[SOURCEMOD] Brutal-OXCE 7.12.1

[anonymous]

Will you update this to OXCE 7.9.6? I want to run it with X-piratez.

[Xilmi]

Will you update this to OXCE 7.9.6? I want to run it with X-piratez.

Version 6.3.9 is up to date with the latest OXCE 7.9.7 from June 12th 2023.

Whenever I release an update I merge with the latest OXCE. Incase I didn't update for a longer time and OXCE got updated, just contact me to do an update.

[eagles980]

Does the AI always avoid mines or (if sufficiently beefy) can they purposefully walk through them in order to clear the way for the rest of the team when the player leaves them no alternate path to move forward like in a single entrance base defense?

Love getting destroyed by the AI so far. lol

[Xilmi]

Does the AI always avoid mines or (if sufficiently beefy) can they purposefully walk through them in order to clear the way for the rest of the team when the player leaves them no alternate path to move forward like in a single entrance base defense?

Currently it will always avoid walking through mines.
I agree that this is not ideal in all situations. Especially not the way it currently is realized.
The avoiding of the mines isn't done in the path-finding but instead as an afterthought. This means if the shortest path to a destination passes a mine, it will never go to the destination even if it could walk around the mine. By handling that in the pathfinding, it could walk around mines, which would be better.

Also I'd say that depending on what they want to do they may be okay with detonating mines if they know they would survive. For example if they want to walk to a spot they can attack from or if they want to go to a spot from where they think they could spot a unit.
However, for this they'd also have to consider whether the mine would hurt their allies if they detonated it.

Would you say that their current reluctance to trigger mines is highly exploitable and it's worth to spend a lot of effort? From my point of view, with the way it currently works, they can be situationally useful to keep aliens off your back and deal with them later. But you can't really generate a massive advantage with them. I still might want to have a look of whether there's low-hanging-fruits in regards to how my AI deals with them.

Also nice to hear that you love getting destroyed by it. I'd always appreciate video-footage of that.

[eagles980]

Well I haven't gone far enough in the game to ecounter a proper base defense as of yet (purposefully avoiding downing UFOs) but did do a brief test of precisely what the AI does when it encounters mines in the mission generator's base defense. Beginner difficulty, Floaters, minimum tech and inherit unit aggression turned on. Turns out the floaters have no problem running into them to their deaths as if they weren't there kind of like base AI with the mission playing out pretty much the same as it does normally: Mine and hide (smoke / doors) until one of the aliens sets it off, reset as needed with occasional instances where multiple enemies would break through and get shots off on my guys waiting inside the doors closest to the elevator while (it did take full advantage of squadsight in these cases naturally.)  The was one moment when one purposefully walked around a mine that was misplaced by one tile but I don't know if that's a coincidence based on where they spawned or what.

All in all I think it would be brilliant if you could find a way to attach some neat pathfinding logic to the AI's consideration of mines but it doesn't seem to be an absolutely critical issue so far (at least when it comes to base game) that can be exploited any more than it was in vanilla. Maybe it's different with a mod like XCF with a hundred+ enemies running around the map with only a few ways to get to you. They might be able to be manipulable into taking the exact path you want them to.

No plans to record any of my gameplay for now but you for sure aren't the only one interested in seeing seeing a lot more footage than is available. I need as many tips as possible considering how many painful losses I'm taking. I'm only heartless towards soldiers with garbage stats unfortunately.

[Abyss]

All in all I think it would be brilliant if you could find a way to attach some neat pathfinding logic to the AI's consideration of mines but it doesn't seem to be an absolutely critical issue so far (at least when it comes to base game) that can be exploited any more than it was in vanilla.

Next turn, after evading mines, enemies should toss primed grenades back to thrower the same turn. If some TUs are saved for that case. Because, AI should win anyhow. That was sarcasm and joke.

Nerfing human win mechanics one by one, what purely stays? A bunch of player-controlled naked guys with low stats that barely can hit when trained soldier hits with 100% probability.
What is this game becomes about? About loosing.

Each hq mission, the player squad can be butchered on turn 2.
Each field mission, player squad can be overtossed with grenades on turn 2
Each mission is smokescreen and pray.

Have you ever played major mods, Xilmi?
And who are these guys, who comment only your branch, with brand new nicknames?

If you nerf win mechanics, give something in exchange, player won't use the product which does not satisfy his needs of victory.

