Weapon Caches and ai Pathfinding

[PsyGuy]

Hello this is primarily directed at AmperCamper, but if anyone else in the community somehow knows the answers to these questions please feel free to provide input. Do the small collectible weapon caches scattered around the map do anything besides give points in the after action report? It's difficult to tell if these have any effect like making the insurgents have less weapons available, or affect their accuracy or some other function that has a tangible effect on gameplay. When certain enemy types spawn in (rifle, rocket, sniper, mg, etc) are these roles set to their specific class for their life duration or can they run into a weapon cache building, either large or small, and rearm in any way? The big weapon caches do of course spawn the rocket guys who then emerge from those particular buildings, but can the enemy effectively change their class at caches? Does destroying big caches and/or collecting small caches result in less machine gun enemies as well, or is it just rockets? Does collecting a sniper rifle mean less snipers, or a grenade cache less grenades, or are small caches all considered one cumulative effect throughout the course of the match? Is there some sort of class downgrade mechanic that dynamically affects the insurgent loadout? While it is satisfying to pick up the small caches and deny the insurgents more weapons, I have no way of knowing if this actually has any effect. This seems difficult to test or I would have, but I can't see enemy actions through walls and relying on clothing as an identifier can be confusing because of either new spawns in the area or other enemies moving around the map from other areas could be wearing the same or similar clothing, or the test subject gets away just enough to blend in with other enemies thus losing track of the observed specific individual for the aforementioned reasons, all assuming the tracked target was actually even using those caches across the map or in his proximity at all.

There will likely be some sort of progression system in the future, but all I can discern from the small caches is that they give you a higher score after the mission. The mission score does seem rather pointless, especially considering that the various modes (Fubar, Hard, etc) don't seem to give any extra points, and the best way to get the highest possible score is to just load a map with max buildings on normal mode and collection speedrun it without getting hit or losing any Fireteam members. My highest score was 8,520 like 6 months ago and I do specifically remember I was only able to get such a high score from easy mode cache collecting while gunrunning around as fast as possible, I think HLZ Wolf on normal with max buildings. (interesting concept but very low intensity even on Hard it needs Fubar mode or just higher baseline enemy count it's probably the most boring map by far because of all the downtime, maybe make it some sort of small-cache themed dynamic spawns map) Currently it seems the best way to max out points is to sprint around without getting hit while collecting caches. It ends up feeling very arcadey, especially for the sake of a redundant scoring system that seems to have little to no effect on actual gameplay. I am assuming the devs have plans to expand the scoring/progression system in the future, but the current state of the aar/rank mechanism is basically meaningless. You can progress up to 2nd Lieutenant but it stops there and has no function, they should have just left max rank at E-9 enlisted as it's strange that it jumps up to officer rank but only to O-1. I would still probably rather be an E-1 than an O-1. If given the choice between two boots I'm siding with enlisted.

Also the most recent hotfix did noticeably improve Fireteam ai pathfinding but mostly for the Fire and Assist and only slightly for the Ready. He still gets stuck alot, though less than half as much as before which was every minute or two, so while it is an improvement it still needs work. I see alot of posts about the Ready pathing his way around buildings and then entering from the other side advancing directly at you while you clear through and he always ends up getting shot. Obviously alot of stuff is a work in progress but even after almost 300 hours in game I still often put a round or two either into or close to the incoming friendly. There is literally not enough time to positive-id him because if you don't immediately blast an immediate range enemy in Six Days then you're just dead so id'ing really only works outside or in relatively distant doorways/windows if within the building, plus the only reason he is advancing at you is because of a pathfinding bug otherwise he would instead be properly navigating the environment. The game inherently trains players to reactively burst down close range enemies in uncleared indoor spaces or when peeking corners. Even on normal mode while crouch-leaning around corners on burst fire with the flashlight off, you will still barely get the drop on an insurgent before he rips a green laser burst into you, and if there's more than one then you have to burst fire mag dump suppress egress as the room turns into a dusty desert sandstorm laser show. This particular issue I would consider to be high priority because of how thoroughly unrealistic and immersion breaking it is whenever this happens, he doesn't even announce his presence or slowly maneuver in any sort of tactical manner he just silently aggressively flies around the corner at you through a dark uncleared building coming from the hostile side of the map. Inversely sometimes friendlies will path through uncleared occupied buildings just to get to where you are outside of the building, either solo sprint clearing the enemy building or getting surrounded and incapacitated inside it and then inevitably dying because I can't safely clear through the whole space to revive him in time, and instead of crawling away he just stays sitting there directly in the middle of the room waiting to catch a stray round. This situation of a single friendly inside of a building being incapped and surrounded would just literally never happen with proper pathfinding as the entire event is basically just a snowballing bug resulting in an unintentional hostage situation.

