I'm pretty much done writing the Python module for reading and writing theater files. This means anyone can easily manipulate them as Python dictionaries and write them back to theater text files. One problem is that the line order is not preserved even though the hierarchy is. Will this cause any issues?

Edit: http://i.imgur.com/Yg3isIr.png

引用自 g o：

i can but it makes everything else harder to see as well. r.i.p. my eyesight already sucks

This is why I'm blind at night haha.

引用自 Blueshy：

引用自 1 share = 1 prayer：

I proceeded to write a script that could generate a curve given the initial value, half-life, and desired number of points. This means if you can find data on the kinetic energy of a bullet you just need to eyeball the half-life and key it into the script. The script generates curves into a format you can just copy and paste into the theater file.

Half lives, kinetic energy, scripts computing it... love it!!! (My degree is in nuclear engineering.)

Although, I would have expected a logarithmic curve fit more than exponential one since at some point the bullet will lose enough energy to tumble and rapidly decay in energy. I still think that might be a more realistic representation of a bullet's actual "killing power." Your referenced source indicates that the *kinetic energy* falls off exponetially, but that is different than the actual *damage* a round might cause. Still, it it is interesting. I would like to see more sources for data. My original comment hinted at creating an equation that we could use to accurately generate quasi-realistic data for any round given a few different parameters that exist in the insurgency game already. That way we could integrate it into the ammuntion script writer to create damage tables for every type of ammunition in Sernix based on the known parameters of every weapon/ammunition. There are still unanswered questions around how may points we can feed into these scripts (the resolution) before we begin to encounter problems. Currently there are only two points and presumably the game is doing some sort of linear interpolation between them.

I think Daimyo's fear of having separate ammunition for each gun on sernix is solved when we switch to generating theater files from a script that automatically computes and outputs all the data. Indeed, for a human to manage changing and tweaking such a large file would be unfeasible but if we manage it in the style of a database...

I think this is where I will focus all my efforts after completing the weapon weight adjustments. I think once we properly set this all up it will free up a lot of time and energy to focus on implementing actual new features/scripts rather than worrying about being able to implement and control all the variables in the theater files manually.

I think penetration and hitgroup multipliers are enough to account for terminal ballistics unless you want to deal with bullet yaw and tumbling. Damage can just be the raw energy of the bullet.

We can actually calculate base damage(muzzle energy) for each gun based on the barrel length and optimal burn length of the cartridge with respect to some arbitrary damage value and generate an ammo for each one quite easily. Dealing with the theater files in it's entirety, however...

Edit: I retract my previous statement, I found a way to read the data into a python dict

I followed Blueshy's comment on improving the damage curve and noticed that the kinetic energy of a bullet roughly follows an exponential decay. Chart here - http://www.brassfetcher.com/Rifles/5.56mm/223%20Remington%20Kinetic%20Energy%20vs%20Range.png

I proceeded to write a script that could generate a curve given the initial value, half-life, and desired number of points. This means if you can find data on the kinetic energy of a bullet you just need to eyeball the half-life and key it into the script. The script generates curves into a format you can just copy and paste into the theater file.

Comparison of current 5.56 damage falloff and curve generated using "realistic" values -
https://docs.google.com/spreadsheets/d/1YqDPszTV-pWDxg3JlOffSL6FBbYaD2-rKGcM3lKNgRo/edit?usp=sharing

Python script (requires Python 3) - https://drive.google.com/file/d/0B-mv4PuuSfrjb2VrblV4a3dTQTQ/view?usp=sharing

引用自 🎄Hodor🎄：

Hm maybe I shd rephrase my earlier post, currently we have fixed most guns into broad catagories of bullet type for ease of modding. There's only so much you can tweak before players start commenting on unexpected handling. But improving draw + reload speed for 9mm/45acp smgs can avoid alot of the above and address some of the biggest complaints of gun handling which are those 2.

Unless you want to create an ammo type for every gun (for more damage refinement) to account for barrel length in real life, geo mean might have more hurdles besides the bullet types we follow for fixed damage catagories already.

I realised that when looking at a small comparison of all the rpk ammos haha.

Lower your screen brightness lmao

引用自 Good-Try Greg：

What about buffing ammo caches? Maybe if they were limited to one use per person, so the whole team could use them?

