OpenXcom Forum
Modding => Released Mods => The X-Com Files => Topic started by: Xilmi on May 15, 2023, 12:44:59 pm
-
Over the past week a Twitch-streamer called "Speedislife" was streaming X-Com Files with Brutal-AI on Superhuman Ironman.
It was a really fun journey to watch until he decided to restart in August 1998. After that first restart he did it over and over whenever something remotely bad happened and almost every time I tuned in it was a new run. I think he now stopped doing it at all.
His reasoning for the first restart was that he was too far behind in progress of what he thought he was supposed to have by that point. As someone with extremely limited experiece with XCF I cannot really judge it.
I personally wish he had played it out until the bad things he expected to happen actually would happen.
The last two things that happened in the run was that he did a Manor, which he won with 4 out of 16 people still alive at the end, 3 of which badly wounded. Really close and thus exciting to watch.
After that a UFO was shot down in the military and he already anounced that he would be too weak to even attempt it and did so anyways. He played uncharacteristically sloppy and got wiped. Almost looked like he wanted to provoke this to happen to make a point.
One thing I noticed being quite different to vanilla-X-Com is the huge discrepancy between Rookies and people that have been in service for a while. It seems like the stats go up way more quickly and also reach significantly higher values. So having to replace 20 veterans with rookies is a massive reduction in overall combat-strength. Another big difference was that the ratio of health to damage was so that most hits resulted in injuries rather than death, meaning that agents often could be saved. And of course that a lot of missions were quantity over quality. Basically tons of weak enemy units, sometimes a few decent ones sprinkled in.
During the time he streamed it, I also made 2 AI-modifications to things I noticed. I think the prior absence of these made the first attempt quite a bit easier compared to the later ones. It was about the behavior of enemies with short-range-weapons, of which there are plenty. They used to use the same peaking-logic as those with long-range-weapons often resulting in dying to reaction-fire with no way to fight back. After the modifications they'd only peak towards locations where they wouldn't expect the enemy yet. This means you'd have to come closer to them and get into their effective range.
I'm not sure whether the mission against the aliens was representative of what to expect a few more months down the line. If so, I can understand it. It looked similar to how my TFTD-attempts against Brutal-AI go. What I also wonder is how much more powerful he could theoretically have gotten before 1999.
I learned a lot about how to approach this mod and should probably give it a serious try too. You need to obtain a lot of knowledge of when and with what kind of gear/stats what kinds of missions are doable. The toughest part I imagine is investing a lot of time into the run before noticing it's actually not salvageable. But that also probably makes it more emotionally engaging.
-
I'm curious if other folks have any insight into using Brutal AI with X-Com Files; I've heard the siren-song of restarting my own run using it a few times. I opted not to on my last restart because I was still pretty rusty at classic X-COM and because XCF in particular seems like a drastic enough overhaul I'd be concerned about how missions set up for regular AI would run with BAI. It is cool that Xilmi got a chance to tweak it from watching someone play though. (I'm not surprised you had to tweak the guys using peashooters and golf clubs).
As to the mod in general, a few of my own thoughts:
I actually managed to lose my first run thanks to low score, on whatever the second difficulty was called. I'm not sure exactly why; I was very trigger-happy with the Abort Mission button to preserve my agents. I assume I had enough Strange Life Forms missions and maybe a couple of the early harder types where enough civvies died, plus maybe despawn penalties from missions I couldn't reach to tip me over.
There's also the fact that the game can get pretty darn monotonous on the tactical level - if for whatever reason you decide to fall into a holding pattern (research, agent training, waiting for funding), you'll find yourself doing a zillion Cultist Apprehension or Cult Safehouse or whatever missions. This can make seeing the bad end in sight more emotionally draining than engaging.
