A special thanks to friend of the channel Alan Zucconi for sharing with me the link to the original blog post by Matthew Davis. It's nice to have friends who share game development blog posts with you. 😄
Apart from balance, randomly picking one of the best options also makes the AI less predictable - if players can engineer optimal conditions to trick the AI into doing what they want 100% of the time, it becomes too easy to cheese the game.
@@timbus2 just wanna put this out there that if you open the console and type “undoturn” into the console it works just like the one-time reset but you can use it as many times as possible.
Exactly, see early xcom 2. A sectoid will raise a zombie, a phantom will make a clone etc, you can manipulate the non-damaging enemy actions to survive a turn.
One of the greatest bits of advice (which I think I've learned from this channel) is that the opponent AI does not need to be competent in the game; it just needs to be fun and engaging.
At university I made a pacman clone with custom levels. I didn't know much about implementing A* or even Pacman-like pathfinding and just made the ghosts move randomly every time they came to an intersection or corner. That, along with the fact I had to run pathfinding twice per frame because for some reason running it once led to ghosts ignoring the walls meant the "AI" was actually quite fun to play against, and my tutor thought it was way more complex than in reality :^)
If I remember correctly, a lot of FTLs popularity came from the fact that it was a Kickstarter that did insanely well, and every news outlet reported on that and then the game. Meanwhile, into the reach is just another release from Subset games. It didn't have that same extra wave of publicity.
I've been working on a grid based puzzle game lately, and the video feels perfectly timed. Even though the video is AI-focused, the concept can apply well to lots of different systems and gameplay decisions. Weirdly enough, I made this realisation through Zachtronics, who's games are known to be quite complex to play, but some of his puzzles like Dungeons & Diagrams can be entirely played on paper. With the game I'm creating, I'm very much always reeling it back in to "good enough", and the gain is that I get to keep things simple.
@@darkhobo Unfortunately not! I’m a software engineer professionally and I had to put it aside to get away from code & the computer at some point during the day. I left it in a place where I can pick it up again though by leaving lots of code documentation & my design docs.
Into the Breach is just an absolute masterclass in all departments it seems... such an elegantly simple game. Feel like basically 99% nowadays turn-based-strategy games could take a note or two.
every task you ever try to do is, or was at some point a puzzle :P Even things that now come with instructions, at one point someone had to figure it out.
I've always felt Into the Breach was best described as "Kaiju Chess." It definitely always felt like a puzzle game to me. A puzzle game with a lot of variables, but a puzzle game nonetheless.
@@michaelcalvin42 Smart of your part, I guess to a certain extend I was using puzzle solving skills but from a more reactionary place. I'm thinking of approaching the game again but with this kind of mindset.
lol I figured that out after playing Into the Breach and having to pause my podcast so I could 'think" about how to win a level, and suddenly I realized I never had to do that during FTL. In FTL it was always just instinctual reaction, but Into the breach makes me stare at a screen in silence for 10 minutes.
Balancing is something that "the doc" and I often talk about. Basically, how good is too good? If a player cannot win, that's gonna suck. But it also kinda sucks if the AI is obviously doing dumb moves. I feel like there's a whole research field on making really good "bad" AI
Yeah it's something you don't really hear discussed in academic circles, and I suspect there's a lot of 'dark arts' hidden throughout the industry on how to do this.
Smart AI is not the same as having a difficult game. You can have an easy game with smart AI, or a hard game with stupid AI. Generally you should just make the AI fun to play against and balance the difficulty through other means.
i'd love to see a video on rain world's AI on this channel, while not the most complex nor the most simple, the interactions it has with the player, the environment and other creatures' AI is astounding
This game is so much fun! Its awesome how it draws you in and the mechanics are easy to understand! I love detailed strategy games yes im looking at you stellaris but its great to learn fast and start solving!
So I linked to my AI 101 video on Utility AI in the episode, and in that video I point to some book chapters that cover it. Plus my recent episode on Halo Infinite also discusses utility AI applications.
Wow I had great time eating lunch and watching you explain the power of simple AI on one of my favorite games. Definitely going to check more videos out!
It would be very interesting to see you break down traditional turn-based DungeonRPG AI, e.g. Wizardry games, or Etrian Odyssey (now that it has PC/Switch remasters instead of being on the DS.)