Really, 2000 missions average to beat the major mod. What is it about?

[panzer]

Each mission is smokescreen and pray.

Everything is like in a real battle Seriously though, just set "Targeting behaviour for Brutal AI" to 1 and the enemies will become significantly weaker.

[panzer]

I have a PC and a laptop. The game runs well on the PC. On the laptop, the processor is only 2.7 GHz. With the latest updates, the aliens started to take much longer to move. In version 6.3.9, the enemy takes almost twice as long to make a move compared to 6.3.0.

[eagles980]

 

Next turn, after evading mines, enemies should toss primed grenades back to thrower the same turn. If some TUs are saved for that case. Because, AI should win anyhow. That was sarcasm and joke.

Heaven forbid one be of the opinion that the AI might be better off using a bit of nuanced strategy dealing with your already impressive toolset and status as a thinking human. That's the equivalent of wanting it to utterly (and unfairly, most importantly) dominate the player! If you want to convince people you're joking, maybe don't inject so much accusatory bitterness in your post especially in the context of your next response.

And who are these guys, who comment only your branch, with brand new nicknames?

I'm me. Howdy. What, lurkers can't post for the first time and then respond to a reply of said post without being viewed without suspicion? You're a strange individual. I'm in agreement with panzer. Turn down your settings and / or the difficulty until the game is fun again and you'll be fine.

[Xilmi]

Well I haven't gone far enough in the game to ecounter a proper base defense as of yet (purposefully avoiding downing UFOs)

With the modified "Aggressive Retaliation"-option winning missions against landed UFOs should also trigger base-defense. It also fixes that they simply wouldn't find your base in certain locations due to how they were searching for it. They got a new searching-pattern, if you enable that option.

Turns out the floaters have no problem running into them to their deaths as if they weren't there

Hmm... I definitely have to investigate it then. This is not supposed to happen. One explanation would be that inherit aggression could lead them to go into sweep-mode. So depends on what their base-aggression is. I prefer using balanced aggression for these tests. As that's what I do almost all of my tweaking on.

[Xilmi]

I have a PC and a laptop. The game runs well on the PC. On the laptop, the processor is only 2.7 GHz. With the latest updates, the aliens started to take much longer to move. In version 6.3.9, the enemy takes almost twice as long to make a move compared to 6.3.0.

I'm not sure which change between 6.3.0 and 6.3.9 could have caused that.
Actually since 6.3.5 it should be faster, not slower.

The new more costly mechanism of finding cover was introduced in 6.0.0. So an increase in enemy-turn-time between 5.2.2 and 6.0.0 could be very well explained with that.
6.3.0 itself with considering more possibilities from where to attack would also be expected to cause a slowdown.
But if something made it slower between 6.3.0 and 6.3.9, I don't know what it is yet and have to still figure it out. Especially if the turn-time outright doubled.

I tested it with the same save to compare:
6.3.0 ~ 6 seconds
6.3.9 (run from Visual studio) ~ 7.5 seconds

So it is slower, not by that big of a margin but measurable in this scenario.

Okay, I think it is this: There's a difference in the peaking-behavior. In 6.3.0 the units repeatedly do indirect peaking, which was not intentional. This way they have less time-units for hiding, which means the algorithm for finding a good hiding spot consideres fewer tiles which means it's cheaper.

I have an idea for a low-end mode where I use a much cheaper algorithm for hiding-spots. It will also obviously be much worse but should significantly speed it up on low-end-hardware.

[eagles980]

With the modified "Aggressive Retaliation"-option winning missions against landed UFOs should also trigger base-defense. It also fixes that they simply wouldn't find your base in certain locations due to how they were searching for it. They got a new searching-pattern, if you enable that option.

Hell no! I like my guy's insides exactly where they are, thank you very much. Unlike certain others I have no problem lowering the recommended difficulty settings for myself to make the game more fun. That fix does genuinely sound to be a good thing but I haven't necessarily had many issues with bases not being discovered if I've taken appropriate risks in triggering the missions.

Hmm... I definitely have to investigate it then. This is not supposed to happen. One explanation would be that inherit aggression could lead them to go into sweep-mode. So depends on what their base-aggression is. I prefer using balanced aggression for these tests. As that's what I do almost all of my tweaking on.

