(1.0)(WIP) Adaptive Cassie Storyteller - Fine tuning

Started by Wanderer_joins, September 07, 2019, 07:28:22 AM

Previous topic - Next topic

Pangaea

Cheers, glad it is helpful. I like to engage with the trade and events of the world, which is why in that save I edited the scenario slightly to add more quests and trade ships. Certainly dangerous to go on the road, especially when you get raids at home with half the people there, but we thankfully managed.

Actually thought the setup I had was a trap maze, but being quite new to the game I'm not all that up to date on all the terms and various exploits/whatever-people-want-to-call-them. After a search I see there are more cheesy approaches, such as this one. I'm always too late to build proper defenses in the early game because I like to take on raiders in the open behind sandbags or chunks. Then we almost get wiped.

Here are a few other saves that may be useful: https://filebin.net/w6bzucj1nf5nh2y2

This is the first mountain base I have played (that survived past the first year). The two saves is from when I triggered the 15-day final assault, and just after we had somehow managed to survive the non-stop raids. You may spot a corpse or two around the map ;D Attacked from hordes of sappers from every direction, sometimes several raids simultaneously, and we were very close to dying. Maybe not getting wiped, but certainly getting in very dire straits when they broke into the base proper, or were very close to doing so.

Although I don't particularly like killboxes, that game in particular taught me that it's required. No way to deal with 200+ raiders unless you have a way to slow them down and attack them with turret support.

5 people died in that save, plus Gillespie was brought back from a vegetable state as she was downed from a one-shot by a lancer that brought her brain to 1/10, leaving her unable to move. Until we, much later, were able to get a healer mech serum and fix her. Don't recall the details very closely now, but I reckon at least a few of those deaths were chased refugees and suchlike that turned out crap and were downed or outright shot to death. Only saw one sarcophagus when loading the save.

Very similar approach between the saves tbh, but I like these semi-open bases with active defenses. I'm therefore intrigued to try out a different story teller to see how that plays out.

As an aside, I see upon loading these saves that item wealth has gone up considerably towards the end. I suppose that is dropped weapons and such from the huge raiding parties? We certainly weren't able to do anything in those 15 days than sleep, heal, and try to repair the defenses.

Wanderer_joins

Right? You need killboxes at some point in the end game, and cheesy tactics like trap mazes artificially reduce the threat scale.

Actually you can play merciless without killboxes, but at the cost of losses, heavy micro and a combat oriented playthrough.
I like merciless for the mood management and other challenges, but the AI constantly pushing makes it really hardcore without cheesy tactics like traps or chokepoints.

That's why adaptive cassie is intended to keep these challenges and to give you significant breathing room if you accept losses.

I'll analyse the additional files this w-e too, thank you.

Pangaea

I was looking through page upon page of mods to look for inspiration, and came upon the Cassandra Hardcore mod, which is also made by you. It has some very neat graphs. If it's not too much work, how does Adaptive Cassie compare with the classic storytellers, say Cassandra Rough and Savage? Would be interesting to see such a comparative graph :)

Don't think I'm good enough to even play on Merciless, never mind without any way to funnel the probably gigantic raids  :( Do you mean without any traps at all? I agree it's kinda cheesy since the enemies can't see them and we basically exploit pathfinding, but I don't really see how we would be able to handle 25-30 manhunting bears or rhinos if they get in our face either :-/ Or 150-200 tribal raiders without any way to thin out the pack a little bit before they all swarm us. I learnt the hard way that it can be a recipe for disaster, even when we have superior guns and armour.

I look forward to what you conclude from analysing those saves. Having looked about the place more, you look very invested into storyteller mods, and I want to try it out to see how the games adapt. I have to admit it gets kinda boring in the end game when I'm done teching, the base is "done", and it's mostly about waiting for raids, and get the ship built. More excitement in this phase would be good. And if the game ever gets more updates, I hope they manage to spice up the end game in some fashion.

Wanderer_joins

The key to playing high difficulty is wealth control. I end up ship built around 250-300k wealth. You also need to develop slowly and invest in your colonists to turn them into bionic soldiers with good equipment:
graphs
graphs

Flat map boreal, taking advantage of marsh and body of water:
base example

Cassandra hardcore pushes this strategy to the extreme, and don't get me wrong, i praise Tynan for keeping the end game challenging.
But in order to play open maps with high wealth and caravaning, without being hardcore and cheesy, the storyteller has to be more adaptive. It's harder to make a simple graph for adaptive cassie since it's based on population / time, wealth points are changed etc.. but i will make a model based on your pop graph and time on recruitment and compare it to your data point. I want to check whether it doesn't get too difficult (ie much harder than vanilla) over time in the best case scenario (0 deaths).