The absolute bonkers gun porn also doesn't help me with my analysis paralysis; a lot of guns are very samey or clearly outclassed and you have to really dig into the Stats for Nerds screen to see that some can be pretty bonkers or have crippling weaknesses (particularly with regards to armor efficacy). On the other hand I do like that this means there's some randomness to what you outfit yourself with if you luck into one of the good ones. Either my second or third game I got the Winchester 1901 shotgun in a very early mission so I basically didn't have to worry about non-standard weapons while I built out my research and Logistics.
Sort-of back to what the OP was talking about: I think simultaneous the most unintuitive and interesting/fun things about XCF are the absolutely sprawling research tree and its affect on game progression. As I kind of alluded to up top, I restarted a couple of times either because I thought I had dead-ended my research or because I researched too much too fast and was getting harder missions without the firepower (or manpower) to deal with them. Once you start to get a grasp of it you can find that it's actually fairly lenient, I think, as long as you pay close attention to what 'Affects game progression'. I do feel like figuring out where your labs, armor, and transports are on the research tree is pretty critical though.
Finally, as far as veteran agents go - it's absolutely the case. Actual rookies are kind of a joke, which is why you kinda want to spend some time rotating folks out in the very early game (and grinding Reactions on those ubiquitous melee missions), and always have people warming the bench (press) getting ready to replace losses. That and having enhancements online early (like the Red Dawn one). Heavy use of dogs and AI Units (once you find them on the tech tree) also helps preserve those vets.
-
XCF on Brutal AI works in like, 95% of cases. Certain mission types, particularly cult HQ missions that cap off those storylines, have over 100 enemies. When the AI is smart enough to remember where the player is and actually push into them, it's basically impossible to get out of the spawn rooms.
Playing Superhuman on XCF is pure ego, IMO. Adding extra enemies to the missions does very little to fill them out since they're already pretty populated already. And XCF is very forgiving of failure, since past a point you stop having to worry about money, and you can definitely play money-efficient if you know what you're doing before that point. You can 100% play it out in almost every circumstance; the penalties for not completing cults "on time" are minimal, and even with really good RNG and strategic-layer play, you weren't going to be prepared for the impending alien invasion.
-
Is there a record of these streams somewhere?
-
The last two things that happened in the run was that he did a Manor, which he won with 4 out of 16 people still alive at the end, 3 of which badly wounded. Really close and thus exciting to watch.
Most of the fight looks like what you would usually expect, except that at the end he loses troops left and right. Looks either under-prepared or B-AI is much better at endspiel.
His reasoning for the first restart was that he was too far behind in progress of what he thought he was supposed to have by that point.
Usually you'll have promotion 3 (defeating one of the cults), and weaponry outlined below. In normal game it's not required for battling aliens per se, but I guess you'll need more options against B-AI.
I'm not sure whether the mission against the aliens was representative of what to expect a few more months down the line.
I'm assuming you're talking about that Downed UFO after manor?
Even with Normal-AI, you can expect to be shredded by aliens if you just try to outshoot them early. With right gear\tactics however, you can complete Downed UFO semi-flawlessly as early as 1997.
That being said, getting inside that UFO with 5 B-AI sectoids seemed really tough, period. Have to either sacrifice a few assaults, or fill the inside of UFO with incendiaries - they go through the shield.
What I also wonder is how much more powerful he could theoretically have gotten before 1999.
Normally, by 1999 you'd get craft for 12-16 troops, increasing amount of possible sacrifices (:.
Weaponry wise - something like a basic vanilla gear - e.g. rockets, cannons. PROXIMITY GRENADES. Better grenade launcher. Maybe lasers, with good RNG. Against sectoids, melee and grenades will remain one of best options.
Armor - there's better ballistic armor. Also you'll get personal armor or slightly better. It gives OK chance to survive plasma blasts, while being knocked down and on fire. Thus requiring squaddie with extinguisher to spend his turn rescuing the victim, possibly just standing there like a dolt.
Early 99, with a bit of luck, you can loot a couple of power suits from specific tough enemies, which give pretty good protection, but will go away if you're not careful. To manufacture good armor you'll need quite a few alien engineers = surviving quite a few UFOs.
