Another good design decision that you touch on some: make enemy npcs non-human. Zombies, undead, monsters, aliens, robots, etc also open up many more design options than trying to be "like a player." And it removes expectations that they behave like players or have similar stats (damage, toughness, etc) as players, or even humans in general
I suppose the way to fix it would be for example if the ai sees a player and the player runs behind a bush the accuracy drops say 50% if the player is running stright drop by 20% zig zagging 25% ect
Tarkov scavs are not as bad as people think. They have a good rush mechanic, group ai, interactions with environmental objects, realistic reactions. They are a bit jank in movement and the balancing is off, if battle-state worked on that (they plan to soon) they would be really good.
I play a lot of Hunt and I live your ides for escaped prisoners and such, and agree that lantern/torch grunts one shotting you with a red barrel is the absolute worst.
Definitely agree that the NPCs need more interactions and more behaviors than just being either threats or noise makers. I'm really familiar with Tarkov, so I'll speak on that. Tarkov scavs in lore (read the books if you haven't, they're not the best books in the world but they're still pretty cool) are all apart of various gangs in the region. The ones we have in game feel like they're only loosely connected though, and they don't even scavenge different parts of the map. Roaming AI should definitely be a thing. On top of that, there's so much potential with having faction reputation and more unique interactions. Imagine if you're USEC you have the ability to join the Rogues, while if you're BEAR you can team up with RUAF remnants/patrols (they're planned, along with GRU hunter-killer teams that chase down USECs). Or as a pscav you can use the different hand gestures to maybe initiate a barter with an AI scav (you can already get them to follow you based on your scav karma using the "follow me" gesture). More idle behaviors across the board too, and I think that all the AI should be toned down to realistic competency for their factions. Rogues should need scopes on their guns to just hit accurate bursts at range, and scavs should definitely have their damage and hit registration bugs fixed to cut down on bullshit deaths. I don't think making the AI not a competent threat in the right situations is necessarily good, but they also shouldn't just be cracked as soon as you run into them, seeing through concealment and cover while hitting shots that would be much more difficult for a player (who is representing a trained PMC), let alone what's supposed to be a gangster. I'm currently working on AI for what I want to make into a single-player extraction shooter-esque RPG heavily inspired by Tarkov, but with a bigger focus on the AI interactions and using them to make the world feel alive. Basically an improved single-player focused gameplay loop of Tarkov but with improved AI and light AI-sim features (sorta like A-Life for STALKER) and trying to connect the player's actions to the world a lot more. Passive and non-aggressive behaviors will play a large part of story and general AI interactions, with hopefully civilian and neutral factions being very present in most normal encounters that you will have to account for (or not, and pay the consequences). How far I'll get with it I don't know, but I hope to see the stuff I'm planning on making also pop up in these games since I love the gameplay loop of them a lot and REALLY enjoy what Tarkov is trying to pull off.
@@dappersloth No name yet. Overall theme is still being worked out, rn the assets I have fit a Middle Eastern setting quite a bit so I'm leaning towards that. Intention is to have wide faction variety and choice in your affiliation with those factions. Currently I'm working on the interaction side of the AI, and once I get that in a place I'm happy with I'm going to switch and start working on the background-sim aspect. My philosophy going into designing the AI is essentially setting-situation-state (just made this up but it describes my design process really well) where behavior should match the setting the AI is in, the individual situations within that setting they can be found in, and the exact state an AI might be in at any moment during those situations. For the interaction-time behavior and decision making part of the AI I'm using the framework outlined by Bobby Anguelov in his archived live stream on youtube "Simple and Scalable Game AI Decision Making." It's a really novel concept that I decided to go with, replacing my previous solution which was a HTN planner. Work has been sorta slow overall since I'm full-time at college, but I'm hoping to make massive progress by the end of next quarter and starting with summer and hopefully have a prototype game just focused on combat and passive behaviors by end of summer. Depending on the reception of that I'll decide what needs more work and go from there. I want to start working on devlogs and have a script written for a devlog 0 sort of describing my design intentions, but I want to get a more working demonstration of what I've done so far before I start with any devlogs. Besides AI, I made a working ballistics system with bullet drop, air drag, material and object based penetration, ricochets, etc., a weapon base that I'll be expanding heavily once I get the AI to a good state and start working on that prototype game, and a bunch of framework systems that will be used for either AI, background-sim, and other systems that will be used to drive the feeling of a living and changing world. Also got some assets I picked up from humble bundles and sales but I'm only going to use them sparingly and when necessary, I'm still going to try and pour as much custom work and content into whatever game I make out of this as possible without going broke.
Another good design decision that you touch on some: make enemy npcs non-human. Zombies, undead, monsters, aliens, robots, etc also open up many more design options than trying to be "like a player." And it removes expectations that they behave like players or have similar stats (damage, toughness, etc) as players, or even humans in general
Dude good video, you deserve more subs
Thanks so much, means a lot!
