[SOURCEMOD] Brutal-OXCE 7.12.1

[Xilmi]

I've tried to convince him to stream on YouTube or twitch again, like he did once before. But mostly he streams directly on Discord to me in a 1:1-session. I'll tell him that there's others too who want to see him play.

He brings like 40 of them to a mission and carpets big parts of the map with them. He basically seals off entire sections.

[Juku121]

Hmm, from an XCF point of view, it'd be easier to just mortar such areas, or better yet, nearby cover. Faster and probably also kills some enemies. Or at least their cover. Throwing isn't necessarily very accurate at long ranges due to no LoS penalty. And the maps are bigger, 60x60 and up, frequently lots of building cover as well. Sometimes pretty high buildings, too.

OG 50x50 maps are probably somewhat better suited for this, but I'd appreciate it if you could convince him to share at least a few videos. Would make for some neat promotional material for you, too.

[Xilmi]

I talked to him and said that he shall stream to YouTube in the future so others can watch too. (Live or later)
He said he'll do it tomorrow.

He mostly plays TFTD, which I think uses 60x60 as default-map-sizes.

Your remark about the buildings is definitely correct. On the mostly flat under-water-levels the zoning with the proxies works a lot better than it does in the terror-missions where they often hide inside of buildings.

Also success is relative. He does a little better than I do but usually is forced to restart soon too. At least with the more current version. The video he already has on his channel is from 3.7.something. A lot has changed since then.

[Xilmi]

Stream will be here:

https://www.youtube.com/watch?v=T0Ydl82-0JU

2023-03-22 at 16 CET

[Xilmi]

4.0.0:

Massive update to the target- and attack-mode-selection-algorithm.
Instead of just comparing the options at the current position it will now also compare options involving walking towards the target and using weapons at a shorter range with higher accuracy or melee-weapons if available. For that it will subtract the amount of time-units it needs walking over and sees how it impacts the amount of attacks it can do, the additional possibilities, whether it can sneak up or is in vision for reaction-fire and the impact on accuracy.

Certain events cause units controlled by Brutal-AI to weak up again when they had already finished their turn, just in case there's now something to do. This was quite taxing on the turn-time as the entire AI-routine was run again. On big maps with many enemies and large sight-radius this could mean a lot of unneccessary AI-cycles. There now is a sanity-check in place for that where a very minimalist pathfinding-check is done to see if reactivaction could make sense at all. Basically when the unit cannot move anymore anyways and also doesn't have enough TUs for any attack, it will not be woken up again and not run another AI-cycle.

Added a new option in Advanced=>Battlescape called "Strict checking for blockages", which by default is off. This allows compatibility with potentially improperly constructed maps containing BigWalls, where they don't belong at the cost of making deliberate bigwalls orthogonally leaky.

Updated OXCE-base to 7.8.14.

Damage vs. Armor comparisons now take into account what damage-distribution-model (UFO vs. TFTD) is being used.

When deciding where to take cover after peaking, cover is now evaluated relative to the best cover in range rather than any amount of cover being considered sufficient.

Units in sweep-mode no longer care about triggering reaction-fire.

[Xilmi]

Brutal-OXCE 4.1.0 - Advanced Cover Assessment Update:

The AI's cover-evaluation no longer depends on what "strict checking of blockages"-option is chosen.

Importance of avoiding line of sight to assumed position enemy could reach when walking towards us is now considered much higher when looking for a safe spot to advance to.

When avoiding to stand close to allied units, units that are behind a wall are no longer taken into consideration.

When evaluating the importance of cover from certain directions the distance towards enemies in this direction is now taken into consideration.

When evaluation the importance of cover form certain direction the cover towards the direction 45° to the left and righ to that direction are now factored in with half the amount of cover-score each.

Reworked AI-cover-evaluation to consider blocking of vision and blocking of damage instead of blocking of movement.
So the enemies shall no longer consider stuff like seaweed, windows or trash-can as viable cover.

Differences in elevation now also are taken into consideraton for evaluating cover.

[Finnik]

Probably I missed that, or may be not, but I will ask - have you considered letting the AI use medkits from inventory to heal wounds if any? In many mods, XCF for instance, you ccan deal with enemies by letting them bleed out even if they have a medkit.

[Meridian]

Being able to use a medikit is a nice idea.
Worth implementing.

However, the rationale above is shaky.
Players who slightly hurt an enemy and then wait for 50 turns for them to die are simply "not a representative sample".
It doesn't even really count as a bug, abuse, cheese or exploit... since such gameplay is more detrimental to the player (in game and in real life) than to the AI.

[Finnik]

Yeah, I agree that is cheesing, not a bug any close. But I've seen it on streams and also, TBH, used it in XCF a couple of times - when wounded not very bulky enemy walked away, you can be pretty sure it will stun itself in a several turns.


Although, enemy healing is a kinda frustrating, so it should not happen really often. In turn, it can be a surprise for the player when the enemy decides to heal up instead of shooting.

[Juku121]

Players who slightly hurt an enemy and then wait for 50 turns for them to die are simply "not a representative sample".

While that may be technically true, almost every XCF streamer I've seen has used it at some point. So this counter-argument is also shaky.

[Meridian]

While that may be technically true, almost every XCF streamer I've seen has used it at some point. So this counter-argument is also shaky.

Yes, almost every XCF streamer I've seen used this at some point as well.
Out of curiosity, or frustration.

And it didn't work, because they couldn't afford to wait 50 turns.

So this counter-counter-argument is also shaky

[Solarius Scorch]

Can we just agree that seeing enemies - and more importantly, civilians! - using medikits to help one another is nice?

Bonus points if an X-Com soldier can walk up to a civilian doctor and expect to be healed...

[Nord]

AI units using healing is quite immersive. Remember though, the example of Afterlight, where one wounded reticulan leads to all his friends to swap guns to medikits, leaving them all defenseless.

[Xilmi]

Probably I missed that, or may be not, but I will ask - have you considered letting the AI use medkits from inventory to heal wounds if any? In many mods, XCF for instance, you ccan deal with enemies by letting them bleed out even if they have a medkit.

Not yet. It's not that high on my todo-list. But I guess I'll eventually get to it.
Would also be nice for auto-play.

[Xilmi]

Brutal-OXCE 4.1.1 - Omniscient AI Enhancement Patch

Fixed an issue where enemy thought it could attack around a corner in melee-range when it couldn't because it had no melee-weapon.

Units in sweep-mode will once again be allowed to hide if they don't have enough TUs to attack anymore.

Fixed an issue that could cause AI to skip the rest of their turn instead of taking cover.

Fixed that AI would think it just needs to turn towards their target despite not having a line of fire.

Omniscient AI no longer assumes that it will be able to attack everything it has a line of fire to. This could lead to them standing out of position when the player used smoke-grenades.

AI now will assume that it will be able to see a unit when it is at minimum range to it.

The peaking-logic now works the same for omniscient and non-cheating-AI.

Fixed an issue with omniscient AI not being omniscient when it comes to knowing whether an enemy was out of position.