Cheating AI?

[Charles22]

Yesterday while screwing around and developing a core force for a possible future campaign (BTW, Norwegians in generated campaign, make for an awful force in 9/39 - NO TANKS!), I went into deployment with human deploy, but didn't move a unit in what was a frozen woody environment. The Ruskies were the opponent. Before my first turn, the preliminary opening possible bombardment phase, roughly about half of the shells fell on the very back line of the screen, where the forces were, and they had absolutely no way of knowing where they were.

To add to this frustration, is the fact that I never see in any game where I've actually bothered to move them from the back, where the AI will bombard back there. On the other hand, I remember how 88's used to be bombarded without their being seen or having fired a round, but that doesn't appear to happen anymore. That fact seems to speak against AI cheating, so then I ask, is the AI cheating at random; sometimes it does, and sometimes it doesn't?

On another note, I would like to say that though there are some wild units in this game, and by wild I mean things like the Maus, the Polish 75mm AA, and the Russian 85mm AA. Note, that when the computer picks such rare units en masse, it almost isn't WWII battle anymore. I have yet to face the Russians, but if force selection for the AI isn't hampered, as is the case with the Poles picking bewtween 6-8 flak guns, then the Russians will have 6-8 85mm flaks against German advances, and how bizzare is that? In order to deal with this problem, the human Gerry opponent has to pick a far disproportionate amount of strange weapons to deal with the problem. One idea, for example, would be to copy the opponent, and indeed, outpurchase him in flak guns. This makes for the ludicrous situation of both sides being able to destroy each others tanks with ease, be that in '39 with the Poles or '41 with the Ruskies.

I think the AI may need a toggle to keep it picking somewhat historical forces for each battle, or have the option to go wild as it is doing, if one feels like fighting battles in another dimension.

BTW, to show the strangeness of some of the forces picked, and yes I know that the computer would like to use all of it's points, so naturally some anomalies might occur, but look at this. My first battle on the western front, wasn't one. It was a fight against the Belgians in 5/40 and they got no points, as it appeared they surrendered before the first round was fired, which gave me desert terrain (I'm talking WWII campaign here).

The next battle was 6/40 which only left me with the choice of North Africa or Eastern Europe. I picked North Africa and fought the French at Reims. Huh? Anyway, this shows radical computer picking, though I think one could make the case for this force perhaps being reasonable, far more than the computer picking large amounts of heavy flak guns. Anyway, I scored like 3902 points, the most in that campaign, while only losing like 37. Strange, but the Frenchies had 170 tanks (I counted them afterwards), of which approximately 150 of them were the dreaded S-35's!!! I perhaps destroyed like 40 of their tanks, before they conceded (they were assaulting).

[Larry Holt]

Originally posted by Charles22:
Before my first turn, the preliminary opening possible bombardment phase, roughly about half of the shells fell on the very back line of the screen, where the forces were, and they had absolutely no way of knowing where they were.

Hmmm, In SP1 I always saw and hated the AI cheat. I've never seen the AI cheat and fire at unspotted units in SPWAW. Good fix Matrix.

On the other hand, I remember how 88's used to be bombarded without their being seen or having fired a round, but that doesn't appear to happen anymore.

Lordy, how I hated that too!

On another note, I would like to say that though there are some wild units in this game, and by wild I mean things like the Maus, the Polish 75mm AA, and the Russian 85mm AA. Note, that when the computer picks such rare units en masse, it almost isn't WWII battle anymore....
Strange, but the Frenchies had 170 tanks (I counted them afterwards), of which approximately 150 of them were the dreaded S-35's!!! I perhaps destroyed like 40 of their tanks, before they conceded (they were assaulting).

I fought over 100 S-35's in a delay. It shows what the French could have done had they concentrated. This is unrealistic (& since it was I did not feel badly about setting the rally ratings real low for the rest of the battle). In SP1, once a certain number of aircraft were purchased, they disappeared from the pick list. Perhaps this could be reused to limit the AI to purchasing realistic numbers of heavy-duty, but rare units.

The next battle was 6/40 which only left me with the choice of North Africa or Eastern Europe. I picked North Africa and fought the French at Reims.

