|
Koramei posted:It'd help a lot of Firaxis actually had a post-release attitude towards expansions and patches that wasn't straight out of 2005. I don't think it's like they need to be watching streams all day long, but especially with Beyond Earth they just seemed so completely out of touch. Tbf personally I don't think AI cheating is a bad thing at all as long as it isn't super obvious, but actually balancing your game around strategies that happen after launch shouldn't be some unattainable thing. Firaxis' attitude towards a game's post-release life simply hasn't changed since civ 4, whereas the rest of the industry (especially EU4) has moved on a lot.
|
#
?
Oct 22, 2015 00:13
|
|
|
|
| # ? Apr 23, 2024 10:41 |
|
|
Koramei posted:It'd help a lot of Firaxis actually had a post-release attitude towards expansions and patches that wasn't straight out of 2005. I don't think it's like they need to be watching streams all day long, but especially with Beyond Earth they just seemed so completely out of touch. Tbf personally I don't think AI cheating is a bad thing at all as long as it isn't super obvious, but actually balancing your game around strategies that happen after launch shouldn't be some unattainable thing. To be fair to Firaxis on that part, I wouldn't want to touch the AI either considering they need to completely fix the rest of the mess that is Beyond Earth before they even start looking into what would make for a good AI. Panzeh posted:Typically difficulty increases AI aggression which is sometimes good, sometimes bad. It's really hard to justify hiding your best AI behaviors behind a difficulty curve, though- if you have a good AI you want to show it off. Definitely does depend though, for example in some of the early HOMM games the AI will actually play smarter defensively as well, keeping tabs on how far the player is from it's castles and such. I agree about not hiding your best AI, especially if harder difficulties also cheat resources and such. It was kind of funny though with AoW3, seeing that many players thought the game was actually too hard even on the easier difficulty settings. Apparently a decent number of people wanted a mode where the AI did literally nothing but sit in town and occasionally build units, the idea of not doing the same themselves being a foreign concept.
|
|
#
?
Oct 22, 2015 00:13
|
|
|
Ofaloaf posted:When I was a kid playing the first Age of Empires, I thought that the "computer" enemies were actually all human players in the employ of Ensemble Studios, and that setting different difficulty levels connected me with players of varying levels of skill. I'd trash talk them with in the in-game chat and call them poopy-head if they were on the opposing side, and thank them if they were fighting alongside me. I thought the same thing, and wondered whether Ensemble released an "inverse" campaign in like China or something, so as I'm playing through the Joan of Arc campaign someone else on the other side of the world is trying to complete the Black Prince campaign.
|
|
#
?
Oct 22, 2015 00:26
|
|
|
Gwyrgyn Blood posted:I agree about not hiding your best AI, especially if harder difficulties also cheat resources and such. It was kind of funny though with AoW3, seeing that many players thought the game was actually too hard even on the easier difficulty settings. Apparently a decent number of people wanted a mode where the AI did literally nothing but sit in town and occasionally build units, the idea of not doing the same themselves being a foreign concept. I'd be curious to see statistics on what difficulty levels are most played for most strategy games, and maybe broken up by age groups. I know I think we get a bit of an inflated sense of most people's skill when we talk on online communities for games too; I have a strong suspicion that the vast majority of the playerbase for most strategy games are actually incredibly poo poo at them. When I play Age of Empires with my friends who don't play it much (but still played it enough to know where everything is etc) we can sit around building stuff for 45 minutes and I'll go raid them and find they haven't even made 12 units yet.
|
|
#
?
Oct 22, 2015 00:34
|
|
|
Yeah, I remember playing Age of Empires, Red Alert, and Warcraft I/II back in the day, and being just terrible at it. I think there's just a grand understanding of the themes of strategy games you need to know to understand how to be good at them, and when you're young that doesn't exist. I think SMAC/X was the first game I figured out enough to be able to play on the top difficulty.
|
|
#
?
Oct 22, 2015 00:41
|
|
Yes, it's like a lava lamp.
|
zedprime posted:To bring it back home, isn't this what they do for EU4? IIRC they just play it themselves for the most part, although they do nerf tactics they consider exploits once and a while
|
|
#
?
Oct 22, 2015 00:42
|
|
|
Panzeh posted:That's also a pretty insane task right there. Lots of apps send very detailed information back to the devs. They do so for a variety of reasons, typically monetization. I think you might be surprised to find out that a mobile dev for instance can figure out every single thing you're doing on the app, what screen you're looking at, how long, which buttons you pressed, etc. The idea of taking that data and using it to improve the game instead of chase $ is not really out there or even hard, so I found the response here was surprising.
|
|
#
?
