[SOURCEMOD] Brutal-OXCE 7.12.1

[jnarical]

What is "Realistic accuracy and cover system"?

I understand vanilla already accounts for cover in the way that if part of the target is covered and shot hits the cover it does not hit the target. Sure this can be calculated and displayed to user for convenience but what exactly is changed?

I feel like my messages here are invisible… ok, Brutal OXCE incorporates this: https://github.com/narical/openxcom-accuracy

This is the separate mod, with old accuracy code reworked too, to improve it and fix bugs.



You can get some insights here under spoiler:
 https://r.tapatalk.com/shareLink/topic?share_fid=91863&share_tid=10967&share_pid=159302&url=https%3A%2F%2Fopenxcom%2Eorg%2Fforum%2Findex%2Ephp%3Ftopic%3D10967%2Emsg159302%23msg159302&share_type=t&link_source=app

Sorry for Tapatalk link, I’m from the phone forum client - it’s actually this very same forum topic, just ~20 messages away from this

Upd: part about 5%/95% is outdated, I’ve changed that long ago but forgot to fix description. Also, current version of algorithm, increasing accuracy for close-distance shot, differs too. For snap/auto, there’s no multiplier bonus. This algorithm is a subject to change, again, 5th time in a row - I plan to add bonus from initial accuracy, exceeding 100%, to workaround target cover.

Here’s an example.

Current state:
Initial calculated accuracy: 210%
Target visibility: 10%
Final accuracy: 210*0.1=21%

Planned change:
Initial calculated accuracy: 210%
Accuracy overshoots 100% by: 110%
Additional bonus: 110/2=55%
Target visibility: 10%
Final accuracy: 210*0.1+55=21%+55%=76%

That way, snipers are actually snipers, and high accuracy really helps against cover

Also, it could help with accuracy with close-midrange shots in dense environments (at least I hope so)

[Xilmi]

Thank you for confirming.
In this case where I can find a list of configuration parameters with defaults? I unpacked the zip but there is no readme there too.

The first post of this thread has the list of the new options.
I'm not sure whether "options" translates to "configuration parameters" though. So not sure this answers your question.

On first launch the new options are automatically set to their defaults and then stored in options.cfg.

[Xilmi]

OXCE launch file is OpenXcomEx.exe
Yours is OpenXcom.exe

Should I launch yours?

I personally would have recommended to put it into a separate folder.
But yes, in order to launch Brutal-OXCE you have to launch the exe from it's zip-file, not the one that previously has been there.

[Abyss]

Started to montage videos.
Tooks a bit longer than expected, plus haven't been working with premiere pro for, like, 6-7 years.
Further will be faster.
This is XCF Red Dawn forward base assault, SH + Ironman, BAI 7.10.2, balanced aggressiveness, 40% randomization. Joy's accuracy settings are off by that moment, too time consuming. 

https://youtu.be/ARGz0PVR4Ek

[Abyss]

"Fun":
"Immersive":
"Smart:"
"Cheater":

Hi Xilmi!

I think all these condensed options should be paused to implement right now, because the development goes on and many things are getting tweaked.
Avoiding proxies and mines - should be totally up to player. 
Some missions are impossible to beat w/o that. Like, underground HQ assaults and base defense vs enemy +1 tier from you. Which happens.
Pre-primed grenades is something I play with from 1 turn, but never use myself. And so on.
Please don't yet upgrade this. Or, if upgrade, make customizable option too.

To Jounarical:

Current state:
Initial calculated accuracy: 210%
Target visibility: 10%
Final accuracy: 210*0.1=21%

Planned change:
Initial calculated accuracy: 210%
Accuracy overshoots 100% by: 110%
Additional bonus: 110/2=55%
Target visibility: 10%
Final accuracy: 210*0.1+55=21%+55%=76%

Here we have balancing issues.

Accuracy overshoots are interesting idea overall, but we rarely counter enemies that exceed 100 accuracy themselves. Plus, 100 accuracy is abstract measure, too. The next thing comes to mind is that any unit which exceeds 100 acc will have too much advantage, while others will miss.

Regarding RA:
- first of all, it should work not only with super-accurate shots, but with every shot
- it should be psychologically reliable: player sees sprites, not voxels. If actual frame of the window makes 80% cover, but the drawing shows large window that exposes all torso and head (33%% cover), player will be dissatisfied with the hitchances. Same with two and a half leaves on a thin tree in front of the unit. Same to wooden fence, which shouldn't stop high caliber at all!   
- fights should be yet dynamic.

