Zombie AI. End goal from fun pimps?

[Maje]

The fun pimps is trying to kill us, and that's okay. However, the way they're doing it is subpar. In A17, the fun pimps did the following.

 

-Increased repair time

-Decreased resource gain

-Zombies attack weakest blocks as one (group think)

-Increased zombie Damage

-Reduced block durability (downgrading gone)

-Removed ability to remove limbs.

-Zombies jump longer

-Zombies Dig

 

The result being the vast majority of bases are now untenable without extensive knowledge of how zombies path to a player. I no longer create bases for hordes, I'm forced to create zombie playgrounds to manipulate their AI.

 

I know the AI is not done yet, I get it... but I want to know the goal of the fun pimps regarding AI. A mission statement. Something like "In an ideal world the AI will do....."

 

With all the above changes, the game is no longer fun for me. I know some people don't share my sentiment, so I suggest while AI is being developed, give players an option in the settings to have the old AI back.

[Maje]

And before I get a bunch of people in this thread suggesting this is a L2P issue and I should simply "git gud". I know exactly how the pathing of the current AI works, and I can create 100% horde proof bases. I just don't find it fun.

 

 

This is NOT a video thread, please do not move it to the video section. These videos simply illustrate the horrible zombie pathing.

[Epicurius]

And before I get a bunch of people in this thread suggesting this is a L2P issue and I should simply "git gud". I know exactly how the pathing of the current AI works, and I can create 100% horde proof bases. I just don't find it fun.

 

 

This is NOT a video thread, please do not move it to the video section. These videos simply illustrate the horrible zombie pathing.

 

Horrible. Zombie. Pathing.

 

I'm really not sure what you are saying here. I have been playing since A8 or so and I am here to tell you that the zed pathing is GREAT. Finally TFP has made this game the 'Gritty Survival Game' that was promoted and not 'Minecraft' with better graphics and zombies.

 

I'm loving the digging zeds. I totally missed them since they took them out in A11 or so.

I'm loving the busting and breaking defenses to kill you. Excitement is had every horde night.

I'm loving the pathing through POIs that are better than my own pathing.

 

Horde nights are now 'How can I stay alive?' verses 'How many kills can I get with my Iron Reinforced Club.

 

I'm really not sure why people play a game to only try to 'trick' the AI verses play the 'Gritty Survival Game' and roleplay what it would be like to be in this world.

 

TFP can't ever win for losing it seems.

[Jackelmyer]

In fairness, zombies really shouldn't cluster up on a single block. It's very unzombie like and kind of boring too. Just kinda hang out with a shot gun over head and hit'em all.

 

IMO horde nights haven't been improved much at all by A17. I made my platform base have a double thick wall. *shrug* Though I have died like 8 times (several due to stupid performance issues/1second 0fps lag spikes). So my horde nights may be nerfed a bit at day 28.

[Cirion]

In fairness, zombies really shouldn't cluster up on a single block. It's very unzombie like and kind of boring too. Just kinda hang out with a shot gun over head and hit'em all.

 

IMO horde nights haven't been improved much at all by A17. I made my platform base have a double thick wall. *shrug* Though I have died like 8 times (several due to stupid performance issues/1second 0fps lag spikes). So my horde nights may be nerfed a bit at day 28.

 

agreed... Luckily I think the zombie AI is still under work.

 

There can and should be advantages to making big sprawling bases - spread out the damage. Right now, though, even with a big sprawling base, without AI cheese generally, you'll still get the whole horde ganging up on one block, which means even on 200% block hp, it takes mere seconds for them to break through - even with several layers of solid steel.

[Cadamier]

agreed... Luckily I think the zombie AI is still under work.

 

There can and should be advantages to making big sprawling bases - spread out the damage. Right now, though, even with a big sprawling base, without AI cheese generally, you'll still get the whole horde ganging up on one block, which means even on 200% block hp, it takes mere seconds for them to break through - even with several layers of solid steel.

 

Yup, yup yup... Even those on FB are complaining.

 

They should go back to pre-A17, but I think the poll on them pushed it to the 'post-a17' zeds...

[Jackelmyer]

Yup, yup yup... Even those on FB are complaining.

 

They should go back to pre-A17, but I think the poll on them pushed it to the 'post-a17' zeds...

 

I don't have the same experience as you two. I play purely vanilla settings, except don't drop toolbelt I think. The default difficulty. Think the default zombies count of 8 or 10. Zombie's take a while to get through my upgraded flagstone blocks.

 

If 5+ are hitting one block, yeah, it can go quick. But I usually still have enough time to get overhead with a shotgun and drop the group in seconds.

