Amazing Russian Tank Gunnery

[Ssnake]

Keep in mind that the computer is NOT operating in terms of pixel size of a target in a sight. There's some gross simplifications going on. First we chack if line of sight is given at all. Then we roll the dice whether a target will be spotted, the likelihood depending on a number of variables. Once that the target IS spotted and the fire control orders of the unit allow to engage, the target WILL get engaged (in the order of priority, if multiple targets are in range of the shooting unit) - unless other factors prevent it (like, whether the available ammunition may actually hurt the target or not).

I don't think that we will keep these simplifications forever. There's alwayas room for improvement, this being one obvious candidate. In defense of the current solution I have to say though that this basic method has been in the simulation since the very beginning (with some refinements in the past ten years, admittedly). MOST OF THE TIME these simplifications works well and produce adequate results. That is evident because it requires special circumstances to bring up this discussion in the first place. I think that for a two-party scenario well over 99.9% of all engagements are actually performed in a plausible fashion.

It's just those "one in a thousand" cases where the immersion breaks down. Given that with most scenarios we have maybe 300 combatants per side (don't forget the dismounts) and that most combatants will have multiple sightings before they get eliminated, "one in a thousand" can actually translate to once or twice per game session. So we'd ideally want a confidence of 99.999% (which is pretty crazy if you start thinking about it).

[hoggydog]

I tried the T-80U scenario with BM-42M rounds and Refleks stowed away. The three tanks started to engage me at 3300 m. I guess ill just used the T-80Us like that for now.

Hoggydog, the T-80U should not engage us from 5km with sabots at all. Given the optics on the Leo 2s and Abrams tanks (excluding the M1A2 SEP) tracking and engaging targets from over 4000m is just not effective. The ballistic computer wont even provide a solution for distances beyond 4000m.

I wonder how T-80Us can pull off stunts like that. With engagement ranges of over 4000 m, a target would not be much more than a few pixels in diameter in the thermals even at max magnification.

If a Challenger back in 91 can get a kill at over 5k I don't think its impossible for a T-80 to attempt the same. Ok there may be reasons why SB is behaving in this way and maybe there is a room for improvement in the modelling but I could see a moving tank with my naked eye at well over 5k I don't think the optics of the T-80 would have any trouble.

[lavictoireestlavie]

SSnake, i changed the difficulty setting and it made no difference.

As far as i know the Challenger 1 was engaging a static t-62 tank from about 5110 m. Neither vehicle was moving.

In my scenario the t-80Us are moving towards me. In another we are both moving and in yet another scenario i am moving sideways while they engage me.

Some T-80 and T-90 info that might be of relevance:

http://www.kotsch88.de/f_t-80_fla.htm

http://www.kotsch88.de/f_1g46.htm#ballrechner

in English:

http://www.microsofttranslator.com/bv.aspx?from=de&to=en&a=http%3A%2F%2Fwww.kotsch88.de%2Ff_t-80_fla.htm

http://www.microsofttranslator.com/bv.aspx?from=de&to=en&a=http://www.kotsch88.de/f_1g46.htm#ballrechner

[Companion]

SSnake, i changed the difficulty setting and it made no difference.

As far as i know the Challenger 1 was engaging a static t-62 tank from about 5110 m. Neither vehicle was moving.

In my scenario the t-80Us are moving towards me. In another we are both moving and in yet another scenario i am moving sideways while they engage me.

Some T-80 and T-90 info that might be of relevance:

http://www.kotsch88.de/f_t-80_fla.htm

http://www.kotsch88.de/f_1g46.htm#ballrechner

in English:

http://www.microsofttranslator.com/bv.aspx?from=de&to=en&a=http%3A%2F%2Fwww.kotsch88.de%2Ff_t-80_fla.htm

http://www.microsofttranslator.com/bv.aspx?from=de&to=en&a=http://www.kotsch88.de/f_1g46.htm#ballrechner

Again, in your test, try using something other than Eastern kit in T-80's place.

I'm pretty sure that when crewed by AI, Leo AS1, M60, etc. would produce similar long range hits as T-80U did.

Ssnake's reply above is what you're looking for; when AI shoots, dice roll dictates hit. It might be that dice rule need some adjusting to keep up with upgrades of the software.

In any case, I highly doubt that there is some pro-Eastern kit bias as some posters here implied. (Quite the contrary, some of the Eastern kits are currently underrepresented)

[Maj.Hans]

The AI does some strange things with long range gunnery. Sometimes it hurts you, but sometimes it helps you too. They might be quite accurate "mechanically" speaking, but for some reason I feel like they don't observe the way their rounds land the same way humans do and that causes them problems when aiming at tanks that were carefully positioned to minimize their exposure.

Just a few nights ago I was running a scenario with a new player in the gunner's seat. We found ourselves on a reverse slope about 3500m away from a T-64 behind a small hill, swapping shots.

After firing a few rounds we realized the T-64 was hitting short. It was firing HEAT or HE and they were landing 10-15 yards short, in the exact same spot, every single time. So we just ceased firing, closed the GPS doors, buttoned up, watched him with the CWS scope and waited a few minutes for him to run out of ammo, then rolled up on top of the hill and engaged.

Couldn't get a good lase so I told the gunner to enter 3500 and fire. We watched it land short, so we added 200, that one went over, so we split the difference and got a solid hit. Then pumped two more into him to be sure.

[Invader ZIM]

I don't think the problem is so much of an AI accuracy one, as much as it's what the AI is actually using.

In my test scenarios, the T-80U seems to be acting exactly as it should, it will take 5km shots with AT-11, and at 3km will use BM-42. Without AT-11's in it's inventory, it will use the BM-42 at just over 3km, where it has a decent chance to hit.

But the discrepancy is that in other member's simulations, the T-80 seems to incorrectly use the BM-42 round about 2km outside of it's effective range, at 5km, when it should be using AT-11's. Or, failing to have AT-11's in it's inventory, should hold fire until within the 3.2km effective range of the BM-42

[hoggydog]

Could it be the AI misreading what ammo to load? It's selection of ammo is sometimes a little odd (e.g. the challenger Co and his obsession with loading WP as an anti everything weapon) This could be no more than that.

[Ssnake]

Each target has a (simplified) number or basic figures denoting its armor resistance from different angles for the different munition types. The computer will typically select the type of ammunition with the greatest overmatch (as it promises the greatest effect on target); it is not necessarily the ammo type with the highest chance to score a direct hit. Also, because if the simplifications involved some ammo type selection decisions may appear counterintuitive.

We don't want the computer to make a perfect calculation prior to his decision to engage or not; human crews wouldn't do that either, and it's simply absurd to assume that every crew has an encyclopedic knowledge about the armor array of a particular target taking into account the angle of impact, the range, and other factors.