Admittedly knowing nothing about the nuts and bolts inner workings of Six Days, I would suggest some sort of tree-branching system of "if-then" relative to their training and the maneuvers they perform around the map. It sounds like an obvious thing but if this tree-branch training system was effectively implemented then, for example, instead of the ai solo sprinting through enemy strongholds they would reach a hitch somewhere along the tree in that decision making process that says "bad idea, refer to other branching behavior; alternatively cancel and continue following Team Lead" or something of that nature. An existing in-game example of this sort of dynamic behavior is their situational awareness for covering sectors. Anywhere the player looks they cover another sector, they constantly adjust to your movement and their environment with both their directional observation as well as their x-y physical positioning on the map, while fluidly and seamlessly continuing this behavior at all times. Their clearing behavior is a potential example as well, it's still somewhat rough but has certainly improved over time "breach door -> clear room -> new door? -> if yes, repeat -> if no, regroup on Team Lead"(preferably choosing to navigate through already cleared spaces to rally up on Team or somehow registering and back-tracing Team Lead's path to avoid oncoming unidentified friendlies) Friendlies do make "cleared" callouts when clearing rooms through buildings, so they currently do have some level of dynamic environmental technical recognition. I know basically nothing about programming or coding but years ago I heard someone say in a game dev interview that developing a video game is like packing a suitcase full of clothes that you shove everything in and barely get it to close, and then when you open it up again (patches, updates, etc) everything comes flying out and makes a huge mess that can in some ways be even worse than before. Seems accurate. Good progress, and good luck.

[AmperCamper]

Hey! Thanks for the questions.

Weapon caches strictly grant XP right now (and boost role score as a result). We've had discussions about how we can make these more impactful, and you've given us some good ideas here.

Noted on the Ready AI. I shared a response in a similar thread, but I want to confirm we're aware there's more work to be done in this area, and our AI team is looking into these as we speak.

I appreciate the tree branching behavior suggestion, too. The entire system is much more complex than this under the hood (especially in a procedural environment), but I follow your thought process.

[PsyGuy]

I'm glad to help provide feedback for the development process, and I have a few further thoughts after reading through the patch update comments regarding the flinch suppression mechanic and the player health/damage model, as well as some ai-pathing and ai-targeting related issues. I was going to post it there but it ended up being long so I'll just add it onto this thread. Please do expedite making those weapon caches useful because I am still not going to stop picking them up even though I now officially know that they are completely useless.

The flinch suppression mechanic, while sometimes inconvenient, is not nearly as egregious as it used to be. I don't know the extent to which it was changed but in previous builds I always found it to be much more frequently intrusive than it is now, to the point where I basically don't even notice it anymore and just register it as a suppression signal to take cover or break contact. It can be problematic for the SAW due to the way it immediately interrupts full auto bursts to the point where you have to fire the weapon almost semi-auto with rapid taps (as in either left mouse or right trigger) in order to overcome the suppression, but whatever changes were made seem to work well for semi and burst weapons as these are inherently tap-fire anyway. I find that I am consistently and reliably able to output effective suppression fire while being engaged, often enabling me to escape multiple enemies by rapidly returning fire as I maneuver away from them. It seems like they increased the reaction sensitivity of player return-fire so it doesn't completely lock me down like it used to and now I can actually manage the situation. I might just be conditioned to it and overestimating any adjustments but the system overall now seems much more functional and far less frustrating. Previously if enemies rushed into the room they could easily lock you down with just their normal bot firing, but that doesn't really seem to occur anymore as long as you make an attempt to immediately return fire while simultaneously pushing into or pulling away from the source of incoming fire. If I do get fully suppressed it's usually because my current cover is between a rock and a hard place rather than being outgunned by the enemy, and therefore better to remain stationary and let the team handle it.