This is a good idea

引用自 Blueshy：

引用自 1 share = 1 prayer：

The player multipliers apply to each column of body part just like armor, thus the distribution remains the same. As for the statistical model, I'm not really assuming 100% hitrate. I'm saying that if the ap has X% more damage at some body parts the hp needs to have X% more damage at other body parts. And I haven't taken into account the accuracy also

If we assume there's the following player multipliers for the sake of example:

Head: 1
Body: 1
Legs: 0
Arms: 0

and we have some ammo multipliers,

Weapon 1:
Head: 1
Body: 1
Legs: 0
Arms: 0
Weapon 2:
Head: 0
Body: 0
Legs: 1
Arms: 1

You can clearly see that the player multipliers will change the damage distribution when computed. The same thing will happen when you use less binary values.

You're not assuming a "100% hitrate" *per se*, but you are in a way assuming that the bullet hits the target because you're calculating the damage based off an average of the multiplier values. You can totally do that, but whether or not the data represents a plausible senario relating to actual gameplay is up for debate...

You can't put 0 into the geometric mean. Multiplying each column by a fixed value just scales the distribution, I suggest you try it for yourself if you still don't believe me.

The average is not really an average but rather a comparison of the trade-offs. The index is meant for comparison between idices and scaling the graph does nothing for this purpose.

引用自 Blueshy：

引用自 1 share = 1 prayer：

By using the geometric mean, I throw aiming out the window anyway. And armor. If you multiply one column by some aim factor or armor factor the distribution would just scale but not skew or anything. The idea is that you can compare comparable groups of ammo such as all the 556 ammo and all the SMG ammo to see if they are equal choices. This doesn't take into account other things such as recoil and such so you can't compare between different guns yet.

That isn't true though. I'm talking about the player multipliers which are separate from the ammo multipliers. Those will skew the multiplier values even further, even if you continue to assume an averaged 100% hitrate across all the body parts because of it scaling the multipliers differently. And you're not throwing aiming out the window, you're just assuming a bad statistical model for it, whether you realize it or not. If the damage model the calculation is based on is flawed the results will be equally as worthless.

The player multipliers apply to each column of body part just like armor, thus the distribution remains the same. As for the statistical model, I'm not really assuming 100% hitrate. I'm saying that if the ap has X% more damage at some body parts the hp needs to have X% more damage at other body parts. And I haven't taken into account the accuracy also

引用自 Blueshy：

I wonder if we could actually tune the penetration and armor to more realistic values by creating more in depth data sets for how they behave. So instead of interpolating by just two numbers, I think you use 1m and 139m(?), you would have the script auto generate a more parabolic decrease in damage/penetration power using more data points. Don't know where you could look to get the real world data for this. I'm sure you could find something online though and perhaps just come up with a generalized equation for the parabolic dropoff of damage/penetration based on bullet type, caliber, barrel length, etc.

Brassfetcher.com seems to have some interesting data for travel distance and penetration. Will it lag the server if you put too many points though?

引用自 Blueshy：

I think you need to rethink your equations/categories. For your magdump one it looks like you're just taking the average of the ammo multipliers times the damage, which is fine I guess, but this approach kinda throws the idea of aiming out the window. It's unlikely that someone would hit both the leg and head equally while spraying down an enemy while also maintaining 100% accuracy. Also some of those senarios don't exist, there is no 7.62x54 30rd gun in the game. You are also missing factoring in the player multipliers which will further skew the data.

引用自 Daimyo：

I'd be very interested in a few things:

-Adding armor value reductions for bullet damage index which we do on our original sheet.
-Pentration power at the various ranges. To get meters, divide range by 52.46. To get wall penetration length, also divide by 52.46.

By using the geometric mean, I throw aiming out the window anyway. And armor. If you multiply one column by some aim factor or armor factor the distribution would just scale but not skew or anything. The idea is that you can compare comparable groups of ammo such as all the 556 ammo and all the SMG ammo to see if they are equal choices. This doesn't take into account other things such as recoil and such so you can't compare between different guns yet.

For now, if you want to simulate aiming for certain parts of the body, just change the column to 1. It has the same effect as removing that hitgroup.

What I see from the data is that if I were to pick the RPK and spray it a lot, I'd pick FMJ ammo.

引用自 Daimyo：

-Pentration power at the various ranges. To get meters, divide range by 52.46. To get wall penetration length, also divide by 52.46.

I did actually manage to extract that data but due to the presentation in Excel it wasn't actually usable so I removed the columns. The left column in the theater file goes to the top header in the Excel so...

If anybody wants I can share my python script but it's a mess, I had no experience parsing text files before this. I would also write my parser very differently after encountering those spaces hiding in the indentation LOL