Pangaea

Thank you. That was interesting reading, and viewing of graphs. I dislike the "need" for wealth control, although I understand it is required to succeed on those levels. Seems impossible to me to get to that stage with only 250,000-ish wealth. Typically I'm at 7-800k, and upwards of 1 million if we survive the 15-day terror (probably due to heaps of dropped weapons, and maybe apparel on the corpses, if that counts).

In my current game I made the best attempt so far at making everybody bionic super-soldiers, but it simply took too long for all the advanced components with nothing else really going on, so I decided to trigger the ship instead. That was with 25 people and around 800k wealth. I tend to craft apparel for trade, but usually don't have heaps laying about, so not sure what accounts for almost 3 times as much wealth.

But think I'd still prefer to stay at a lower difficulty (although Rough or Savage isn't exactly easy) and largely ignore wealth control. The game is a lot of fun up to the base is basically done. Then much dead time, and then fun and great challenge when the ship onslaught is triggered. In the OP you write "wealth is less the main driver of raids size and increasing the number of colonists will actually helps you defend your colony", which sounds like it would fit my playstyle quite well.

You have presumably checked out the save file, but here are a few pictures for others. In fairness I do have heaps of resources here, especially steel, becasue it became a big issue in the last game I played despite having 7000 or so when I triggered the end attacks. Probably rather a lot of wealth just there  :-X

Base main
Base killbox (sorry)
Graph (many caravans and mining trips)
160+ pop tribal raid (corridor bought us time to sort-of get into position)
Nasty mech ship

Wanderer_joins

There's a large overlap between difficulties. The psychic ship in your screenshot is worth my merciless ships, and beyond.

The different level are mostly relevant early to mid game, then it depends how you handle wealth and deaths.

Merciless is a flat 40% modifier, here your wealth is 100% higher than in my examples and each of your colonists weight 80% more than mines since their value is tied to wealth too. You end up dealing with significant harder threats on savage than mines on merciless.

Plus you didn't die, so adaptation must be high and savage weight adaptation more than merciless, paradox of lower difficulty.

Pangaea

#21
Ooof, that is downright scary. This game is "only" on Rough  :o

We survived the onslaught, but 4 raids in a little over 1 day was hard. Two mech raids and a pirate raid blew up everything. All turrets were gone or otherwise out of commission. Half the colony hospitalised. We had to run away from the killzone to avoid getting killed there. Later a double-whammy pirate drop pod attack + siege. I had a double-walled ammo depot, and 1-2 pods dropped on top of it. Didn't see it because attention was elsewhere at the time, but it somehow blew up the whole thing, including many of the double-walls (not just one layer), plus much of the undamaged granite mountain (900hp). There was an antigrain warhead in there too... Glorious hit from the one that was thankfully already in a mortar, mind you :) Then the buggers shoot not one, but two doomsday rockets in my face. Thankfully the people downed from the first one somehow survived. Thank you marine armour + shield belt.

Had a save from when that ship in our fields touched down, so I tried to up the difficulty to merciless and pop it. Does that "take", so the points are increased to merciless, or is the ship content generated earlier? In any case, more centipedes came out (15 vs 8 earlier), but actually fewer numbers combined. About 40 (tried a couple times) vs 47 in the original Rough game.

I'm a slow player that almost always play at 1x, so games take time (and I get very attached to them, hence the probably too detailed coverage above for our purpose here), but I'll fire up a new one now and use your Adaptive Cassie storyteller. I'd like to try without a channel+trap corridor, but since I don't like to (ab)use wealth control, it scares me.. The raid sizes in the end-game are colossal, and from my meagre experience it's much worse on Savage vs Rough. This is surely not a linear effect (re: overlap in your post above), but according to the XML files, Rough has a threatscale of 1.0, Savage 1.55, and Merciless 2.2.

Wanderer_joins

Sure, it's linear ( the 40% is the jump from savage 1.55 to merciless 2.2). But when you play at higher difficulties, it's harder to make silver, you've more deaths, and adaptation weight less, all factors which tend to reduce the threats.

For example you can have a look at the " fun points" on my graphs (adaptation), they are almost halved at the start of one of the ship sequences. I had one or two deaths, then the storyteller gave me a 50% discount on raids. If you try hard not to lose anyone during the ship defense (or save scum) it's much harder.