Oct 22, 2015 01:11
|
|
|
Rakthar posted:Lots of apps send very detailed information back to the devs. They do so for a variety of reasons, typically monetization. I think you might be surprised to find out that a mobile dev for instance can figure out every single thing you're doing on the app, what screen you're looking at, how long, which buttons you pressed, etc. It's not the getting information- it's parsing it and then trying to glean anything from it that's the problem. Humans have enough problems deriving the right lessons from trends as it is, now said humans have to make a program that does this, much less sorting through this data. Koramei posted:I think we get a bit of an inflated sense of most people's skill when we talk on online communities for games too; I have a strong suspicion that the vast majority of the playerbase for most strategy games are actually incredibly poo poo at them. When I play Age of Empires with my friends who don't play it much (but still played it enough to know where everything is etc) we can sit around building stuff for 45 minutes and I'll go raid them and find they haven't even made 12 units yet. It's kinda interesting how this works- people tend to play difficulties too high for them on game speeds too slow and maps too big. Honestly, 4x games should probably not give the option to players to make dumb games, because they inevitably whine when the game isn't great that way. The success of civ 5 is probably a testament to the fact that people want to build a few cities, trounce some AIs, and call it a day. Koramei posted:It'd help a lot of Firaxis actually had a post-release attitude towards expansions and patches that wasn't straight out of 2005. I don't think it's like they need to be watching streams all day long, but especially with Beyond Earth they just seemed so completely out of touch. Tbf personally I don't think AI cheating is a bad thing at all as long as it isn't super obvious, but actually balancing your game around strategies that happen after launch shouldn't be some unattainable thing. Most of the good strategy games have bad AIs, from Imperialism to SMAC. If the game was good people would get around the AI, but it's clear the base is bad and that's the thing that needs to be fixed.
|
|
#
?
Oct 22, 2015 02:53
|
|
|
VostokProgram posted:Make the game phone home with every move every player makes, do some machine learning on the collected information, release AI patches monthly. This is roughly as useful advice to an AI developer as 'invoke the seventh sphere of Jupiter at midnight to bless your AI code'.
|
|
#
?
Oct 22, 2015 08:09
|
|
|
Suffice it to say that with current technology, having "good AI" for any game with the complexity of EU4 is impossible. I don't care how much data you have, you can't get AIs to behave like players all the time. Oh hey, Wiz beat me to it
|
|
#
?
Oct 22, 2015 08:10
|
|
|
Panzeh posted:Honestly, 4x games should probably not give the option to players to make dumb games, because they inevitably whine when the game isn't great that way. Please do not develop any games.
|
|
#
?
Oct 22, 2015 08:11
|
|
|
Wiz posted:This is roughly as useful advice to an AI developer as 'invoke the seventh sphere of Jupiter at midnight to bless your AI code'. Do you not?
|
|
#
?
Oct 22, 2015 08:13
|
|
|
To put it another way, learning AI is complete fantasy outside a few highly specific behaviours.
|
|
#
?
Oct 22, 2015 08:15
|
|
|
Maybe it wouldn't be if you offered the proper sacrifices to Jupiter.
|
|
#
?
Oct 22, 2015 08:42
|
|
|
TBF I would put the chance of sacrifices to Jupiter resulting in good AI significantly higher than 'get a bunch of data and just add some ~machine learning~'.
|
|
#
?
Oct 22, 2015 08:49
|
|
|
Hellenism confirmed for EU4.
|
|
#
?
Oct 22, 2015 08:56
|
|
|
zedprime posted:To bring it back home, isn't this what they do for EU4? Yeah no kidding didn't Wiz like manage to shoot down every way to "game" the Hundred Years War over the course of however many of EU4's patches?
|
|
#
?
Oct 22, 2015 08:56
|
|
|
GSD posted:Hellenism confirmed for EU4. I still want Rome 2. Only with less focus on Rome because at that game's start date Rome is just a regional power and the Diadochi are cooler and more interesting. Also the map should extend to India, seriously. EU: Rome's map was tiny and had almost nothing on it.
|
|
#
?
Oct 22, 2015 09:00
|
|
|
Can you at least promise me that HoI4's AI will know how to conduct a goddamn naval invasion?
|
|
#
?
Oct 22, 2015 09:01
|
|
|
Dibujante posted:Firaxis' attitude towards a game's post-release life simply hasn't changed since civ 4, whereas the rest of the industry (especially EU4) has moved on a lot. I met some Firaxis people in a bar after Gamescon a last year. I wanted to ask them about their post-release stuff in general, but they left while I was still complaining about XCOM bugs/patch support 
|
|
#
?
Oct 22, 2015 09:04
|
|
|
Darkrenown posted:I met some Firaxis people in a bar after Gamescon a last year. I wanted to ask them about their post-release stuff in general, but they left while I was still complaining about XCOM bugs/patch support XCOM's one of their less buggy games though. Compared to Beyond Earth it's a towering masterpiece of flawless game design.
|
|
#
?
Oct 22, 2015 09:05
|
|
|
gradenko_2000 posted:Yeah no kidding didn't Wiz like manage to shoot down every way to "game" the Hundred Years War over the course of however many of EU4's patches? HYW and Byzantium were my honeypots, yes. Every patch people would figure out a new strat for them and post it on reddit/the forums, and I would be watching.
|
|
#
?
Oct 22, 2015 09:06
|
|
|
I guess you only have Three Mountains left for that. Or maybe some WC games. Edit: I am pretty sure Three Mountains is impossible with the current patch.
|
|
#
?
