New Arty Routine

[Marek_Tucan]

Haven't played this game as much as some other members here but still I do consider myself experienced, though lacking a bit in the infantry department
The AI back in old SP2 times WAS Borg-ish a bit, but it was compensated by being not-so-bright. It had to be Borgish to be able to compete atleast with inexperienced PI (Player Intelligence).
Andy and Don made a great effort to bring the AI as close to PI as possible. While doing so they've removed Borginess as it's not needed anymore now the AI can react with its arty fire plan very close to PI.
And the AI dispersion is working right the same way as PI dispersion. Enjoyed myself in one game how AI was lobbing salvo after salvo into thin air where my tanks have once been, Loathed the AI when it started shelling hill behind which my TD's were lurking. But never saw Borgish behavior in SPMBT/SPWW2. Just an intelligent, some times almost human behavior.

[Mobhack]

You were warned about breaking board rule number 3 by your referring to DRG's "lame claims". You were not banned, that is a formal process which further behaviour of that sort may require.

Please read and digest the Shrapnel message board rules, especially No. 3, and if you are prepared to abide by them in future then we have no problem. You can find these by clicking on the link marked "board rules" at the top of the page.

Andy

[Pyros]

Hi BigJim,

In order to prove you that there isn't anything wrong with the artillery I have initiated a series of tests concerning the artillery routines.

I think that by presenting you the data of one of those tests you will be convinced that the game code works the same way for both the human and the AI.

I am waiting for your comments concerning the following test scenario:


TEST TWO

Second set of test (use of HQ for calling arty missions without LOS)
i) number of tests for AI is 4 samples
ii) number of test for human is 4 samples


The second set of tests is also based on the simple method of mirrored settings with two testings modes:

A. AI vs AI
B. Human vs Human


For the second test (without LOS) I preplanned several arty missions by the use of a group of mortars (stacked in a single hex). As a consequence of this action the rest of the available enemy AI artillery (a second group of mortars) took the opportunity to design several arty missions directly on top of the previous point of interest (mortar concentration with preplanned arty mission).

The results from the AI vs AI mode are extracted by interrupting the game sequence by hitting the "ENTER" and going into human control (thus investigating the assigned by AI artillery missions and the scattering of the mean point of impacts).

For the second series of tests (without direct LOS) the results proved that the Artillery performs equally for both the human and the AI modes (in other words the AI is using the same code for determining the artillery accuracy over the mean point of impact).

Here are the screenshots of the test:

1. General view of the setting
2. Results of AI arty missions
3. Results of Human arty missions


1. General view of the setting

The "pre-planned Arty group" (first group of mortars) will start a sequence of pre-designed artillery missions aiming on a fixed on-map hex (somewhere to the middle of the map) in order to create the appropriate conditions (smoke and activity) and this way to attract (trigger) the other side AI group of "Active mortars" to design several arty missions directly on top of the "pre-planned Arty group".




2. Results of AI arty missions




3. Results of Human arty missions

[Uncle_Joe]

Amusingly, I've often found myself scratching my head as the AI randomly shells areas that are nowhere NEAR my troops.

And that seems to happen far, far more often than the AI pinpointing one of my troop concentrations that are clearly out of LOS. Like narwan, I have played this game a LOT too, and I have never seen the problem that BigJim is stating. Has the AI gotten lucky from time to time and blasted my men out of LOS? Sure, but more often than not, I see the opposite as well.

[BigJim]

Ok pyros I may have found my problem?? I have cut the map down to 60 by 60 and with all the equipment I have now it is almost impossible to position my stuff within the area alotted me at setup (therefore the AI CANNOT miss no matter where it fires). In your test you have only setup opposing arty so I can see what you mean about it being random based on that setup.

Andy as for my remarks regarding Don, I felt his remarks were to say the least abit caustic and responded in kind.

I have played these games since day 1 when SSI first released them as Gary Grigsby programmed them and I recall a large hugh and cry back then about atry effectiveness.

First off this game is primarily a surpression game, ie if your troops are surpressed they cannot see, move, or fire with any accuracy. To me the 81mm motar is just waaay to effective vs armour in this representation. I think I would have to look very hard to find any real life events where tanks were disabled by motar fire (I am sure there must be some but I feel it must be rare, since I am a baby boomer and so grew up with tons of WWII Vets around me my own father having fought at Iwo Jima).

Now the root of my problems to a large degree are that I play for ONLY a decisive victory (this maybe a flaw in my chracter) so when odds like 95% to hit are given and the result is 6 misses or no effect in a row it tends to raise my ire.

That said I still enjoy and play the game daily as I am retired so have some time to mess around and I don't wish to minimize the effort you guys have put in to bring back this old classic.