You can have a look at these graphs in devmod, debug graph. It's a strong incentive to have one or two rookie fighters in the frontline, ready to die for a purple heart...

Once again, in my view you're fighting merciless level raids. The next step will not be combat but mood management.

But with this discussion i think the soft cap of adaptive cassie should be vanilla.

Pangaea

Very interesting. Looked at a save from after the onslaught with dev mode on, and here is the screenshot. We took heavy damages and lost many limbs, but no people. The peak is at slightly over 1100 "fun points". About 920 right at the end, when the hell-ish onslaught was over.



Damage taken (from Statistics) probably includes buildings and so forth as well, and probably enemies too, because it increased too much to be only vs (my) pawns. About 420,000 at the end, and 275,000 when the reactor was turned on.

I shall fire up a new save with your storyteller, maybe on Savage, and hope I can survive long enough to get an impression of how it compares with previous Cassie Savage attempts.

Wanderer_joins

Ok, so i've finally opened the save back home. It's working as intended, the threats with adaptive cassie in your end game are slightly lower than vanilla.

Here are the relevance of your pawns (purple)


As a comparison, in green, a merciless playthrough combat oriented, with an end game around vanilla's level.

Most of your pawns are relevant for my storyteller.

Two outliers:
Vebo the most experienced after 11 years, by a long shot.
Linnea the least experienced after 8.7 years, did you try to protect her from the frontline?

Sofia got her experience real quick.

Pangaea

Quote from: Wanderer_joins on October 06, 2019, 10:25:07 AM
Two outliers:
Vebo the most experienced after 11 years, by a long shot.
Linnea the least experienced after 8.7 years, did you try to protect her from the frontline?

Sofia got her experience real quick.
Wow, that's interesting and unexpected. Especially when I look at this when I loaded the save with the Numbers mod active:



Loads more must go into the experience than kills, because Vebo is in the middle of the pack, and Breixo is first by a fair distance.

Unfortunately I can't recall how I used the pawns in this game, but I'd say that typically everybody would be drafted. Does caravans play a semi-important part perhaps? Linnea being a Night Owl, she probably never went on a caravan or mining trip (or rarely), and spent much of her time mining or doctoring. Or perhaps she rarely if ever happened to man a mortar, which in some cases would rack up many kills, probably important for this hidden experience. If a raid happened and the night owls were tired or in bed, I would typically wait until the last moment before drafting them, or maybe even let them sleep (depending on what we faced). If we could attack with mortars I'd do that with 6-8 people, and then draft the rest when we had to.

I also sent out many caravans and mining trips, some of those faced fierce battle. Perhaps Linnea never went on these, while Sofia did? I honestly don't remember. Didn't particularly try to protect Linnea, but perhaps she happened to mostly be in a position where she didn't deal much damage.

Hmm, I actually remember how Vebo was used now, since I see he has a mini-gun. He was usually placed in a vulnerable position in the killbox, so he would be in decent range with the mini-gun. Meant he took quite some incoming hits and was downed a fair bit, but probably means he dealt a lot of damage against centipedes and whatnot that came into the killbox too.

This is the killbox I had towards the end of the game. Vebo would almost always be in the green circle. Melees behind mini-turrets trying to keep them alive.


Hmm. Looks like Linnea may have used a sniper rifle. That could mean she was placed north of the auto cannons in the above image. Perhaps she just sucked with it and didn't hit all that much, and rarely took incoming damage, being further away from the action.

Of course there was a lot of fighting without using the killbox too, especially at the end of the game, but those were chaotic as fuck, where I generally used whoever was close enough to intervene. Sappers caused me no ends of problems.

Do brawlers/melees do poorly in this metric? Against mechs for instance it was often way too dangerous to send them into battle (20+ centipedes sometimes, like this lot), so they would often stand back a bit and try to repair mini-turrets. Most likely meant they took a lot of hits and damage, but not necessarily dealing out much themselves.

Cool to hear the onslaught might have been a little less severe with your storyteller. It was horribly brutal, especially in this Savage game.

Have started a new game with Adaptive Cassie on Savage. Have only played about two seasons/quadrums, but I think it was noticeably less violent in the start. Only one raid in the first season. Will be fun to see how things develop.

Wanderer_joins

OK, i think they're just flukes and within the random distribution. The main factor is time. Melee is factored in as well and your melee fighters are doing well.
I could add more stats in the metric, but i'm not sure how it would impact the load of the cpu, so i'd rather keep it "simple".

Adaptive Cassie has the same rythm as Cassandra Classic, so you should only feel the difference on the threat scale (size of the threats). Notably when guys are out caravning.