I found that while the theater selection button is there, it does not actually work until there is a choice at that time in the war.




------------------
An old soldier but not yet a faded one.
OK, maybe just a bit faded.

[Hauptmann6]

I have had the same problems only with the germans in the desert. I was playing as the US at kasserine pass(WWII camp) and the germans were spotting my units on the first turn and getting 40% shots against my DUG IN units after moving full at 30 hexes...

[troopie]

Then there is the cheat found in ALL SP games. AI's arty can always find my troop lorries, without FO or line of sight from any unit. I've sometimes lost half of my lorries disabled from AI arty fire in one turn. Haven't used lorry infantry in SPWAW yet but suspect it's there.

troopie

[nikb]

Troopie

I suspect they are still here. I also wonder about cheating with the placement of mines and the avoidance of mines by the AI.

Nik

[Wild Bill]

The argument that the AI "cheats" has gone back and forth ever since the game came out.

Some "in the know" swear this is not the case. Other affirm that there is. I think the latter is true.

I remember seeing somewhere once a list of numbers related to the AI cheat potential. I forget how they applied themselves to the game or how they were to be used.

On the other hand, the AI needs a little help occasionally. After all, we have all complained about how dumb it is, so why not allow it a slight handicap

My argument falls on deaf ears as the reader who has just been pummeled by deadly AI artillery utters a "Yeah, right." and quickly closes this post.

Wild Bill

------------------
Wild Bill Wilder
Coordinator, Scenario Design
Matrix Games

[Mike Wood]

Hello...

I really hate computer opponents cheating! In SPWAW, the computer opponent does not cheat at any time.

Originally posted by nikb:
Troopie
I suspect they are still here. I also wonder about cheating with the placement of mines and the avoidance of mines by the AI.
Nik

For artillery, the computer opponent goes through a loop 8 times during the turn. In each loop units move and fire and assign artillery. Every unit on the map tries to spot every enemy unit on the map. A list of spotted enemy units and who spotted them is recorded. The program then looks at who has the best artillery values and what targets are the best, such as 88mm AAG, and then assigns the targets. It assigns targets about the same as a human player, although it is very precise in the way it approaches the subject. It has the same delays and uses the same fire code that humans use. For pre-game bombardment, the locations are chosen randomly, with proximity to objectives being a factor. No cheating.

The computer places the mines where it calculates the humans will travel, based on terrain and proximity to objectives. Computer controlled units have the same chance of hitting a mine when it enters a hex with mines as a human controlled unit. No cheating.

There were some cheats in the legacy code, but I removed them.

Thanks for your interest...

Michael Wood,
Matrix Games


[This message has been edited by Mike Wood (edited 06-01-2000).]

[Charles22]

Mike: You may not be able to find a cheat, so I don't have a problem with that, but I see no other reason, other than random cheating, to explain why the computer would target half of it's artillery at a non-essential area (the very rear deployment line). Surely, there must have been something to tell it that my units had not deployed, for as I say, It's NEVER done that before. I would know, for I would have been laughing my head off if it had.

This is the only verifiable case I know of to prove AI cheating, because, as I say it seems to operate fairly in othe regards, Actually, now that i thingk about it, this is a GOOD THING. Why? Because the only instance I know of AI cheating involves punishing someone for deploying like an idiot (I wasn't trying to deploy in that game). I was hoping to make the point, that if noone thought the AI was cheating I at least have one great example, but hey, if that's the only case, then that's a GOOD CHEAT. I wonder if leaving your troops on the rear line, after non-deployment, will always result in the AI seeing them?

[This message has been edited by Charles22 (edited 06-02-2000).]

[Wild Bill]

In defense of Michael, he is one of the best programmers around. Time and time again we have argued over bugs, anomalies, problems.

We really get down to it

Invariably, Michael is right. I have come to accept his word as final. If he says he hates cheats and he removed them, I believe him.

And, maybe its me, but I have not seen the unforgiveable accuracy of computer based artillery that I have seen in other SP games.

I have seen the computer drop a barrage on its own units, or fall behind my rapidly advancing units more than once.