 

I just wish the damn zombies didn't path right around spikes. That.... Is super annoying. But... Also inconsistent. Sometimes they won't path to gaps they've made in my spikes. I dunno. It's all weird.

[Cirion]

On default settings you probably won't notice the pain too much.

 

But try x64 blood moon hordes. Also, once you get to high gamestage, even on default settings, you'll get wrecked, so it's not just a matter of turning down difficulty. Turning down difficulty merely increases the time before you start getting wrecked.

 

Yeah you will rarely get more than 5 zombies hitting a block on default settings at low gamestage for sure. Once you get 20+ hitting one block, most of which are green zombies, then let's talk

 

If you don't already know, eventually you'll get 64+ zombies per BM, even with "8 zombies" set, so it's not a matter of reducing the setting lower, you'll still eventually get wrecked, as I mention.

 

All the x64 setting does is make earlier BM's a little trickier.

[fragtzack]

Horrible. Zombie. Pathing.

 

I'm really not sure what you are saying here. I have been playing since A8 or so and I am here to tell you that the zed pathing is GREAT. Finally TFP has made this game the 'Gritty Survival Game' that was promoted and not 'Minecraft' with better graphics and zombies.

 

I'm loving the digging zeds. I totally missed them since they took them out in A11 or so.

I'm loving the busting and breaking defenses to kill you. Excitement is had every horde night.

I'm loving the pathing through POIs that are better than my own pathing.

 

Horde nights are now 'How can I stay alive?' verses 'How many kills can I get with my Iron Reinforced Club.

 

I'm really not sure why people play a game to only try to 'trick' the AI verses play the 'Gritty Survival Game' and roleplay what it would be like to be in this world.

 

TFP can't ever win for losing it seems.

 

Agree, AI is a vast improvement in a17. Very well said on that last sentence about win for losing.

 

Sure, AI could use some improvement.

Just like everyone on the real earth could use improvements also.

 

but..

 

At some point, you have to accept your wife/girlfriend as she is with all her beauty and flaws.

 

Machines and software are made by humans, machines and software all have flaws too. If we humans spend every minute trying to remove every flaw in our life, life would be miserable and get nothing meaningful accomplished.

[Jackelmyer]

On default settings you probably won't notice the pain too much.

 

But try x64 blood moon hordes. Also, once you get to high gamestage, even on default settings, you'll get wrecked, so it's not just a matter of turning down difficulty. Turning down difficulty merely increases the time before you start getting wrecked.

 

Yeah you will rarely get more than 5 zombies hitting a block on default settings at low gamestage for sure. Once you get 20+ hitting one block, most of which are green zombies, then let's talk

 

 

No that's totally fair. Though... okay... So, totally not being a smart ass. But why aren't you turning down the number of zombies? x64 I assume is 64 zombies on blood moon? ... w... I don't understand. That... to me... sounds like... you shouldn't be running a setting like that without either setting difficulty WAY down low and/or ensure you're with a group of 3+ players. And... bases would need to be build waaay different. Like... 4 rows of electric fence wire zombies have to run through in order to just reach your base. So you all have time to bring down some of them.

 

X64 doesn't seem like just upping the stakes. Seems like you're requiring things to be done way differently. Like... Base to Zombie Horde... "NO TOUCHY!".

[Cirion]

No that's totally fair. Though... okay... So, totally not being a smart ass. But why aren't you turning down the number of zombies? x64 I assume is 64 zombies on blood moon? ... w... I don't understand. That... to me... sounds like... you shouldn't be running a setting like that without either setting difficulty WAY down low and/or ensure you're with a group of 3+ players. And... bases would need to be build waaay different. Like... 4 rows of electric fence wire zombies have to run through in order to just reach your base. So you all have time to bring down some of them.

 

X64 doesn't seem like just upping the stakes. Seems like you're requiring things to be done way differently. Like... Base to Zombie Horde... "NO TOUCHY!".

 

Maybe let me try to rephrase, and give a little "story" as background to maybe help my point along.

 

I'm an engineer and do a fair bit of coding professionally. When I do a new code, if its intent is for a large scale high computational solution, I first start small, and work my way up. Once I have validated it for small scale application, I then test it for large scale. More often than not, I find it works just fine for a small scale application, but blows up / fails / otherwise doesn't work right in the first large scale test.

 

My point of this story - Large scale applications tend to "expose" the problems. And that's what I see here.