Pangaea

#27
Darnit, this is sad  :'( Always a huuuuge worry when Cassandra doesn't send anything for a while. About a season with no action. Then it happens, and it's three groups of mechs. "Only" 6 of them in fairness, but that is hard when we don't have proper weapons and no armour, nor any traps and only a few sandbags for cover.

Opened up a door in the west because most went there, and we managed to take down two scythers. But apparently they broke through in the east as well. Smashed up a wind turbine before we could get over. A few more of them go down. But a lancer is hiding behind a solar panel. Truth with the EMP grenades from a previous raid is hiding behind a tree and throws a nade while other people are handling two other mechs. BAM. Truth is murdered. Sign of the times...

Very hard blow for the colony as he was our best cook and only artist. The rest of us got cooked too. Two others almost died to lancer shots in the chest, and if it weren't for the pelvis and sternum being indestructible, they would have been blown to dust as well.

I've put the poor chap in the freezer, and will build two sarcophagi there, and hope we can get a resurrection serum down the line.

Popped into dev mode to see what happened to "fun points". Kind of a dip allright. Hopefully Cassandra won't throw death on such levels at us for a while. [edit: Err, guess not. Few days later: Siege!]
Can you tell us anything about what these points/scale actually mean?





Even without the friendly fire he would have been killed (by a hair). Can barely see the exclamation mark of another person downed further south. Shot in the torso.


Pangaea

#28
Figured maybe a save using this storyteller from somebody other than yourself could be useful?

Naturally this is still quite early game. Huge drop-off in recovery, whatever those "days" mean, after the Truth murder + siege. Currently we have a bunch of prisoners from several groups of sappers. Unusually, a lot of them survived and were only downed.



No traps or killbox yet, only hiding behind sandbags, which is bound to bite us in the arse sooner or later. We shall see.

Managed to get the save just under 1MB by using 7zip compression and attach it here. Zip or tar.gz was too big.

Pangaea

#29
Slightly touched on the pig incident elsewhere. Coupled with other stuff going very badly, I don't think we could have survived this on the default Cassandra Savage storyteller (knock on wood, we seem sort-of back on our feet now). She probably would have sent something worse after us sooner.



The 22 manhunting pigs very, very nearly did us in. Practically everybody down. Only Mandrill and the slow as heck Onesan was sort of up, and she's incapable of doctoring. Thankfully a few got up, and I managed to self-tend some of their wounds in the field before they dropped again at 45% blood loss (it lowers consciousness). The lousy and injured doctors had to do what they could, being the only people up, and I had to micro hard with loads of people 1-3 hours from dying. Because they were wading in blood and guts in the hospital, we got hit with 4 infections. Used all the 13 (heh, lucky number) normal meds we had acquired from rare traders, the last 3 on infections. Some got lousy 20-ish percent tends, but that can happen with so bad doctors (skill 0-1, and 4), with them being injured on top.

Mie was worst hit with the infection. But before I could do anything about it, a sapper raid hit us. There was absolutely nothing I could do about it. Most of the colony was in hospital, either out cold or fiercely struggling against serious infections. Could only put the various animals where they ended up going through into the inner base, and hope for the best. Thank goodness I remembered to uninstall and remove the infinite chemfuel gadget!

Several animals died from getting fragged, including the pig -- plus a great prisoner (Fox). They all got exploded into nothingness. But we eventually managed to make them flee and didn't lose any of the two people able to fight them. The cows are pretty darn good, and one of them got very seriously wounded. I used a pawn to fix it up as it was only hours from dying. Then I noticed the infection on Mie's leg had got to 90%. Camlaban, badly injured, was in the process of removing it, but was too slow. Mie died on the operating table with knives poking her infected leg. Despite being a good doctor, Camlaban only had 58% surgery success chance due to all the wounds, so he may have failed anyway.

Suddenly it dawned on me that both Truth and Mie were doctors, and we're now down to only Camlaban being capable. But on the upside, the three surviving prisoners have now been recruited, so we're up to 10 pawns. The mortar and 35-ish shells got exploded by the sappers, and looking at the pink dots, we're probably due another raid soon. But at least we have flak armour now, so hopefully we'll be able to stand up better in gunfights. I'm concerned about the basically total lack of defenses, but that giant drop in recovery and fun points probably meant we're still alive -- so there is that :)

Oh, and one of the pigs from earlier was out cold forever and had 94% malnutrition, we managed to rescue the girl, and she is now our new pet pig. Can't haul yet, but pigs are easy to train up.