Questions and observations on CEP

[gbethel]

How is CEP supposed to function in CMANO? Do CEP values represent best case or real world? Are there other factors that are taken into account in weapon accuracy besides just CEP? Things like visibility, weather, sea state, deflection, target speed, sensors, etc.

My understanding is that 50% of rounds should strike within the CEP value, 93.7% within 2x CEP value and 99.8% within 3x CEP. Only 0.2% should strike at greater than 3x CEP.

My observations are that at least for some weapons that this is definitely not the case. Using the 76mm/62 OTO Melera Super Rapido DART, DAVIDE, STRALES most impacts are greater than 2x CEP. This weapon is supposed to be a guided projectile. The witnessed accuracy of the DART/DAVIDE is just deplorable, even if it were unguided. With the 76mm/60 China H/PJ-26 (AK-176 Copy) I have seen rounds impact as much as 13.2 nautical miles from the target. I don't even know how that is possible because the weapons range is only 6 nautical miles. First time I noticed this was because I could see rounds impacting on the screen all over the place and then I looked in the message log and all the log entries were for impacts 13.1-13.2nm from the target. In this instance the weapon was mounted on a Type 056 Jiangdao.

If this isn't a bug that needs to be corrected then please help me understand what causes this behavior or what I can do as a player or scenario designer to achieve the published CEP values. Are there any passive or active sensors that may help to improve accuracy, such as laser range finders or weapon directors or fire control radars? These sensors were present on the platforms when this type of behavior was witnessed but I didn't check to see that the sensors were actually active at that time.

Also, at what distance is the CEP value calculated? Is CEP taken from the weapons maximum range or at some fixed distance, for example 1 nautical mile. CEP of 80 meters at a distance of 6nm is bad when converted to MOA but if it is calculated at 1nm then that is just atrocious. Is accuracy supposed to improve at reduced ranges, because I am really not seeing that?

All of these observations were made using v1.14 build 998.10 and database 474.

[gbethel]

This is an example from my message log. Fire control radar is active. Laser range-finder is not.

10:33:18 PM - 10:33:18 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:17 PM - 10:33:17 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:17 PM - 10:33:17 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:16 PM - 10:33:16 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:16 PM - 10:33:16 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:15 PM - 10:33:15 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:15 PM - 10:33:15 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:15 PM - 10:33:15 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:14 PM - 10:33:14 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:14 PM - 10:33:14 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:13 PM - 10:33:13 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.8nm
10:33:13 PM - 10:33:13 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:13 PM - 10:33:13 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:12 PM - 10:33:12 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.8nm
10:33:12 PM - 10:33:12 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.8nm
10:33:11 PM - 10:33:11 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.8nm
10:33:11 PM - 10:33:11 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.8nm
10:33:10 PM - 10:33:10 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.9nm
10:33:10 PM - 10:33:10 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.8nm
10:33:10 PM - 10:33:10 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] missed Fishing Charter [47m] by 19.8nm
10:33:09 PM - 10:33:09 PM - Weapon: 76mm/60 China H/PJ-26 HE Burst [2 rnds] has malfunctioned.

[c3k]

What's with the very first message which states that the weapon has malfunctioned? Perhaps that accounts for the misses?

[Dimitris]

Can you please post a save of the issue.

[Dimitris]

bump

[SeaQueen]

I've noticed some issues with CEP as well. It looks to me as if the distribution from which the miss distances is drawn is the uniform distribution, not the Gaussian distribution. As a result, you get too many far misses and not enough near misses.

I know... I have no life.

[Dimitris]

Bumping this again as we are going over the CEP formulas for a do-over. gbethel if you are able to reproduce the multi-nm CEP issue please post a suitable save here for investigation. Thanks.

[SeaQueen]

I don't know if this might be useful for not, but the standard algorithm for producing Gaussian distributed random numbers is called the Box-Muller transform. The thing that confuses people about it, is that it's apparently easier to produce two random numbers than it is to generate one, so it actually produces two and you can just throw one away.

https://en.wikipedia.org/wiki/Box%E2%80 ... _transform

http://mathworld.wolfram.com/Box-Muller ... ation.html

https://theclevermachine.wordpress.com/ ... transform/

[AndrewJ]

Or you can just stack three or more uniform distributions, and get something that approximates the normal that way. I've no idea what's the most computationally efficient. It works in Excel! [:D]

[SeaQueen]

Box-Muller by far, because it's just performing some math on uniformly distributed random numbers.