 

Adding a lot of zombies exposes the "flaw" of zombie AI far more easily than having not many zombies. Which you basically say in your post. It's not that bad when you only have a few zombies ganging up on one block, on an easy difficulty. The flaw in the AI is not readily noted, because as you say, you have no problem dealing with it. You would only notice flaws in the AI at this point, if the flaw was even more severe (Like a full on glitch, being able to walk through walls or something like that... I know, extreme example, but hopefully you get my point). That doesn't mean the current AI is OKAY, that just means you can deal with it fine, so it SEEMS ok. But increase the scale -> problem more readily exposes itself.

 

There certainly are workarounds that I have found that work to an extent, when you realize how the AI works, but that doesn't mean it's balanced by any stretch of the imagination considering zombies should be DUMB.

[Cirion]

you shouldn't be running a setting like that without either setting difficulty WAY down low and/or ensure you're with a group of 3+ players. And... bases would need to be build waaay different. Like... 4 rows of electric fence wire zombies have to run through in order to just reach your base. So you all have time to bring down some of them.

 

X64 doesn't seem like just upping the stakes. Seems like you're requiring things to be done way differently. Like... Base to Zombie Horde... "NO TOUCHY!".

 

never said X64 shouldn't be harder or even require different strategies than X8 or whatever and I absolutely agree that adding more traps like fences can absolutely be very helpful on bigger hordes and that adding more players can and should make it easier to do.

 

All I am saying is that zombies should not gang up on one single block. That's it. That's the crux of the debate. This is regardless of default settings or hard mode settings.

 

I don't necessarily have an opinion one way or another for the new pathing otherwise (seems still a little weird for zombies to be smart enough for that, but I will say it has improved combat in POI's since now they path to you instead of destroying blocks... sometimes anyway.) So I'm not saying necessarily remove that pathing (but the fact they avoid traps is annoying), just them having cheat vision / group think to gang up on singular blocks.

[Khulkhuum]

Zombie pathing still needs a bit of tweaking, but i wouldn't call it subpar. Take into consideration the following:

1. Compare zombies walking/running straight at you, destroying most of things in their path (unless they can spare a second to walk around it) to the current version where they tend to find the best route towards you.

2. Standing on an overhang and looking down at a horde that focuses in one spot, compared to a horde seeking a route through the building the overhang is built upon (if no direct route present, then they mash supports).

3. Enemies running in circles, trying to decide the direction they should be going, compared to mobs picking up specific behaviours and if focused on going after you, they do that.

 

Sure, you may not like the current system (for any number of reasons), but IMHO it is far more complicated and will lead to a far better resolution when tweaked to perfection than the A15 or A16 ones.

 

BTW, great videos. One such tweak i have in mind is some additional verification to pathing vertically above ground - just like in your example, when you sidestep to a close by platform, where they should know closest would be to jump to you after climbing the first section.

[Jackelmyer]

Maybe let me try to rephrase, and give a little "story" as background to maybe help my point along.

 

I'm an engineer and do a fair bit of coding professionally. When I do a new code, if its intent is for a large scale high computational solution, I first start small, and work my way up. Once I have validated it for small scale application, I then test it for large scale. More often than not, I find it works just fine for a small scale application, but blows up / fails / otherwise doesn't work right in the first large scale test.

 

My point of this story - Large scale applications tend to "expose" the problems. And that's what I see here.

 

Adding a lot of zombies exposes the "flaw" of zombie AI far more easily than having not many zombies. Which you basically say in your post. It's not that bad when you only have a few zombies ganging up on one block, on an easy difficulty. The flaw in the AI is not readily noted, because as you say, you have no problem dealing with it. You would only notice flaws in the AI at this point, if the flaw was even more severe (Like a full on glitch, being able to walk through walls or something like that... I know, extreme example, but hopefully you get my point). That doesn't mean the current AI is OKAY, that just means you can deal with it fine, so it SEEMS ok. But increase the scale -> problem more readily exposes itself.

 

There certainly are workarounds that I have found that work to an extent, when you realize how the AI works, but that doesn't mean it's balanced by any stretch of the imagination considering zombies should be DUMB.

 

 

So I totally see what you're saying. Though... I think there's an alternate view to this story.

 

Where you could say that scaling up will inherently show the flaws, I do agree with this. To a point.

 

What I often find in scaling up anything, is that "how" you accomplish the same goal is VERY different. And, to a degree, that's, IMO, a real natural design. And I mean like, reality, physics, nature type of natural. Lemme splain.

 

When you take that code, find it works great small scale and ramp it up, yes, you can definitely find flaws and enhancements that would improve both the small scale and large scale applications of the same code. However...

 

And this applies to technical solutions as well (think fix width format data storage flat file compared to full SQL database cluster on disk arrays, etc).

