aligns Finland, they are not surprised....so the set-up routine needs to keep that in mind. And just because a paradrop is possible, you have to consider how likely it is. A single PARA corps against a city is not very likely to succeed...is there also enough ground support factors in range to double the strength? Even if so, might it be better to dare the enemy to drop, then walk back and crush their lonely PARA?

And as Paul mentions, enemy action limits should be considered. Going back to Finland, one of the checks is for a threat of a DIV invasion. In 1940, this is important. In 1941 at the beginning of Barbarossa, this is irrelevant as the USSR won't be able to take a combined impulse.

As USSR, I would risk a combined impulse even in 1941, if I have a chance to put Finland out of war.

I think Finland has a special status as a minor country and refining, or double checking, the AI for Finland is worth it.
The armies of Finland are often active the whole campaign because Finland is not easy to conquer, as such this minor country can have a good impact on a campaign.

I would not want to see Finland fall early just because the AI would not think to protect Helsinki with a unit at least I mean.

Only my personal newbie opinion of course.

[micheljq]

ORIGINAL: wfzimmerman

Peskpesk is doing an amazing job with these AI setups, but I wonder if a player could exploit the if ... then logic of the setup choices. Forex, a lot of the setups check to see if there is a transportable amph or para within range of the target; if not, they defend something else. What's to stop a player from "tricking" the AI by setting up with the key unit one impulse away, then moving it into position on the next impulse?

Maybe the AI of the minor country should always leave at least a corps or division in its capital.

[Orm]

ORIGINAL: micheljq

Maybe the AI of the minor country should always leave at least a corps or division in its capital.

In many cases that would severly handicap the defence of the minor. A good AI should be able to leave the capital undefended. And from what I have read so far on the forum about the planned AI it looks like it will be a good one.

If the AI in some cases leaves the capital empty and the player can take advantage of it is in itself not completely bad. As long as it does not happen all the time (with the same conditions). I've made errors when I set up minor countries myself. And if the AI sets up the minors as good as a human player with all the errors a good human player does then I am very happy with the AI.

I would be more troubled with a to cautius AI that always uses a capital defence because then it would be to easy to plan for it and to win.

I want a good AI and an AI that always leaves a corp on a minors capital makes a decent AI and not a good one. If that lets the AI to make an occasional mistake then so be it.

-Orm