Tbh I wished for a bit more indepth discussion. Most of what was said I could gleen when the game releasedy but was very, very interested in the more intricate "small balance decisions" that make sure the levels stay solvable. Unfortunately these detaiks were handwaived, which makes me question of the author thinks of the viewers not interested/capable of understanding, or that they just don't know either, and are just regurgitating what they have read in a recent interview.
that "unsolveable" state is why I hate turn based games. few devs even consider such a situation, and even when handled at best it's unreasonable to expect it to be completely gone. RTS' almost have a similar issue, but not to the extent turn based games do
I've come to appreciate that, while some turns may be actually unsolvable, it's not the end and mission turns to damage mitigation. The Kobiashi Maru was an apt comparison. I've had missions where I could barely complete one mission and lost 3 grid, but went on to finish the island.
Good video but I don't really see the point in categorizing this as a "utility AI system". Basically every kind of AI out there is a "utility AI system" in that it almost always uses some form of heuristic or scoring system to make its choices. It seems like a very broad category that doesn't actually say anything about the AI.
I already know the Vek are designed to not be smart or optimal. Thats to prevent you from baiting them into "optimal" play the human exploits Plus the Vek are giant Kaiju bugs. Makes more sense if they are dumb.
I really, _really_ wanted to like this game, but something, some deficiency in my brain, leaves me hilariously and utterly unable to understand strategy games, even ones distilled to their essence, like ITB. For another example of how bad I am, there's a scenario in one of the Advance Wars games that places you and the AI on a small, roughly symmetrical island with choke points on the top and bottom of the map; the closest I have ever gotten to beating that scenario is a stalemate. By contrast, Doom Eternal on Nightmare difficulty Extra Lives mode (I am not nearly good enough for Ultra Nightmare) was very much a surmountable challenge for me, so I don't think I'm a *total* idiot. I just wish I could get strategy :(
Winning reliably seems like it depends on RNG until you realize how the bug AI works and bait them into artacking your own mechs, which you can simply move to negate the attack, instead of letting them target buildings which are much more expensive action-wise to save.
I feel like there's at least a small amount of Myopia around ITB, it's a fun game but it neither draws me back to play it nor even really left any memorable moments. The AI wasn't terribly challenging beyond the first few attempts when I was still getting used to it, the terrain was far more of a gameplay affecting element.
Keep it simple stupid isn’t great for writing though, despite what a certain Bethesda head writer may say. I miss when the writing in Bethesda games was actually good. Like the amount of depth to the lore of Elder Scrolls is insane but then you play Skyrim and it just felt like they never tried to add any depth to the main story or most of the side quests. Don’t even get me started about the awful writing of the 3d Fallout games not named New Vegas (different dev). I know this is completely unrelated but Bethesda just made me despise that phrase.
thankfully the AI is not pulled out often enough to notice it like you don't have to make moves at the same time the AI is also doing it and thus learning the AI would help, but for turn-based with relatively few moving parts, it is perfect also doesn't help that you are probably holding your head because sometimes it's wtf there are 4 threats and 3 of you
This video is 90% padding and has 2 intros before you even begin explaining something. And then you don't even explain how the developers actually managed to keep the game solvable, which is the real question I have about the AI.
A special thanks to friend of the channel Alan Zucconi for sharing with me the link to the original blog post by Matthew Davis. It's nice to have friends who share game development blog posts with you. 😄
Apart from balance, randomly picking one of the best options also makes the AI less predictable - if players can engineer optimal conditions to trick the AI into doing what they want 100% of the time, it becomes too easy to cheese the game.
Exactly. I was kind of hoping for more secret tips that would help me be better at the game, but at least this one does not help much. 😅
It's still basically possible to do this, it's just unreliable, and also extremely hard to do in general.
@@timbus2 just wanna put this out there that if you open the console and type “undoturn” into the console it works just like the one-time reset but you can use it as many times as possible.
Exactly, see early xcom 2.
A sectoid will raise a zombie, a phantom will make a clone etc, you can manipulate the non-damaging enemy actions to survive a turn.
One of the greatest bits of advice (which I think I've learned from this channel) is that the opponent AI does not need to be competent in the game; it just needs to be fun and engaging.
At university I made a pacman clone with custom levels. I didn't know much about implementing A* or even Pacman-like pathfinding and just made the ghosts move randomly every time they came to an intersection or corner. That, along with the fact I had to run pathfinding twice per frame because for some reason running it once led to ghosts ignoring the walls meant the "AI" was actually quite fun to play against, and my tutor thought it was way more complex than in reality :^)
There is an unreasonable amount of power in simplicity, I like that the devs chose not to over complicate things :)
Killer thumbnail 🫡
How do you not have replies and 500 likes yet
I discovered Into the Breach before FTL and was allways puzzled why it was less popular than FTL. It is such a good game.