What if make it hybrid:
partial weights:
Overshoot mechanic 25%
Weight of the cover/sight 45%
bullet pre-roll damage 20% (0 = 0%, 100+ = 20%)
distance (yet I somehow believe it should matter) 10%

[Juku121]

Planned change:
Initial calculated accuracy: 210%
Accuracy overshoots 100% by: 110%
Additional bonus: 110/2=55%
Target visibility: 10%
Final accuracy: 210*0.1+55=21%+55%=76%

That looks a bit better. Personally, I'd instead use accuracy overflow as cover reduction. E.g. in the above calculation, 100%(cap)*(0.1+(0.9-0.9/2.1)) ~= 57%.

But the big problem with this approach is that now aimed shot is king and snap/auto is relegated to knife-fighting ranges only. A reversal of the original 'auto-shot or go home' problem, if you will. Good-bye, heavy machine guns and miniguns, it was nice knowing you. 😢

it should be psychologically reliable: player sees sprites, not voxels.

This is a core feature of OXC and one should not touch this lightly. It's the responsibility of the people who create sprites and match them with voxels (i.e. the modders) to do this. If the engine starts having its own ideas of how a unit/sprite should look like, there's going to be no end of trouble(shooting). Maybe the modder wanted to create a shoot-through hologram, but the engine knows better and makes a giant shot-blocking bedsheet...

[Alpha Centauri Bear]

How do I install/transfer mods to the freshly copied Brutal-OXCE directory? Just copy "user" subfolder?

[Abyss]

If the engine starts having its own ideas of how a unit/sprite should look like, there's going to be no end of trouble(shooting). Maybe the modder wanted to create a shoot-through hologram, but the engine knows better and makes a giant shot-blocking bedsheet...

Oh ho, you got me wrong. I was suggesting not about fucking up engine, but either about these two bushes that lower hitchance from 50% to 5%, as from the player's perspective is weird. The entire passage was about bumping up chances.

To Joy:
I still think of the formula. Let's take 50% snapshot from rifle which degrades over 15 tiles. At 30 it will be 30%. The bullet power is 35
1) Make weight of the cover 0.40-0,50. Then plain hitchances are 25-30 and 15-18%, respectively
2) Make accuracy overshoot weight 0,25, e.g. all above 100 divided by 4. This particular shots are not affected = +0%
3) Make bullet pre-roll power-to-accuracy 0,15: (0,15*35/100) = + ~5%
4) Maybe distance to cover factor is not that bad idea. The farer the unit, the better chances it has to move a bit while projectile flies... Like 0,1 (100 tiles -10%, 1 tile 0  ...or... 100 tiles -5%, 1 tile +5%)
5) everything else that seems relevant to balance miniguns, rifles, sniper rifles and other stuff
It's better to make discussion with more modders, because this is relevant to incorporation of RA into design after some time passes.

As per me, all shots on covered units should degrade no more than by 50-60% of plain accuracy

[Xilmi]

How do I install/transfer mods to the freshly copied Brutal-OXCE directory? Just copy "user" subfolder?

Yes, this should work.
If you have the Documents/OpenXCom-folder, you don't even need to transfer them.

[Xilmi]

Please don't yet upgrade this. Or, if upgrade, make customizable option too.

This was the idea: Have presets that modify the other options but leave the other options there.

Something really scary was that I found bugs in the way the AI cheated when you enabled targeting-mode 4 and bug-hunt-mode. Basically they didn't always make use of this information. I fixed that. This is totally next level!

Another thing is that some settings I thought weren't as difficult are actually more difficult than others that I thought that were.

Like the lower-aggressiveness is actually weaker when there's a lot of cover for the player because when the AI operates with outdated information you can sneak up on them. If it plays more aggressive, it often actually is stronger.

This tells me that the algorithm that determines the ideal behavior hasn't been found yet.

[Abyss]

To Xilmi

This tells me that the algorithm that determines the ideal behavior hasn't been found yet.

Yep, and next video will be very interesting and informative in the way how algorithm is not balanced yet.
It will be Red Dawn HQ night assault with 16 agents vs ~100 enemies. Same BAI 7.10.2, 40% random, aggressiveness 2, dynamic aggressiveness, TargB 3.,
I have NO idea how would I do that during the day. No chances. Completely different game.
But when BAI doesn't see enemy, even if it knows it is somewhere there in the dark - it is kind of pleasant to do with. 