 

But a more simplified way of looking at scaling and how one process just may not fit multiple scales is irrigation. Just moving water where you need it to go. Simple right?

 

Well... for Average Joe sitting on his... I dunno... 20 acre land. He can dig some ditches, run a large ditch to a river near by, and connect the two. And... roughly... he's good to go.

 

Now take the industrial farms running 100's of acres. their water flow can't simply be managed by a simple set of ditches and tapping into the river. Volume of water will wash away near river areas or water will never reach the distant areas at all because there's not enough flow and water volume.

 

 

For Joe, the amount of land he's watering has a much lower water volume requirement. So he can run a water flow through a ditch dug into soil and it doesn't overly disturb the irrigation ditches overall by much.

 

However the industrial sized facility has to concrete their ditches. Run pumps to traverse the distances. Setup a flow control system to manage the heavier water pressure and overall flow.

 

I kind of see the same thing with zombies.

 

Sure. With 10 zombies on horde night and the default difficulty, Joe Farmer and I are having a hay day and are living the life.

 

You and Mr. Industrial fancy pants however... can't really expect to do things the same. And with so many zombies, you absolutely would need flow control mechanisms. Literally... Like water flow.

 

- - - Updated - - -

 

never said X64 shouldn't be harder or even require different strategies than X8 or whatever and I absolutely agree that adding more traps like fences can absolutely be very helpful on bigger hordes and that adding more players can and should make it easier to do.

 

All I am saying is that zombies should not gang up on one single block. That's it. That's the crux of the debate. This is regardless of default settings or hard mode settings.

 

I don't necessarily have an opinion one way or another for the new pathing otherwise (seems still a little weird for zombies to be smart enough for that, but I will say it has improved combat in POI's since now they path to you instead of destroying blocks... sometimes anyway.) So I'm not saying necessarily remove that pathing (but the fact they avoid traps is annoying), just them having cheat vision / group think to gang up on singular blocks.

 

I think I agree with the single block gang up. I'm not as worried about them blowing through walls. But more would like to see them actually flail against the walls spread out a bit and a bit more zombie horde'ish.

 

- - - Updated - - -

 

I would... like to see more zombie animations though. I don't like that they only run, hop over, or destroy things in their path. There should be some kind of crawling animation for crawling over 1 block or two block high objects. Like crawling over the hood of a car as opposed to beating on it till it explodes.

[Cirion]

Zombie pathing still needs a bit of tweaking, but i wouldn't call it subpar. Take into consideration the following:

1. Compare zombies walking/running straight at you, destroying most of things in their path (unless they can spare a second to walk around it) to the current version where they tend to find the best route towards you.

2. Standing on an overhang and looking down at a horde that focuses in one spot, compared to a horde seeking a route through the building the overhang is built upon (if no direct route present, then they mash supports).

3. Enemies running in circles, trying to decide the direction they should be going, compared to mobs picking up specific behaviours and if focused on going after you, they do that.

 

Sure, you may not like the current system (for any number of reasons), but IMHO it is far more complicated and will lead to a far better resolution when tweaked to perfection than the A15 or A16 ones.

 

BTW, great videos. One such tweak i have in mind is some additional verification to pathing vertically above ground - just like in your example, when you sidestep to a close by platform, where they should know closest would be to jump to you after climbing the first section.

 

Again in case my point(s) get misinterpreted, I don't hate *everything* about the AI.

 

The main things they changed are

 

1.) Zombies can find shortest path to you

2.) Zombies can and does avoid traps, oftentimes to the extent they would rather bust through solid steel, than walk through a clear open path (with traps) to get to you

3.) Zombies will gang up on ONE block instead of spread out damage

 

#1 is unrealistic, but I can get on board with it, because you need some level of unrealism to increase fun.

 

#2 is unrealistic, and should be removed as feature entirely.

 

#3 is what I hate a lot also from the new system.

 

So I like #1, but hate #2 and #3.

 

I'm guessing #3 happened because of #1.

 

There are multiple ways to fix it. One way would be to make taking "open paths" a bigger priority for zombies than it is now, and to not avoid traps. Now you can have a clear open path (even without traps) and still zombies will often times ignore this opening, to bust through the wall behind you instead, even if said wall is a solid layer of steel. How is this the "easiest" path to get to you?

 

I have found you can work around this system if you build a wall thick enough, but if your base is very large, this solution is quite impractical as it can require up to 10+ layers of walls. So again, I have already found solutions to this problem, and can work with it, but I still think it's not an ideal system.

 

They often pick the "shortest" instead of the "easiest" path, there's a distinct difference. IMO, the "easiest" path should be used - which is the path of least resistance - or least amount of blocks between you and the zombie... (not counting trap blocks like spikes), with "air" blocks not counting towards this calculation (Like it is currently).

