Question about breaking target lock

[Claymore]

Question: In CMBB, when engaging in AFV combat and the target lock is broken (loss of red line), however momentarily, does the "to hit" probability reset back to "first shot" status?

For example in Zitadell you can engage in some very long range (1600m) tank/gun/tank duels. It is very common that sometime during firing the target lock is broken, if only for 1-2 seconds. It appears that although the facing of the tank/gun is "sticky" to the target, the range data is lost immediately. This is of course wildly inaccurate. CMBO suffered from the gross error of resetting the hit probabilites back to first shot status in these cases.

Has this been tweaked/fixed in CMBB?

(insert a "praying Graemlin" here if anyone can locate one)

Cheers

MRD

[ September 10, 2002, 04:04 PM: Message edited by: Claymore ]

[Claymore]

I realize it's been just a short time so this is a shameless self-serving bump to the top just before I knock off work.

Anyone else just a little bit interested in an answer to this question? From looking at all the chatter about MODs, doodads, and demos I would have thought that questions about the gritty basics would have generated more talk.

Sheesh! Moaning and whining has dominated this forum for the last couple of days people. :mad:

Cheers

MRD

[Liebchen]

We're all moaning and whining because we don't have our CDs yet! :mad:

[Claymore]

Another couple of hours goes by and nothing (except one piece of dryer lint) gets posted to this thread. Blech!

Nobody curious at all??

No response from BFC? Bueller? Bueller?

[John Kettler]

I am, and I can't even run the demo, much less CMBB.

Regards,

John Kettler

[gunnergoz]

Obviously, the BTS crew doesn't have target lock on this thread yet...

[Treeburst155]

This is a good question, Claymore. I'd be interested in the answer.

Treeburst155 out.

[Dirtweasle]

What would be an acceptable amount of time from loss of target to loss of range data where there is a reset for first shot probability?

Wouldn't it also depend on if there were other targets about?

What if they begin to lay the gun tube onto another target and the first pops back into sight?

[John D Salt]

Originally posted by Dirtweasle:

What would be an acceptable amount of time from loss of target to loss of range data where there is a reset for first shot probability?

Recall that most of the error budget for anti-tank shooting in this period is accounted for by range estimation error. This error is typically quoted as being 20 to 25% of true range for an initial visual estimate. I don't know the mechanics of how CM calculates such things, but it seems obvious that a "first-shot penalty" of this magnitude would not be justified if the gunner has once established a corrected range.

I should have thought that the main penalty for switching targets from one to another and back again would be the time taken to re-lay. Where repetition time is controlled by loading speed, even this would not seem to matter much.

All the best,

John.

[Claymore]

[/tweak ON]

It seems like all the people at BFC are handling the "softball" questions these days. Lots of quick responses to stooopid Demo questions and self-congratulatory noises about the new movement orders though.

[/tweak OFF]

Rexford's excellent treatise on ballistics is brimming with information on the errors, and indeed John, range estimation error is at the head of the pack. 20-25% is spot on.

Ok John, lets ennumerate the possiblities and keep it really simple.

1. Gun Breaks Target Lock

1.1 W/O acquiring new target

1.1.1 if Delta T < 2 seconds NO PENALTY TO PROBABILITIES

(I'd even make an argument for up to 5 seconds)

1.1.2 T > 2 sec as time increases then range estimation error should increase proportionally.

(if this is too difficult for the algorithms then a Heavy-Side Function at T = 2 (or 5 sec) could be applied)

1.2 Gun acquires new target

1.2.1 Target within 50m of previous target NO PENALTY

1.2.2 Target outside of 50m Range error set to maximum - back to first shot probabilitis.

Cogent arguments can be made that 1.2 is impossible to implement with the current s/w engine. Hence I suggest that if a gun breaks target lock and aquires ANY new target then you devolve back to first shot probablities.