It seems like the stats go up way more quickly and also reach significantly higher values.
Stats increase in same way they do in vanilla, but there are many easy missions to boost them. This is the main difference, as you have good chance to bounce back if you lost your A-team. Regarding how caps (https://xcf.trigramreactor.net/master/article/STR_SOLDIER) are different I dunno. There's also many researchable one-time trainings that increase stats down the line.
Another big difference was that the ratio of health to damage was so that most hits resulted in injuries rather than death, meaning that agents often could be saved.
Yes, common firearms rarely one-shot agents, especially those in armored vests. There's still grenades, rockets, melee critters and gimmicky units though.
So having to replace 20 veterans with rookies is a massive reduction in overall combat-strength.
It depends on equipment. E.g., most players prefer to play around snipers. Sniper rifles have long range and cannon-level damage but very quickly dwindle in accuracy bellow 100 firing, thus require veterans to use. But there are ways to make most of the rookies.
Playing Superhuman on XCF is pure ego, IMO. Adding extra enemies to the missions does very little to fill them out since they're already pretty populated already.
It increases enemy stats, too. Not that much compared to veteran, but still.
And XCF is very forgiving of failure, since past a point you stop having to worry about money, and you can definitely play money-efficient if you know what you're doing before that point. You can 100% play it out in almost every circumstance; the penalties for not completing cults "on time" are minimal, and even with really good RNG and strategic-layer play, you weren't going to be prepared for the impending alien invasion.
Yep, fully agree. That wasn't always the case, though...
-
I wound up deciding to do a little test-run with Brutal AI and the short-range peeking behavior seems very tedious and sometimes.. well, beyond brutal. Melee units and your average pea-shooter cultist will hunker down in one tile, necessitating a full sweep with two agents (and often going to the 'bug hunt' mode phase of the fight). I've considered flipping on Aggressive mode, at least for the early game, though I can definitely see this behavior being interesting when both sides have more than two under-geared units.
It's nice that melee units don't just kind of wander back and forth waiting to die, but unless it's an open desert it just turns into a jump-scare slaughterhouse as you crawl up two steps to maintain overwatch like you're playing XCOM2012. (It'd be nice if units had greater than 90 degree FOV, too, though that might benefit the enemies more)
edit:
And apparently stunned enemies immediately get their TUs back if they recover??? Was shocked to have an agent die right after tasing a dude and ducking back to cover, whirled around with the other guy and saw the guy with the axe back on his feet.
-
And apparently stunned enemies immediately get their TUs back if they recover???
Yeah, the default value of 'tuRecoveryWakeUpNewTurn' is 100. I use 50 myself.
-
Is there a record of these streams somewhere?
Twitch normally stores all the streams as VODs for a couple of weeks at least.
So you can find them here: https://www.twitch.tv/speedislife/videos
And apparently stunned enemies immediately get their TUs back if they recover??? Was shocked to have an agent die right after tasing a dude and ducking back to cover, whirled around with the other guy and saw the guy with the axe back on his feet.
It's not about the TUs. The mechanics are the same as when your own troops wake up. They also wake up with full TUs. The issue was that the base-AI deliberately skips their first turn after waking up. Of course the brutal-AI sees no reason to limit itself in such arbitrary way.
X-Com-Files actually has a really useful counter to that: Handcuffs! Normally they are kinda unnecessary. But now they actually are a really helpful tool in your agent's arsenal to prevent the exact scenario you described.
-
X-Com-Files actually has a really useful counter to that: Handcuffs! Normally they are kinda unnecessary. But now they actually are a really helpful tool in your agent's arsenal to prevent the exact scenario you described.
Cuffs cost 40 TUs to apply, so you'll need a teammate to do it. So you won't prevent this exact scenario - tase somebody, get into cover with 0 TUs, and get fried by stunned hostile the next turn.
You'll prevent other scenario where some guy pretends KOd and takes a jump at you. Wow, it's actually lore-friendly (:.
But yeah, cuffs kinda seem more useful now.