Same
If I remember correctly, a lot of FTLs popularity came from the fact that it was a Kickstarter that did insanely well, and every news outlet reported on that and then the game. Meanwhile, into the reach is just another release from Subset games. It didn't have that same extra wave of publicity.
Oh no, you gonna make me play it again!
Mwa ha ha ha
Yup, just seeing this video is giving me the itch for it as well.😆
I've been working on a grid based puzzle game lately, and the video feels perfectly timed. Even though the video is AI-focused, the concept can apply well to lots of different systems and gameplay decisions. Weirdly enough, I made this realisation through Zachtronics, who's games are known to be quite complex to play, but some of his puzzles like Dungeons & Diagrams can be entirely played on paper. With the game I'm creating, I'm very much always reeling it back in to "good enough", and the gain is that I get to keep things simple.
You ever finish this game spyder638?
@@darkhobo Unfortunately not! I’m a software engineer professionally and I had to put it aside to get away from code & the computer at some point during the day. I left it in a place where I can pick it up again though by leaving lots of code documentation & my design docs.
Into the Breach is just an absolute masterclass in all departments it seems... such an elegantly simple game. Feel like basically 99% nowadays turn-based-strategy games could take a note or two.
Into the breach needs more love in the gaming community. Such an amazing game
Never thought of Into the Breach's scenarios as a puzzle to be solved. Really interesting.
every task you ever try to do is, or was at some point a puzzle :P
Even things that now come with instructions, at one point someone had to figure it out.
I've always felt Into the Breach was best described as "Kaiju Chess." It definitely always felt like a puzzle game to me. A puzzle game with a lot of variables, but a puzzle game nonetheless.
@@michaelcalvin42 Smart of your part, I guess to a certain extend I was using puzzle solving skills but from a more reactionary place.
I'm thinking of approaching the game again but with this kind of mindset.
lol I figured that out after playing Into the Breach and having to pause my podcast so I could 'think" about how to win a level, and suddenly I realized I never had to do that during FTL. In FTL it was always just instinctual reaction, but Into the breach makes me stare at a screen in silence for 10 minutes.
I've definately pitched this game to friends as "Pacific Rim, but a highly stressful puzzle game"
Balancing is something that "the doc" and I often talk about. Basically, how good is too good? If a player cannot win, that's gonna suck. But it also kinda sucks if the AI is obviously doing dumb moves. I feel like there's a whole research field on making really good "bad" AI
Yeah it's something you don't really hear discussed in academic circles, and I suspect there's a lot of 'dark arts' hidden throughout the industry on how to do this.
Smart AI is not the same as having a difficult game. You can have an easy game with smart AI, or a hard game with stupid AI. Generally you should just make the AI fun to play against and balance the difficulty through other means.
i'd love to see a video on rain world's AI on this channel, while not the most complex nor the most simple, the interactions it has with the player, the environment and other creatures' AI is astounding
I thought rain world had one of the most complex ai ...
Easily one of my favorite games of all time.
This game is so much fun! Its awesome how it draws you in and the mechanics are easy to understand! I love detailed strategy games yes im looking at you stellaris but its great to learn fast and start solving!
Thanks for sharing the source, it's really interesting but difficult to find good articles about videogame AI that implements Utility AI
So I linked to my AI 101 video on Utility AI in the episode, and in that video I point to some book chapters that cover it. Plus my recent episode on Halo Infinite also discusses utility AI applications.
Wow I had great time eating lunch and watching you explain the power of simple AI on one of my favorite games. Definitely going to check more videos out!
Thanks for watching and commenting. Much appreciated! 😁
This game showed me that there are no inescapable situations in life
yeah my takeaway is, squint hard enough at a hopeless situation and you will find an absurd combination of moves that bails all your buildings :)
Great video! Very clear and concise explanation, and overall polished video, subscribed
It would be very interesting to see you break down traditional turn-based DungeonRPG AI, e.g. Wizardry games, or Etrian Odyssey (now that it has PC/Switch remasters instead of being on the DS.)
one of my favorite games and a great video while the AI may be simple if definitely feels like they know your next move sometimes😅
Thank *YOU* so much! 🙏
Tbh I wished for a bit more indepth discussion. Most of what was said I could gleen when the game releasedy but was very, very interested in the more intricate "small balance decisions" that make sure the levels stay solvable. Unfortunately these detaiks were handwaived, which makes me question of the author thinks of the viewers not interested/capable of understanding, or that they just don't know either, and are just regurgitating what they have read in a recent interview.
that "unsolveable" state is why I hate turn based games. few devs even consider such a situation, and even when handled at best it's unreasonable to expect it to be completely gone. RTS' almost have a similar issue, but not to the extent turn based games do
I've come to appreciate that, while some turns may be actually unsolvable, it's not the end and mission turns to damage mitigation.