Cheers

MRD

[John D Salt]

Originally posted by Claymore:

[snips]

Ok John, lets ennumerate the possiblities and keep it really simple.

1. Gun Breaks Target Lock

1.1 W/O acquiring new target

1.1.1 if Delta T < 2 seconds NO PENALTY TO PROBABILITIES

(I'd even make an argument for up to 5 seconds)

1.1.2 T > 2 sec as time increases then range estimation error should increase proportionally.

(if this is too difficult for the algorithms then a Heavy-Side Function at T = 2 (or 5 sec) could be applied)

A what function? I'm afraid I can't follow your banter, Biggles.

Originally posted by Claymore:

1.2 Gun acquires new target

1.2.1 Target within 50m of previous target NO PENALTY

1.2.2 Target outside of 50m Range error set to maximum - back to first shot probabilitis.

Cogent arguments can be made that 1.2 is impossible to implement with the current s/w engine. Hence I suggest that if a gun breaks target lock and aquires ANY new target then you devolve back to first shot probablities.

Drastic, but believable, and pretty straightforward. What I don't like about it is the "cookie-cutters" of 2 sec in time and 50m in distance, which I think can be done away with.

Given that the weapons under discussion here are not strictly "locked" on targets, ISTM that faithful modelling of this question would require each weapon to remember its "correction of the moment", at least in range.

A rather rich model of direct-fire gunnery might work as follows; exactly how CM's model works I have no idea.

1. Once the target has been visually acquired (skating hastily over the complications of the Night Vision Lab's optical acquisition model here) the weapon commander makes an estimate of the range to the target. This estimate will be made by drawing a random variate from the normal distribution, with the expectation being the true range and the s.d. depending on the accuracy of the range estimation method used (visual estimation, map-reading, co-ax MG, stereoscopic rangefinder, reference to measure range-marker). The estimate will be "remembered" by the weapon entity as its "range of the moment".

2. Once the time for range estimation plus whichever is greater of loading and laying has elapsed, the first shot is fired. For this shot, a second normal random variate is generated, based on the ballistic properties of the gun and ammunition nature, and all the other things that go into an error budget (laying error, trunnion cant, trunnion jump, ballistic hump, blah, blah) except range estimation error. This is then applied to the aim point defined by the "range of the moment" to determine whether the shot is a hit, falls plus, or falls minus.

2a. For a really de-luxe model, a third random variate could be drawn to determine whether the round has been sensed by the firing crew. This would depend not only on where the round fell, but on the type of ammo (17-pdr APDS was notoriously hard to sense) and visibility conditions.

3. If the result was a hit, the firing crew repeat until the target is seen to be destroyed or they find something better to do. If the result was a miss, a new correction is made, depending on whether the round fell plus or minus. The accuracy of the correction, as well as the accuracy of the initial estimate, may depend strongly on the gunner and commander's skill rating. As a first-order guess, I would say that making the magnitude of the correction proportional to miss distance would be about right for direct fire corrections. A similar process to the one described here could be used when ranging mortars or HE at long range, and there might be standardised bracketing distances laid down for this kind of shooting before FFE is started.

4. When a switch is made between two targets, the probable error in the revised range estimate is based not on the range to the target, but on the distance between old and new targets. "Short switches" between closely-clumped targets are therefore likely to be pretty accurate.

Although it took a while to describe in words, this is a pretty straightforward scheme that requires only two or three random variates per shot, and needs the weapon entity to remember only one thing, namely its "range of the moment" in range. It should be pretty obvious how to adapt the method to account for errors in line as well as range.

One of the possibilities such a model would simulate is the case of a gunner making a "false correction". For some high-velocity weapons that are ballistically inaccurate but relatively insensitive to errors in range, it may sometimes be prefereable to fire a repeat shot wihtout making a correction.

All the best,

John.

[BeauCoupDinkyDau]

Target Lock?! "Goose, I'm too close for missles! Switching to guns!"