-
My experience with recovery from stun is that there is 50% chance that just recovered from stun enemy unit has their TU (and can react) and 50% chance that he has no TU (and cannot react). Approximately the same is with my own units: sometimes they have TU, sometimes not. The difference is that just recovered enemy unit NEVER walk, but my units can do that if they happen to recover with TUs. Initially I thought that it depends on when exactly a unit recovered from stum: on your turn or on enemy turn. But that does not explain why enemy units never walk after recovering.
-
My experience with recovery from stun is that there is 50% chance that just recovered from stun enemy unit has their TU (and can react) and 50% chance that he has no TU (and cannot react). Approximately the same is with my own units: sometimes they have TU, sometimes not. The difference is that just recovered enemy unit NEVER walk, but my units can do that if they happen to recover with TUs. Initially I thought that it depends on when exactly a unit recovered from stum: on your turn or on enemy turn. But that does not explain why enemy units never walk after recovering.
The explanation is in the code:
if (_AIActionCounter >= 2 || !unit->reselectAllowed() || (unit->getTurnsSinceStunned() == 0 && !unit->isBrutal())) //stun check for restoring OXC behavior that AI does not attack after waking up even having full TU
-
Interesting to see that's how stun recovery works - I've so seldom had my own units stunned I didn't realize it was different between AI and player.
That said, my experience after having played some more is that handcuffs are just way too TU intensive - holster (or if you're still using Kevlar and shotguns, drop) your weapon, pull out handcuffs, spend 40 TU to apply handcuffs. If the agent has very high TUs you might be able to do that all in one turn. Also for the privilege of being able to do all that, you lose out on a whole grenade's worth of inventory capacity.
That said, a much better solution... Just tase (or beanbag or whatever) the guy again. There's no hit indicator, but it absolutely will apply more stun damage (same trick with lethal ammo is also good for dealing with zombies that got stunned instead of killed). Unfortunately it doesn't work with melee weapons.
Also it's neither here nor there, but I find it amusing that handcuffs can put a unit into overstun and start killing them.
---
As for Brutal AI in general, I've noticed that there's a lot of slowdown on the player turn on more populated maps (I'm early enough this pretty much means certain Strange Life Forms missions and a Citizens vs. Monsters map). In particular, the CvM map was buttery-smooth while the megascorpions rushed the desert fighters, but when it came to my turn I had hefty framedrops whenever my units would move. Most of what I've seen mentioned about slowdown was on the AI turn so I figured I'd mention it.
-
...I guess making the cuffs faster is an acceptable option.
-
...I guess making the cuffs faster is an acceptable option.
Not enough! Because...
handcuffs can put a unit into overstun and start killing them.
And that is why right now handcuffs do more harm then good. Not only handcuffs not doing it's primary job (you still have to babysit downed enemies), but they also kill! And since you probably use them on important prisoners, it is even more frustrating.
Also, if you insist on enemies can free themselves, then at least provide a notification about that if happened.
-
No, handcuffs don't kill with overstun, unless target dies from existing overstun on first turn after KO. Rather, it seems that overstun is cleared by handcuffs.
-
That does sort of match my experience - it'll display the Overstun symbol, but it's always gone on the next turn (sometimes I get frustrated with an agent missing a prone figure one tile in front of them, so I do still use them occasionally).
As for buffing handcuffs, my two cents - as I just said the 'just shoot 'em again' strat is a bit risky - you absolutely can miss and I think agents 'aim' at a standing (or perhaps kneeling) target rather than the ground (I haven't done enough to have a good guess to the behavior). And if you have to shoot more than once, you've likely spent about as much as you would have inventory shuffling the handcuffs. If that's the deciding factor about cuffs needing a buff, it's not as fool-proof as it might seem.
That said I do still think for how low-utility they are in the base game they're way too cumbersome over just bringing another flashbang or KO grenade. To go through the whole cycle of putting away weapons to getting weapons back out takes between 53 and 76(!) TUs depending on exactly how you juggle the agent's inventory, plus if you're using, say, a taser or pistol and electric club, you've got to reserve 2 of your 3 2x1 slots for both of those.