The Kobiashi Maru was an apt comparison. I've had missions where I could barely complete one mission and lost 3 grid, but went on to finish the island.
Now that I think about it, there are a bunch of in game dialogue that just straight up calls the Veks stupid
Great episode! :D
Does anyone know what Subset Games has been working on since AE released?
sooo how did he prevent them from making moves that result in an unsolvable state? I'm guessing it can still happen then?
This video changed the way I Look at the game
Good video but I don't really see the point in categorizing this as a "utility AI system". Basically every kind of AI out there is a "utility AI system" in that it almost always uses some form of heuristic or scoring system to make its choices. It seems like a very broad category that doesn't actually say anything about the AI.
I already know the Vek are designed to not be smart or optimal.
Thats to prevent you from baiting them into "optimal" play the human exploits
Plus the Vek are giant Kaiju bugs. Makes more sense if they are dumb.
I really, _really_ wanted to like this game, but something, some deficiency in my brain, leaves me hilariously and utterly unable to understand strategy games, even ones distilled to their essence, like ITB. For another example of how bad I am, there's a scenario in one of the Advance Wars games that places you and the AI on a small, roughly symmetrical island with choke points on the top and bottom of the map; the closest I have ever gotten to beating that scenario is a stalemate.
By contrast, Doom Eternal on Nightmare difficulty Extra Lives mode (I am not nearly good enough for Ultra Nightmare) was very much a surmountable challenge for me, so I don't think I'm a *total* idiot.
I just wish I could get strategy :(
19% defense: _whoosh!_ *PING!* "Blocked!" _whoosh!_ *PING!* "YES!" _whoosh!_ *PING!* "OMG. HOW."
27% defense: _whoosh!_ *CRASH* "Dammit." _whoosh!_ *CRASH!* "FUCK!" _whoosh!_ *CRASH!* "...bruh."
Really annoying how the game title keeps jumping to different corners during the various clips.
Can anyone recommend any good game developer blogs?
I wish I knew more examples, but Factorio devs have a very nice blog.
@@BlueDog15391 cool
Hmm. I never got the hang if this game. I'll try again with a puzzle game approach.
Winning reliably seems like it depends on RNG until you realize how the bug AI works and bait them into artacking your own mechs, which you can simply move to negate the attack, instead of letting them target buildings which are much more expensive action-wise to save.
I feel like there's at least a small amount of Myopia around ITB, it's a fun game but it neither draws me back to play it nor even really left any memorable moments. The AI wasn't terribly challenging beyond the first few attempts when I was still getting used to it, the terrain was far more of a gameplay affecting element.
Into The Breach is a problem solving game, not a puzzle game. Puzzles have one solution.
I feel smarter than a bug
Occam's razor
Keep it simple stupid isn’t great for writing though, despite what a certain Bethesda head writer may say. I miss when the writing in Bethesda games was actually good.
Like the amount of depth to the lore of Elder Scrolls is insane but then you play Skyrim and it just felt like they never tried to add any depth to the main story or most of the side quests. Don’t even get me started about the awful writing of the 3d Fallout games not named New Vegas (different dev).
I know this is completely unrelated but Bethesda just made me despise that phrase.
I disagree. Simplicity is what makes game AI feel so dead.
Of course complexity doesn't necessarily translate into more varied responses.
Trying to put chat gpt or whatever overcomplicated thing into your game doesn't make it better. You gotta know when to stay simple.
thankfully the AI is not pulled out often enough to notice it
like you don't have to make moves at the same time the AI is also doing it and thus learning the AI would help, but for turn-based with relatively few moving parts, it is perfect
also doesn't help that you are probably holding your head because sometimes it's wtf there are 4 threats and 3 of you
Uh oh, you mentioned Edge of Tomorrow without mentioning the original novel, that's kinda cring
Uh oh, you also mentioned Edge of Tomorrow without naming the original novel. That is equally cringe 😉
Ngl it pains me to watch the game footage because there's so many cool moves you could have pulled but you're just meandering ;_;
Professionally meandering thank you very much. 😉
chess simulator pretending to be battletech game. thumbs down.
ChEsS SiMuLaTor PreTeNdinG tO Be BaTTleTeCh GamE
Don't worry bub, it's probably just too smart for you.
failed game , compared to FTL this is a joke of a game
Ftl?
This video is 90% padding and has 2 intros before you even begin explaining something. And then you don't even explain how the developers actually managed to keep the game solvable, which is the real question I have about the AI.