[Claymore]

JDS: "A what function? I'm afraid I can't follow your banter, Biggles."

Heavyside function...a step function by any other monicker

JDS: "Drastic, but believable, and pretty straightforward. What I don't like about it is the "cookie-cutters" of 2 sec in time and 50m in distance, which I think can be done away with."

My predjudice towards hard "constants" is from my career. In the varied and complex instruments I have delivered to my customers I have tried a number of different s/w approaches to "tuning" a system for optimum operation. The most flexible I have found involves implementing a numerical multi-variate representation of a process. The instrument performance can then be tweaked by changing the value multiplicative constants. As these constants typically appear in a separate text file independent of the compiled code, s/w modes do not imply the transfer of Mbs of mainline code. Obviously though if a new stimulus arises which cannot be adequately modeled the s/w algorithm must be changed.

To enter any discussion about what is possible or probable for BFC to implement is a complete non-starter...neither you or I have even the remotest idea as to the s/w structure.

But there is considerable grist in a discussion of what a possible implementation COULD look like.

JDS: "A rather rich model of direct-fire gunnery might work as follows; exactly how CM's model works I have no idea.

1. Once the target has been visually acquired (skating hastily over the complications of the Night Vision Lab's optical acquisition model here) the weapon commander makes an estimate of the range to the target. This estimate will be made by drawing a random variate from the normal distribution, with the expectation being the true range and the s.d. depending on the accuracy of the range estimation method used (visual estimation, map-reading, co-ax MG, stereoscopic rangefinder, reference to measure range-marker). The estimate will be "remembered" by the weapon entity as its "range of the moment"."

I agree with your premise so far. Pragmatically, Steve and BFC as a whole have truncated discussions quite abrupting by using arguements hinged on the premise that no valid historical data is available for some particular phenomenon to be modeled. As such, I predict that BFC would choose not to change their current algorithms based upon the dearth of derived or recorded information on range estimation using the above equipment.

JDS: "2. Once the time for range estimation plus whichever is greater of loading and laying has elapsed, the first shot is fired.

In my view there should also be a variable time for laying, loading, and firing based upon statistical models. The current CM model is rigid in its timing.

JDS: "For this shot, a second normal random variate is generated, based on the ballistic properties of the gun and ammunition nature, and all the other things that go into an error budget (laying error, trunnion cant, trunnion jump, ballistic hump, blah, blah) except range estimation error. This is then applied to the aim point defined by the "range of the moment" to determine whether the shot is a hit, falls plus, or falls minus.

2a. For a really de-luxe model, a third random variate could be drawn to determine whether the round has been sensed by the firing crew. This would depend not only on where the round fell, but on the type of ammo (17-pdr APDS was notoriously hard to sense) and visibility conditions. "

Couldn't agree with you more here John. Dust from the muzzle blast, for example, was such a problem that it was SOP for an AT gun commander to be displaced laterally. Discarding sabot rounds were incredibly difficult to see at distance, especially so for the 17pdr which suffered from a considerable dispersion. Allied optics had only fair quality and a limited field of view, making the job even more difficult at Ostfront distances.

JDS: "3. If the result was a hit, the firing crew repeat until the target is seen to be destroyed or they find something better to do. If the result was a miss, a new correction is made, depending on whether the round fell plus or minus. The accuracy of the correction, as well as the accuracy of the initial estimate, may depend strongly on the gunner and commander's skill rating. As a first-order guess, I would say that making the magnitude of the correction proportional to miss distance would be about right for direct fire corrections. A similar process to the one described here could be used when ranging mortars or HE at long range, and there might be standardised bracketing distances laid down for this kind of shooting before FFE is started.

Give me a few hours (or a day) and I'll try to have some input as to the various forces' training procedures for firing.

JDS: "4. When a switch is made between two targets, the probable error in the revised range estimate is based not on the range to the target, but on the distance between old and new targets. "Short switches" between closely-clumped targets are therefore likely to be pretty accurate.