Seems to me that dropping the base cost by 10-15 TU would work well; agents carrying one weapon spend about as much as the base cost, while agents using both hands are probably using 70-80% of their TU (assuming they're close-ish to the training cap).
---
And on the subject of Brutal AI again, I've encountered Shadowbats and they both take an absurd amount of time to process and are acting strangely (closing to my agents and then just wandering around in front of them to get punched - though it seems like if they start their turn close enough they do bite). I dunno if this is typical for flying enemies, something to do with them also being melee-only, or just me playing without the forced aggression option. I've attached a couple of saves for the mission in case they might help.
-
As for Brutal AI in general, I've noticed that there's a lot of slowdown on the player turn on more populated maps (I'm early enough this pretty much means certain Strange Life Forms missions and a Citizens vs. Monsters map). In particular, the CvM map was buttery-smooth while the megascorpions rushed the desert fighters, but when it came to my turn I had hefty framedrops whenever my units would move. Most of what I've seen mentioned about slowdown was on the AI turn so I figured I'd mention it.
I think the most likely cause of that is that I'm caching the tiles in the unit's line of fire into all directions rather than just the visible tiles into the direction the unit is looking at. And I just realized that I don't even use these anymore!
I think it was probably part of my old cover-seeking-logic. I guess I can take it out again to help with that.
-
And on the subject of Brutal AI again, I've encountered Shadowbats and they both take an absurd amount of time to process and are acting strangely (closing to my agents and then just wandering around in front of them to get punched - though it seems like if they start their turn close enough they do bite). I dunno if this is typical for flying enemies, something to do with them also being melee-only, or just me playing without the forced aggression option. I've attached a couple of saves for the mission in case they might help.
Thanks for the saves. I'll look at them later. Flying-units with lots of TUs always have put some strain on my AI due to the vastly increased of possile options they look at each turn. Having the maps be like 9 tiles in height does not exactly help in that regard. Wandering around to get punched definitiely sounds wrong and something I should do something about. I recently fixed something that could be related to that.
-
I looked at it. The vast majority of the additional time they take from the tile-visibility-checking. Disabling it reduced the time for "thinking" for them from 6 to 1 seconds.
Being able to fly adds just so many more tiles to what is being checked.
I could just assume that they'd be visible if they are more then 1 tile above the floor without doing the actual checking.
As for the going back and forth near the units I still need to analyze. I suspect that all melee-units could have this issue but it's more prominent for the ones who's attack costs significantly more TUs compared to moving around.
-
Hey everyone!
I would like to take part in the discussion!).
Xilmi, this particular message is written for you, as I somehow believe you haven't played the XCF mod.
That far I can explain my experience in XCF as 3 months in total split within last 8 years, 6 playthroughs I guess. XCF is my favorite game ever, and I am a bit nerdy about it.
4 of them were made on superhuman and they were, at my perspective, easier than the same playthoughs on lower levels of difficulties. I will explain why below.
I would say that playing superhuman is often gets you to better results progress-wise.
1. More enemies = huge stat boost to crew at the end of the mission. Thus, doing next mission is easier.
2. More enemies = more score!! I have always gotten better score results on difficulty 5, even with worse outcomes. Like, more civilians dead etc. And more score = faster progress (more funding, more good random events).
3. More enemies = more arms on the battlefield. When it comes to the middle-game - more enemies = more loot. Like, you always have a chance that MiB powersuit guy gets unconscious, you save him and rip the powersuit off at the end.
4. More enemies = you have a better chance to capture the specific pawn alive you are doing this mission for. Early game, each cult mission is about capturing someone alive.
5. More enemies = stronger feel of tension and fun, but as we already had it discussed with Xilmi, quite subjective experience.
Also, I am the one who abused the mechanics like: camping, flashlight on/off, getting all possible stuff from MiB, getting all possible from promotions, getting alien alloys from UFO as early as possible etc.