Jim

[Mobhack]

have played these games since day 1 when SSI first released them as Gary Grigsby programmed them and I recall a large hugh and cry back then about atry effectiveness.

As we found when we had the code in our hands - the AI was getting large "hints" as to enemy location, as well as the ability to "teleport" just-arriving fires across the map AFAIR - but we removed such code back in 2001 or whenever the DOS version of SPWW2 came out. However - the SSI team would have had only the usual 1 year to market game development cycle, and we have been refining WW2/MBT for much longer than that. So I had the time to do such a specialist AI routine (and tune it too), whereas they most likely did not.

However there seems to be a "folk memory" of borg-AI arty in the SP community, despite the repeated statement of fact by the programmers of the Camo verion that we have removed the original SSI code borg, and instituted a human-style decision matrix on the AI. This "folk memory" may still apply to non-Camo versions of the code, but frankly, we are tired of it repeatedly being ressurected as applying to ourproduct .

First off this game is primarily a surpression game, ie if your troops are surpressed they cannot see, move, or fire with any accuracy. To me the 81mm motar is just waaay to effective vs armour in this representation. I think I would have to look very hard to find any real life events where tanks were disabled by motar fire (I am sure there must be some but I feel it must be rare, since I am a baby boomer and so grew up with tons of WWII Vets around me my own father having fought at Iwo Jima).

In the original SSI code, if HE pen was equal to the armour - it went though. But as stated somewhere in the release notes - now HE pen needs to have signifigant overpenetration to have a chance of penetrating, and WH size of the HE round is assessed. So little 81mm mortars are no longer very effective armour killers. In SP III the 81mm mortar was the primary human M1A1 killer of choice (cost/benefit analysis) because in SP III with its multi-vehicle platoon unit lumps, it could not differentiate between killed and mobility killed (disabled vehicles were not separated off the lump as individuals) - so an 81mm track hit removed (=killed)the tank. (I have no idea whether or not this behaviour was modified for the SPWAW code - based on the SP III code).

And the first Tiger II met by the Allies was killed by a British 2 inch mortar - opinion varies as to wheter it dropped down the turret hatch, or blew up the ammo truck parked beside it but the result was the same - one cooked tigger.

Now the root of my problems to a large degree are that I play for ONLY a decisive victory (this maybe a flaw in my chracter) so when odds like 95% to hit are given and the result is 6 misses or no effect in a row it tends to raise my ire.

And we have already discussed the random number generator in SP, and done dumps of 1000s of numbers, published (whether here or on the old Yahoo board) which proved entirely normal distribution of the random numbers, and no unusual chains of numbers in the output stream.

Sometimes you have a "bad hair" day and miss the barn door several times, and others you get the opposite effect and get a few low percentage pops on target. Such is life. Such are random number generators (which are not "random" but pseudo-random, if you want to be pedantic .

Cheers
Andy

[Charles22]

It's a real shame that he over-reacts to when the AI does hit his unseen troops or that he thinks things were as the old SP was, because this game is much better than either.

One of the things I like to do is get an ammunition truck and park a couple of 75IG's and 150IG's beside it and fire away. I pick IG guns because I don't have to get them transport because they can move away if needed. It's none too infrequent that I've had the shelling go for 10 straight turns before getting any return bombardment. I guess the AI just thinks there's more important things to bombard, particularly the seen things, because there's certainly an unavoidable amount of smoke in a close area this way.

[Charles22]

BigJim: I see now that you have figured out your problem, and the remedy is all too apparent, but I would like to add just off the cuff that playing a map with the height of 200, adds a whole new dimension to the game that you may not have thought possible (I wouldn't recommmned going with the maximum width, however, unless you like a lot of turns of moving forward, which of course can be got around by adding extra turns to the game anyway). Imagine, you park an 88 in the middle of the map, and it cannot reach to either end. That's what height does, it allows for more successful flanking and more individual group strategies. Those fire brigades take a lot longer to get to the attacked area this way too. I like it so much I can't imagine playing on a narrower map. It makes the game start to approach the operational to some extent and therefore adds more strategy than possible before. The fact that we can adjust the visibility on the fly just makes this even more enjoyable. I generally play those sort of maps with between 110-130 units, and to say I'm vulnerable for at least a good portion of each game in no understatement. The use of every unit is very critical in this especially when you're fixing to be assaulted by an AI which may or may not be gathering en masse on something of a flank.

[Starmyth]

I am late to this discussion. I searched all of BigJim's posts and can find nary a file attachment. When i was a noob, I learned quickly that it was unwise to place anything where I would fire turn 0 suppression fire if I was the enemy and got to go first. This gets a bit tricky with a 200 unit core. I always save some of my turn 0 suppression arty for counter-battery. My turn 0 arty fire is always to suppress the enemy's arty. (computer generated long campaigns)