SOP platoon formations of AFVs extended over fairly short distances. AAR's from the Ostfront mirror just this effect. A few rounds required to strike the first target and then consumptions verging on a single round per successive target.

JDS: "Although it took a while to describe in words, this is a pretty straightforward scheme that requires only two or three random variates per shot, and needs the weapon entity to remember only one thing, namely its "range of the moment" in range. It should be pretty obvious how to adapt the method to account for errors in line as well as range.

One of the possibilities such a model would simulate is the case of a gunner making a "false correction". For some high-velocity weapons that are ballistically inaccurate but relatively insensitive to errors in range, it may sometimes be prefereable to fire a repeat shot wihtout making a correction.

All the best,

John.

Good stuff boyo.

[ September 11, 2002, 01:19 PM: Message edited by: Claymore ]

[John D Salt]

Originally posted by Claymore:

JDS: "A what function? I'm afraid I can't follow your banter, Biggles."

Heavyside function...a step function by any other monicker

Rats. I was hoping in was something weird and wonderful.

Originally posted by Claymore:

[snips]

I agree with your premise so far. Pragmatically, Steve and BFC as a whole have truncated discussions quite abrupting by using arguements hinged on the premise that no valid historical data is available for some particular phenomenon to be modeled. As such, I predict that BFC would choose not to change their current algorithms based upon the dearth of derived or recorded information on range estimation using the above equipment.

Yes, but the rangefinding method that matters far more than any other is visual estimation, so I don;t think it would hurt terribly badly to assume that was the method always used.

In any case, I have estimates for the accuracy of infantry rangefinders, map-reading, ranging with HE and RMGs from a PRO file I'm too lazy to check the number of right now (but is one of the WO 291 series on Operational Research, I think) and Shephard, Hartley, Heysman, Thorpe & Bathe's book "Applied Operational Research" (many of the problems in which are clearly based on real OR investingations now archived under WO 291).

As so often in research into "hard" performance numbers, most of these are of the nature of being a "single data point". This is pretty unsatisfactory, but a marked improvement on "zero data points".

Originally posted by Claymore:

In my view there should also be a variable time for laying, loading, and firing based upon statistical models. The current CM model is rigid in its timing.

...and, despite earlier cautions about specualting on the internals of CM's software, I have to wonder if this is responsible for the relatively frequent occurence of two opposing AFVs killing each other.

Originally posted by Claymore:

Couldn't agree with you more here John. Dust from the muzzle blast, for example, was such a problem that it was SOP for an AT gun commander to be displaced laterally.

Another advantage for towed anti-tank guns, hooray!

Originally posted by Claymore:

JDS: A similar process to the one described here could be used when ranging mortars or HE at long range, and there might be standardised bracketing distances laid down for this kind of shooting before FFE is started.

Give me a few hours (or a day) and I'll try to have some input as to the various forces' training procedures for firing.

Anything you can find on this would be most welcome for addition to my "WW2 weapons effectiveness" file. I already have things like the ranging procedure for the 3-in mortar (day and night) and short and long bracket distances for HE shooting with some British WW2 tank guns, but the PRO is pretty light on such things for other countries.

All the best,

John.

[ September 11, 2002, 02:01 PM: Message edited by: John D Salt ]

[Claymore]

Things that go BUMP in the forum.

[MikeyD]

I'm pondering your question.

Basically you're asking (correct me if I'm wrong) if you've got a 1-in-6 chance of hitting a target and you break lock after two rounds fired then reacquire, do those first two rounds still count in the 1-in-6 average?

I'd bet that in the game each individual shot is given the same odds and hit probability isn't really accumulative. If you stop halfway through flipping a coin the 50/50 odds of heads-or-tails still applies. This is just my opinion, only the guy who wrote the game's code knows for sure in what manner the hit probability odds are modeled.