Mechanics of the mod are quite simple, yet brilliantly polished for player to play on superhuman vs vanilla AI.
Mission-wise it is polished the way you should do next tier mission each month.
Eg. Apprehension in month one (get the cultist), safehouse month two (get lvl 2 cultist), .... base month 4-6 (get one more cultist, hahaha), HQ month 6-9 (kill them all). Otherwise, you are already beneath the progress and will likely fail getting control over the invasion.
There are three (or four, depends on how you count) layers of global enemies:
1. Four cults - earlygame (Church, Exalt, Red Dawn, Ninjas) to deal at month 8-9, at least, with one. Each having 5 steps, mean you not only have to do the mission correct, but also do the research of the cultist prior of the month, otherwise next mission will not spawn in the next month.
2. Advanced organizations with little different mechanics (Syndicate, MiB, Hybrids, Cyberguys), that give you bonuses (Syndicate = alien alloy munitions unlock, Cyborgs = advanced protection etc.) to deal with, at least by some middle point in year 2. While some tier 2 organizations are straight-forward in the progress of the story, others are random-based and need a lot of research.
3. Tier 3 - like lunar Nazis or terror from the deep aliens are sometimes quite stronger than aliens themselves.
4. Aliens with random race.
When it comes to mid game (after alien invasion), strict timing gets less punishing and the game fits to your research progress (most missions are bound to researches you make, except terror, random and storyline missions)
About soldiers cap: With Ironman/superhuman playthroughs I had most like 100-150 soldiers trained by the mid year 2, with like 100-150-200 dead. This is, somehow, the soft-hard cap, because human resources cost funds, not just for buying, but also for keeping them.
Not to tell, that each replacement for experienced unit costs a lot of the time (20-40 missions + at least half a year of training at base).
When getting to alien invasion, my troops usually getting vanished in increased manner because there is the gap between alien firepower and player's troops armor. Usually, hit from plasma weapon = death of the soldier. It is the game rules. You always get incremental.
Now, when you buy newbie soldiers and equip them same as veterans, they are on 10% of effectiveness of veterans with the same armor and equipment (shooting accuracy + moving speed + whatever). Dealing with 10000-15000 enemies by year 3 with 200-300 soldiers is what XCF is about. Please keep this in mind. Also, please keep in mind the craft design: it is almost always open-back until mid game, while even low-tier enemies possess incendiary grenades/dynamite/whatever.
With Lack of the resources forever until year 4-5, losing 50 veterans is gameover. And it is quite punishing experience for the player who spent real-time hours, because player loses it's investments (many many psychology here). Also, losing the craft is sometimes gameover too.
The missions are 95% times incremental over your current progress. Like I mentioned it previously, you will almost never get the mission where enemies are weaker than your crew.. And if you do get that type of mission - you are either on agent training (stat boost, or just score-loss prevention) of you are done and failed (close to score loss and restart).
Of course, there are certain types of missions, where your agents going in while being absolutely naked, like Syndicate HQ, facing 250 TU minotaurs and super-soldiers each of which worth 3 of your agents, Minigun-robots and a couple of bosses, of course. So, you have no armor at all and there's no chances to beat such mission with AI anyhow more advanced than vanilla AI.
That is why I believe that giving to AI some fun tweaks is not "arbitrary", it just matter of integration of the AI into major mods. The total experience with your AI over global mods should be even more entertaining (fun, strong, tense, non-ordinary, difficult but beatable), and I strongly believe it is achievable. The game is made for entertaining and joy, not versa!
-
The purpose of my AI is to maximize kills and minimize losses with the tools it is given.
There are ample ways to reduce difficulty. Like disabling squad-sight, forcing the AI to rush, lowering the difficulty-level and even turning it off completely.
If you want the AI to behave in specific ways that you imagine to be more fun to play against, which are neither provided by the base-AI nor my AI, you'll have to code it yourself.