Yeah, pretty much confirmed this pattern of behavior on superhuman as well in order to check the terrorist's behavior (none spawned earlier) and evaluate the danger of blaster bombers with squadsight. As in the first test the AI became far more conservative the fewer in number there were to the point they stopped charging and I had to hunt them in the hanger / flush out some of the positions near the elevator they were presumably using to peek and definitely using to camp. Managed to avoid convincing the enemy that launching their bombs would be a good idea as I keep the hallway pretty well smoked throughout with the troops hiding within the doors usually. Overall ended the mission with 6 deaths and 2 drone scrappings, some of which were caused by human error. Wish I'd have thought to have taken some screenshots of the more interesting moments, though.
in any case, if this behavior wasn't intended then I can see why introducing dedicated pathing logic for mines might be a good idea as it'll solve multiple problems at once on top of making things more interesting. To give a nod to alledged nerfing concerns certain others have brought up in the case of mods like XCF, it would certainly make more powerful and heavier mines capable of killing and incapacitating tanky units viable again as the AI wouldn't want to trigger them if at all possible if serious damage or death were likely unlike the less lethal cousins. Food for thought. ...You can take your sweet time getting around to adjusting the AI's behavior on inherited aggression for now, though. I like winning fairly convincingly. lmfao.

[Xilmi]

I tried reproducing the Aliens triggering the mines.
I used debug-mode to place mines like shown in the attached screenshot.
On the first alien turn none of them were triggered and aliens respoected them.
However, on the second turn, one Floater ran into one of the mines after he had killed a soldier of mine.

So I'll now try to reproduce it in order to look at the debug-log.

Note that in the case of this screenshot, there's nothing that pathfinding could help them with as there's no way to walk around any of these mines anyways.

Edit:

Okay, I've analyzed it. It is a pathfinding-issue afterall!
For their planning my AI uses the Dijkstra-pathfinding but for the actual movement it seems like it uses bresenhamPath and if this doesn't work A*.
So there's basically a discrepancy. If there's different ways to walk to the same target, for example: "north-east, north-east, east" and "east, north-east, north-east" then there's a slight difference between plan and execution. A difference, that may very well mean life or death depending on whether there's a mine on one path but not the other.
The very same underlying cause could also make the difference between walking through reaction-fire or not.

So my goal is now to figure out how I can make the AI to actually use the same path it had planned and not make a new one that may or may not be different from that.

Edit 2:

No idea why it first attempts bresenham and only uses A* when that fails. A* and Dijkstra seem very similar and in my test also have the same result. So by skipping the attempt of bresenham the Alien no longer triggers the mine. Now putting the mine-check into the path-finding-itself, maybe in a similar way to how Sneaky-AI works should help to make it find paths around mines, when possible.

[eagles980]

Okay, I've analyzed it. It is a pathfinding-issue afterall!
For their planning my AI uses the Dijkstra-pathfinding but for the actual movement it seems like it uses bresenhamPath and if this doesn't work A*.
So there's basically a discrepancy. If there's different ways to walk to the same target, for example: "north-east, north-east, east" and "east, north-east, north-east" then there's a slight difference between plan and execution. A difference, that may very well mean life or death depending on whether there's a mine on one path but not the other.
The very same underlying cause could also make the difference between walking through reaction-fire or not.

So my goal is now to figure out how I can make the AI to actually use the same path it had planned and not make a new one that may or may not be different from that.

Edit 2:

No idea why it first attempts bresenham and only uses A* when that fails. A* and Dijkstra seem very similar and in my test also have the same result. So by skipping the attempt of bresenham the Alien no longer triggers the mine. Now putting the mine-check into the path-finding-itself, maybe in a similar way to how Sneaky-AI works should help to make it find paths around mines, when possible.

Thanks for the explanation! The rabbit hole goes even deeper than I thought. Happy to know I tipped you off on this issue. if you make any more progress on implementing a fix I'd be happy to hear it.

Just to confirm, the intended behavior you meant for the AI is to stay back behind any mines you drop and wait for you to approach instead, launching bombs if it "hears" or sees clusters of your troops in the interim?

[Xilmi]

Yes, that is the intended behavior.

I've already fixed it for the next version.

I also implemented a 4th option for aggressiveness between balanced and sweep-mode, that is supposed to be a compromise for faster processing without being completely reckless either.

While testing I saw a peeking-bug, which caused units to walk back and forth thinking they could uncover an enemy when they couldn't and thus wasting a lot of TUs. Not really related to any of the changes but I'll try to fix it too for the next version.