With the video, I will make a suggestion of "Sea Battle" suppression mechanism. 

This was the idea: Have presets that modify the other options but leave the other options there.

As most casual+ players come for mods, it's good to be checked through these.
Also, I suggest more dialogue between you guys.
 
Btw, I came with particular question: How does dynamic aggressiveness setting exactly interferes with balanced aggressiveness (2)?

[Xilmi]

Btw, I came with particular question: How does dynamic aggressiveness setting exactly interferes with balanced aggressiveness (2)?

Dynamic ignores whatever you have set as aggressiveness. It starts out at max and reduces it when morale goes down. If they get kills that restore their morale they become more aggressive again.

I also think that in cases where the AI doesn't see much like you described being more aggressive clearly would help.

I just tried again against higher aggressiveness and did much better this time. This is also because of the snow-bally nature of panicking. Even a small difference can have a massive impact on the outcome.

Now gonna test aggressiveness 0.

[jnarical]

It will be Red Dawn HQ night assault with 16 agents vs ~100 enemies.

That’s what I was talking about. Why not 200 enemies? Or 1000? This is nuts, dude. How do I suppose to make accuracy mechanics which works similar in terms of fun both for original game with 15 missions per month / 15-30 enemies vs this nonsense?)

[jnarical]

Accuracy overshoots are interesting idea overall, but we rarely counter enemies that exceed 100 accuracy themselves. Plus, 100 accuracy is abstract measure, too. The next thing comes to mind is that any unit which exceeds 100 acc will have too much advantage, while others will miss.

We’re talking about 16 agents vs 100 enemies here… this game IS asymmetrical, where you
play with 1:5-6 ratio against enemy forces. It was never was about fair fight. There’s sheer fun obliterating hordes of enemies from afar, and I don’t know why ask for enemies with 100+ accuracy.

Regarding RA:
- first of all, it should work not only with super-accurate shots, but with every shot

We already have that) it’s called vanilla accuracy. Works the same ignoring any cover, and even distance if you disable extender accuracy. I didn’t have an intention to copy that, I’m trying to make something different. Not “one-size-fits-all” and not “better in any possible way”, just different kind of fun. (Better in many ways is still the goal tho).

- it should be psychologically reliable: player sees sprites, not voxels. If actual frame of the window makes 80% cover, but the drawing shows large window that exposes all torso and head (33%% cover), player will be dissatisfied with the hitchances. Same with two and a half leaves on a thin tree in front of the unit. Same to wooden fence, which shouldn't stop high caliber at all!

We are talking about the game full of conventions. One of them is the fact, that deep in our imagination we have a mix of turn-based and real-time action. We know the game represents real and fast-paced battle. That’s why it’s totally ok to have 50% cover in high crops. Grass doesn’t give a protection against bullet, but in real life it gives you a chance to hide and mislead enemy about your true position. Put in other way, covered visibility affects chance to hit. Nothing confusing for players here. You get highly covered enemy? Deal with it, that’s how the game (in this accuracy mode) works.

What if make it hybrid:
partial weights:
Overshoot mechanic 25%
Weight of the cover/sight 45%
bullet pre-roll damage 20% (0 = 0%, 100+ = 20%)
distance (yet I somehow believe it should matter) 10%

In my experience, even one new variable which seems to be clear and have predictable effects - turns out as unpredictable, given the sheer amount of different mods and weapons’ types, and leads to balancing hell and a bunch of iterations of rewriting code… your approach with “let’s introduce a bunch of variables with totally arbitrary values and hope for the best” won’t work, it doesn’t work even for one variable. Things tend to break in most unexpected ways)

I don’t say anything against your ideas, just that implementing should be iterative. And I still can’t configure working toolchain to build the game under Windows, yesterday’ve spent 2+ hours in addition to 4.5 earlier, still no luck.

[jnarical]

Oh ho, you got me wrong. I was suggesting not about fucking up engine, but either about these two bushes that lower hitchance from 50% to 5%, as from the player's perspective is weird.

Every unit checks three different positions for fire, shifting left and right (with off-centre shooting on, which I highly recommend, cause don’t know hot to force turn it on)  and chooses the one with better enemy exposure. If after that you’ve still got low visibility - the very most of the time you’ll get it with in-game sprite representation, too. I personally didn’t have such issues you’re talking about.