[MikeyD]

{post deleted. Double posting } :mad:

[ September 12, 2002, 11:21 AM: Message edited by: MikeyD ]

[Claymore]

Sorry Mike, the probability to hit a target goes up with each successive round, just like in real-life.

The CM engine in CMBO modelled this faithfully as long as your "lock" didn't break, however momentarily. For example, you are tracking a AFV as it parades in front of you at 600m. You miss with the first two shots and the LOS to the enemy vehicle is lost for 1-2 seconds as it rolls behind a house. Although BFC tweaked the "stickiness" of the turret facing in CMBO and your AFV doesn't immediately begin to slew to another target when LOS is lost, you do start climbing the shot probability ladder completely from scratch.

My question to BFC, responded to so far with stunning clarity is...has this been changed in CMBB?

Sheesh...I mean...is it so hard to answer? Steve has been very active in other threads.

Maybe its my breath?

Cheer

MRD

[Battlefront.com]

Claymore,

Sorry, but with the plethora of posts things get pushed back several pages pretty quickly. I generaly come on line and respond to the first 1-2 pages only as that is generally all I have time for (if that). I did look for this thread yesterday but stopped at page 3 without finding it

Some comments to your comments above:

I agree with your premise so far. Pragmatically, Steve and BFC as a whole have truncated discussions quite abrupting by using arguements hinged on the premise that no valid historical data is available for some particular phenomenon to be modeled. As such, I predict that BFC would choose not to change their current algorithms based upon the dearth of derived or recorded information on range estimation using the above equipment.

Correct. We have looked into this stuff many times in the past, and the simple fact is that there are no conclusive facts There is, however, data that we have managed to cobble together that at least gives us ballpark estimates to model. And that is about the best one can hope for unless someone travels back in time and rides with a couple hundred crews of various skill levels into combat with a stopwatch

In my view there should also be a variable time for laying, loading, and firing based upon statistical models. The current CM model is rigid in its timing.

This is not true. Times are first of all variable due to crew Experience. There is also some degree of variability within that. Pehaps it should be more extreme (Pvt Numbskull dropping the AP round or grabbing HE instead and having to switch back again), but variability is in both CMBO and CMBB.

Couldn't agree with you more here John. Dust from the muzzle blast, for example, was such a problem that it was SOP for an AT gun commander to be displaced laterally. Discarding sabot rounds were incredibly difficult to see at distance, especially so for the 17pdr which suffered from a considerable dispersion. Allied optics had only fair quality and a limited field of view, making the job even more difficult at Ostfront distances.

True, but unfortunately variable to the nth degree and quite subjective at best. So there is only so much we can explicitely simulate.

SOP platoon formations of AFVs extended over fairly short distances. AAR's from the Ostfront mirror just this effect. A few rounds required to strike the first target and then consumptions verging on a single round per successive target.

Yes, and this makes sense.

Unfortunately, each unit does not remember the range and relative positions of everything it shoots. Accuracy does improve vs. a single target with successive shots, but once contact with that unit is broken the benefit ends.

Why is this significant factor not simulated? The simple reason is that it is a lot more complex than it might appear. At least from a simulation standpoint. Charles looked into coding up such behavior and found it to be very difficult to do with the existing engine (even back in CMBO development). Therefore he passed on doing more with this.

Thankfully most situations would not benefit from this information retention. Range estimations are next to useless (in practical terms) if the situation changes even slightly. The most favorable situation is one stationary shooter vs. several closely spaced, stationary targets all within LOS. This situation does not come up that often in regular games, and therefore is (by and large) not an issue.

However, that is not an excuse to say this is an unimportant feature. Quite the contrary, we regret not having it coded into the game, even if we still feel it was the right decision to make. And that means this will very likely (almost certainly) simulated with the engine rewrite along with a much more detailed accounting of gunner issues instead of the more abstracted factors we have in there now.

Thanks,

Steve

[Claymore]

Thanks Steve.

Just another reason to await the engine re-write.

Cheers

MRD