Oct 22, 2015 09:10
|
|
|
I get why people would try to cheat their way to victory with Byzantium, but what's the deal with the HYW events? I never heard about any of this.
|
|
#
?
Oct 22, 2015 09:59
|
|
|
Gort posted:XCOM's one of their less buggy games though. Compared to Beyond Earth it's a towering masterpiece of flawless game design. BE wasn't out then, came out a month or 2 after.
|
|
#
?
Oct 22, 2015 10:38
|
|
|
Wiz posted:TBF I would put the chance of sacrifices to Jupiter resulting in good AI significantly higher than 'get a bunch of data and just add some ~machine learning~'. Even for a wizard? You can't say some incantations and just magic up an algorithm?
|
|
#
?
Oct 22, 2015 10:41
|
|
|
Kavak posted:I get why people would try to cheat their way to victory with Byzantium, but what's the deal with the HYW events? I never heard about any of this. It's back when England and France started at war, and England winning would net them France in a PU. There was quite a bit of effort spent in finding the best and quickest way to win the war, which would get fixed the next patch, then someone would dig up something new. Among other things it led to Military Access being a two-way street.
|
|
#
?
Oct 22, 2015 10:43
|
|
|
Wiz posted:TBF I would put the chance of sacrifices to Jupiter resulting in good AI significantly higher than 'get a bunch of data and just add some ~machine learning~'. Machine learning in principle is not a bad idea, but the actual implementation would be nowhere near as good as the ~self-learning AI~ that people who suggest this stuff dream of. People think "man wouldn't it be cool if the AI could do that", and it sounds cool, but the truth is, no, it would not be all that cool.
|
|
#
?
Oct 22, 2015 10:55
|
|
|
you might be the richest software developer ever if you make a consistently useful machine learning system though. so if you think you have such a great idea, time to swing for the fences!!
|
|
#
?
Oct 22, 2015 11:00
|
|
|
Larry Parrish posted:you might be the richest software developer ever if you make a consistently useful machine learning system though. so if you think you have such a great idea, time to swing for the fences!! If you could make an way to mine data and use ~machine learning~ to glean something from it you could do a lot better than making video game AIs. The Sharmat posted:Please do not develop any games. I remember when I was testing Star Ruler 2 and a bunch of people complained about how boring it was and it was universally because they jacked up the planet count to 500 so you just spent three hours connecting dots where the game was tested and tuned for settings closer to 10-20 planets per player. A lot of players are really loving stupid and go "HUGE MAP/MARATHON/MAX OUT EVERYTHING" because they think that's fun. Panzeh fucked around with this message at 11:32 on Oct 22, 2015 |
|
#
?
Oct 22, 2015 11:27
|
|
|
I say let the stupid players be stupid because they will buy your game anyway rather than punish people that aren't stupid by taking away options they may want to use for some crazy gimmick game or mod or something.
|
|
#
?
Oct 22, 2015 11:42
|
|
|
Wiz posted:TBF I would put the chance of sacrifices to Jupiter resulting in good AI significantly higher than 'get a bunch of data and just add some ~machine learning~'. I sacrifice a goat to Odin every Midsummer and Yuletid, HOI4 AI is looking really good. So I guess it works?
|
|
#
?
Oct 22, 2015 12:38
|
|
|
Does this mean it can manage amphibious ops?
|
|
#
?
Oct 22, 2015 12:40
|
|
|
The Sharmat posted:I say let the stupid players be stupid because they will buy your game anyway rather than punish people that aren't stupid by taking away options they may want to use for some crazy gimmick game or mod or something. Then leave in the capacity for it to be modded in. I really hate this notion that more options is always, always better, because the inevitable result is always sloppy games that bear no relation when actually played to the game as it was designed.
|
|
#
?
Oct 22, 2015 12:43
|
|
|
Saying a game should not always be hardlocked to 10-20 systems per player is a bit different from "more options is ALWAYS better".
|
|
#
?
Oct 22, 2015 13:03
|
|
|
Pharnakes posted:Does this mean it can manage amphibious ops? This man understands
|
|
#
?
Oct 22, 2015 13:17
|
|
|
Define "manage"? No specifically naval invasions are a hard nut to crack in AI and do it well. Not the area I've been working with in the AI but I have seen while testing my more "The Big Picture AI" stuff actually planning a Naval invasion of Britain and succeeding. But the AI can always be improved upon.
|
|
#
?
Oct 22, 2015 13:23
|
|
|
Groogy posted:Define "manage"? That's 100 times better than Darkest Hour at least. 
|
|
#
?
Oct 22, 2015 13:30
|
|
|
All the more reason to do an antiquity game where ships couldn't carry that many troops, were mostly for supply, and rarely if ever sailed out of sight of the coast.
|
|
#
?
Oct 22, 2015 13:31
|
|
|
|
| # ? Apr 23, 2024 10:41 |
|
|
Pharnakes posted:Does this mean it can manage amphibious ops? There was a DD early on where Ethiopia managed to do a naval invasion of Italy. It caused some consternation because a) the AI wasn't supposed to be able to do naval invasions yet and b) Ethiopia doesn't have a fleet
|
|
#
?
Oct 22, 2015 14:15
|
|