@@dapperslothI can’t believe you don’t have more subs too. Your editing is tight and your scripts are insightful and meaningful. You do Good work.
Maybe Marauders can add a mode that NPC is not human.
Alien monster is not feeling out of place for Marauders😆
I suppose the way to fix it would be for example if the ai sees a player and the player runs behind a bush the accuracy drops say 50% if the player is running stright drop by 20% zig zagging 25% ect
Tarkov scavs are not as bad as people think. They have a good rush mechanic, group ai, interactions with environmental objects, realistic reactions. They are a bit jank in movement and the balancing is off, if battle-state worked on that (they plan to soon) they would be really good.
I play a lot of Hunt and I live your ides for escaped prisoners and such, and agree that lantern/torch grunts one shotting you with a red barrel is the absolute worst.
Definitely agree that the NPCs need more interactions and more behaviors than just being either threats or noise makers. I'm really familiar with Tarkov, so I'll speak on that.
Tarkov scavs in lore (read the books if you haven't, they're not the best books in the world but they're still pretty cool) are all apart of various gangs in the region. The ones we have in game feel like they're only loosely connected though, and they don't even scavenge different parts of the map. Roaming AI should definitely be a thing. On top of that, there's so much potential with having faction reputation and more unique interactions. Imagine if you're USEC you have the ability to join the Rogues, while if you're BEAR you can team up with RUAF remnants/patrols (they're planned, along with GRU hunter-killer teams that chase down USECs). Or as a pscav you can use the different hand gestures to maybe initiate a barter with an AI scav (you can already get them to follow you based on your scav karma using the "follow me" gesture). More idle behaviors across the board too, and I think that all the AI should be toned down to realistic competency for their factions. Rogues should need scopes on their guns to just hit accurate bursts at range, and scavs should definitely have their damage and hit registration bugs fixed to cut down on bullshit deaths. I don't think making the AI not a competent threat in the right situations is necessarily good, but they also shouldn't just be cracked as soon as you run into them, seeing through concealment and cover while hitting shots that would be much more difficult for a player (who is representing a trained PMC), let alone what's supposed to be a gangster.
I'm currently working on AI for what I want to make into a single-player extraction shooter-esque RPG heavily inspired by Tarkov, but with a bigger focus on the AI interactions and using them to make the world feel alive. Basically an improved single-player focused gameplay loop of Tarkov but with improved AI and light AI-sim features (sorta like A-Life for STALKER) and trying to connect the player's actions to the world a lot more. Passive and non-aggressive behaviors will play a large part of story and general AI interactions, with hopefully civilian and neutral factions being very present in most normal encounters that you will have to account for (or not, and pay the consequences). How far I'll get with it I don't know, but I hope to see the stuff I'm planning on making also pop up in these games since I love the gameplay loop of them a lot and REALLY enjoy what Tarkov is trying to pull off.
That’s so cool! Do you have a name for the project or any more information you can share? Thanks for the input on Tarkov too!
@@dappersloth No name yet. Overall theme is still being worked out, rn the assets I have fit a Middle Eastern setting quite a bit so I'm leaning towards that. Intention is to have wide faction variety and choice in your affiliation with those factions. Currently I'm working on the interaction side of the AI, and once I get that in a place I'm happy with I'm going to switch and start working on the background-sim aspect. My philosophy going into designing the AI is essentially setting-situation-state (just made this up but it describes my design process really well) where behavior should match the setting the AI is in, the individual situations within that setting they can be found in, and the exact state an AI might be in at any moment during those situations.
For the interaction-time behavior and decision making part of the AI I'm using the framework outlined by Bobby Anguelov in his archived live stream on youtube "Simple and Scalable Game AI Decision Making." It's a really novel concept that I decided to go with, replacing my previous solution which was a HTN planner. Work has been sorta slow overall since I'm full-time at college, but I'm hoping to make massive progress by the end of next quarter and starting with summer and hopefully have a prototype game just focused on combat and passive behaviors by end of summer. Depending on the reception of that I'll decide what needs more work and go from there. I want to start working on devlogs and have a script written for a devlog 0 sort of describing my design intentions, but I want to get a more working demonstration of what I've done so far before I start with any devlogs.
Besides AI, I made a working ballistics system with bullet drop, air drag, material and object based penetration, ricochets, etc., a weapon base that I'll be expanding heavily once I get the AI to a good state and start working on that prototype game, and a bunch of framework systems that will be used for either AI, background-sim, and other systems that will be used to drive the feeling of a living and changing world. Also got some assets I picked up from humble bundles and sales but I'm only going to use them sparingly and when necessary, I'm still going to try and pour as much custom work and content into whatever game I make out of this as possible without going broke.
Hunt Showdown is the best FPS!!! 🤠🤠
Yeah who ever figures out a good way to mimick players is gonna make alot of money.