 

Currently they take the shortest block distance, which includes air blocks, trap blocks, and wall blocks, but I argue that neither trap blocks nor air blocks should count in this calculation.

 

Through experimentation, I DO think air blocks are weighted slightly less than wall blocks, but not as logical as I would think. A small building, you can get by just double walling, and they'll ignore them if you have an opening.

 

But I have tried triple, quadruple walling a large structure, and they still bust through all 4 layers of wall rather than take an opening in a game with a friend of mine. I keep adding more layers and they still won't take my opening.

 

after more reflection, I dunno how you can have #1 without #3. Even picking "easiest" path in some cases may result in them targeting only one block...

 

At the very least, as long as they can make it so they pick "path of least player built blocks" to you, I'd be happy enough. (As long as "trap" and "air" blocks don't count in the calculation)

 

edit - one more thought I had

 

I am pretty sure that the Quality of material is NOT used in the calculation of "path of least resistance". I double walled a simply wood-structure and it didn't seem to matter if it was wood or steel, just that it was double walled.

[SnakeWildlife]

The way zombies are right now we need to remodel the entire zombie thing out of the game, and have lots of Morty's running around.

 

 

And we're the Ricks trying to get rid of annoying Morty's.

 

We'll rename the game to "7 Days till the Neutrino Bomb goes off"

 

These just arent zombies.

[badmojo]

The zombies also need to learn how to better attack blocks in the direction of the player. For the past 3 or 4 horde nights we spent the entire night on the second floor of a large power station building, with me literally hanging off a staircase block meleeing the horde just below me. The zombies seem to ignore the blocks above their heads and just run in circles below us. They did manage to destroy the stair block I sit on, twice, but I suspect it was more of an accident than intentional. The cops are the exception to this, but their damage can usually be repaired and them focused and killed before they do any serious damage.

 

It seems silly to be able to melee a zombie and not be in any danger of being hit or having the horde target the block (well within their range) that I am camped on.

 

We have been building a horde night base since the start but it usually gets pushed to the back burner so even though we are on day 70 we have yet to actually finish and use it. Instead we used a small house for the first two nights and then for the last 8 we have used the power station, the first 5 blood moons we just sat on the roof and waited for morning, but the last 3 we have been going down to the second floor to farm more xp. Gamestage is probably around 300 by now, on default settings.

 

Thing is, I feel like the zombies will never bring down the power station building entirely. And the zombies will likely never build themselves a way to climb up it. Therefore the whole 'building a base' thing becomes entirely unneeded. We will still do it, because building can be fun, but its depressing that with all the intelligence these zombies have they still just run in circles moaning and groaning when you do something as intuitive as removing the bottom two blocks of a staircase in a building. How was this not tested and addressed?

 

What I would suggest is to make zombies that cant pathfind to the player attack any block they can reach in the players direction, but only blocks above their waist which they cant pathfind to. This would result in zombies jumping to the highest surface and attacking the blocks above, it would also cause them to make their own vertical pathways, as soon as they clear a block above their waist they jump up and repeat the process.

 

With something like the above, walls suddenly become avenues for zombies to build their own staircases to reach you. That new pathing would also probably help them more effectively attack big buildings, because currently all I ever see is two, maybe three vertical blocks being removed, and thats it, anything higher than 3 blocks from the ground seems to only be removed by spitting cops or zombies jumping on each others heads and accidentally hitting higher up.

 

Tldr; make zeds climb and attack upwards.

 

Here is what it looks like from where I have been sitting...

[apostrophe]

What if zombies could climb up another zombie's back, like a ladder?

[RipClaw]

What if zombies could climb up another zombie's back, like a ladder?

 

They're already doing that. I have already seen how 3 zombies stood on top of each other and that is as far as I know also a feature of Alpha 17. They can't necessarily do it in perfection yet but they are zombies and not circus artists.

[apostrophe]

They're already doing that. I have already seen how 3 zombies stood on top of each other and that is as far as I know also a feature of Alpha 17. They can't necessarily do it in perfection yet but they are zombies and not circus artists.

 

Of course.

[meganoth]

I know the AI is not done yet, I get it... but I want to know the goal of the fun pimps regarding AI. A mission statement. Something like "In an ideal world the AI will do....."

 

You probably won't get such a statement. What the AI-developer Fataal has been saying is that eventually he wants them to attack from all sides and that they will be dumbed down again when the AI is good enough to also work for bandits.

 

I think he also said that for A17.2 block damage bonus of zombies hitting on the same block was changed.