And I have played dozens of scenarios in this game time and time again.

I understand your frustration, Charles. It is that of any battlefield commander, but I'll have to go with Michael on this one.

Wild Bill

------------------
Wild Bill Wilder
Coordinator, Scenario Design
Matrix Games

[Charles22]

Well, like I said, I actually consider what it did good, but when it's tentacles reach that far, in a situation where it's knowing where my troops were, couldn't be more remote, it definitely makes me wonder. I might try to recreate this situation this weekend just for curiousity's sake, to see if it always sees an idiotic non-deployed deployment. If that's the only super-snoop it has, I definitely don't have a problem with it, but it would seem to indicate it has that capability.

Some of the other AI cheats people think they have seen with this game I haven't seen AT ALL, such as AI sniffing out trucks. My natural response to such a charge, would be to expect that the player hasn't realized that trucks are pretty easy to spot, moving or not, but if they had them behind hills and two to three hexes away from their guns, they wouldn't have that problem. It does make me wonder though, because if I swallowed my own advice and applied it in it's most radical sense, that wouldn't have saved me in the AI possibly sniffing out my non-deployed deployment discussed earlier (although technically I'm not sure whether it was trucks hit or not).

[kfbaker]

Usualy speeking the best so called AI do cheat, the trick is making it look as if they don't.

Secondly I would suspect the augment for enermy opening barrage would be based on rougth intelegence gather before. In order to effect this you would basicly take the units actual possition and then randomize around it to create a relistic effect while actual cheating to some extant. Of course this works fine with any normal deployment unless some one placers all his units in a line as you had, and as such the must have a fair percentage landing on the units, maybe not at the one centered on, but since the units are all contrated in the same area the result will look as such.

Just A whiled stab, but from past experence of programming so called AI I wouldn't be suprissed as I used simlair cheats myself years ago when programming war games on the old ZX Spectrum. Saying that we only had 40k to play with, (8k was screen if your wondering).

[Charles22]

Let's say there were 16 hexes hit in total; perhaps there were more. In any event, 8 of them were upon that thin line of troops at the back (I might add, though I know nothing about AI, that it has been admitted that the generated campaign is a bit unstable [Was Mike Woods the one who said this?]), and, afterall, my instance of seeming AI cheating was from the generated campaign. Perhaps AI is different for WWII campaign/generated campaign/scenario? The other relative 8 hexes that were hit, were somewhat behind a fairly non-consequential hill. From all of my WWII campaign (in SPWAW) playing I've seen no instance of anything resembling AI cheating, but then I didn't go do any non-deployed deployments as some great strategy either.

[Windo von Paene]

Well I have noticed that if I haven't specified a target for artillary, and then click on the "Show Target Hex" it will take me to the back row of the enemy's side of the map. Perhaps that's what is happening here, as the AI can't spot any units, so it leaves it at the default. Now why it would then attack that hex anyway, I can't fathom...

[O de B]

Hi
i heard that in the Old SP1, there was a sort of 'trigger' in the Radio value of the unit, that would be its 'rarity'. I never fully understood the process, but basically if the radio chance of the unit was written '52' in the OOB, then the unit whould have a 50% chance of having a radio, and would be very rare (for purchase by the AI). Is this sill applying in SPWaW ? (And if yet, i'd like some explanations on how it works, i never managed to have it work in SP).
Thanks.

[Charles22]

If you run the oobedit program you will see that every unit has a radio percential rating. I tired another generated Norwegian campaign against the Russians, with a non-deployed deployment and I didn't get any bombardment on those rear hexes. Only problem is, I didn't get any bombardment at all. They mustn't have picked any, or in nay case would plot it post-opening phase, but I didn't stick around to find out.

[Tombstone]

I bet that if you dont move a unit from it's non-depoloyed hex then the computer actually still has line of sight on it. You know how if you dont move and you had been spotted in the old sp3 that you'd stay spotted? Maybe there's something wierd about the initial deployment that makes them visible to the computer... and to artillery attack on the first turn. Who knows. Anyways, there's some total speculation from someone who has no idea what's really happening there.

Tomo