The increased player health and reduced required bandaging is definitely a step in the right direction but I feel that the current regenerate-until-bandage damage model is decidedly overtuned considering that players can repeatedly take such immense fire with such consistent frequency. If it hits the SAPI plate then great, but I think the classical tactical damage model of permanent damage to specific body areas that RoN has continued to use and refine over time would undoubtedly suit Six Days far better than the current model, so much so that I feel it is highly likely that the present form is only a placeholder until the developers implement such a system for a permanent damage model with limited healing capability. For example when damaged by enemy fire the player starts to bleed, the player is then able to bandage to stop bleeding and pause the HP pool loss, which does not regenerate, instead gradually degrading whenever taking damage until progressing into an incapacitation state wherein the team can then revive, with the body armor and helmets being able to take the appropriate damage in the appropriate places, and with an improved suppression system and effects. It would be extremely cool to have a system that allows for taking a near-impact mortar or rocket and have significant but survivable damage to extremities and then being able to continue fighting in that state with reduced capabilities for movement and accuracy. Slightly video-gamey but a far superior solution to magical regeneration, and the limited layers of health would serve to add to player immersion and decision making rather than detract from it like with the current placeholder Rambo-regen system. Additionally the visual aspect of having character models covered in blood with no real persistent damage is slightly immersion breaking and it would be much better to look at someone and see in real time where the real damage persists via the location of blood stains on a permanent model. I am curious as to whether the ai and the player do/will share the same type of health systems, or if they have their own unique ai health system. One current example of this difference would be that the ai Fireteam can bandage wounds and also be revived from incap, while enemy ai can do neither they just fall over dead.

Also the Fireteam ai pulls off some sick 100 meter rooftop headshots all the time no problem, but then when they get into a close-range fight both they and the enemy just dance around each other until one of them (usually the insurgent) rips a full auto burst to the face, entirely bypassing the incapacitation mechanic and just instantly killing them. Something is causing the ai to override the minimum required standoff distance to engage each other with their weapons as they currently seem to have a particular set of issues at close-range where they register the enemy but get too close too engage, or are maybe pseudo-suppressed or something, whereas with long-range they don't register the enemy at all past a certain point of ~125 meters and therefore simply won't engage them even when targets are in clear view thus requiring a manual Suppression order on enemy troops in the open. Ready or Not had a very long and rough road for both friendly and enemy ai that I remember all too well and it is only relatively very recently that RoN has finally become an actual full-feature release, I think for RoN the first legitimately and comprehensively constructed version was like v1.2 as v1.0 was rushed out for the holidays and had some very persistent ai issues between the superhuman enemies and barely functional friendlies and was definitely not what one might consider a full 1.0 release, but all of that faded into the past as development continued and it now has 183,000 Steam reviews and is an unbelievably incredible successor to the tac-shooter throne for the legendary SWAT series.

The RoN enemy ai aimbot headshots through walls was a huge meme it happened to me dozens of times and it was way worse than 6DIF enemy aimbot ever has been, plus up until 1.0 you had to scour every single inch of the entire map for the last remaining civilian or suspect or evidence and depending on how bugged it ended up being it could take almost as long as the actual mission duration. It was literally the equivalent of trying to find one single small weapons cache in an entire map of cluttered dark buildings in Six Days. There were such long update gaps during its development that for a long time alot of people thought RoN was more or less an elaborate and expensive vaporware product, but after the dust settled when it was finally finished it went on to become one of the greatest games ever made. I would assess that in terms of gameplay the primary differences between the two are the quantity and intensity of killing enemies. RoN's approach to combat is more of a "Think about maybe shoot a bad guy or two" whereas Six Days is "Shoot lots of bad guys whole time nonstop" and that is the key differential between them and why my 6DIF playtime quickly managed to overtake years of RoN playtime. The violently impactful gunplay, meaty shooting mechanics, and the deep satisfaction of wasting 100+ insurgents in just a few minutes in relentless point blank cqb fights is why I keep coming back. That is the core strength of Six Days and it is what should be focused on and improved and iterated upon, with the general Fireteam concept (for both players and ai) and procedural map generation combining to form the uniquely and intricately built frame that supports the extreme intensity of the experience. When this game works it's incredibly fun and engaging, and I'm onboard all the way but for the moment it is still in v0.4 early access and has a long way to go.

Regarding team commands to blow caches, until pathfinding is improved the Team Lead will have to manually order Fire and Assist out into the street while either Team or Ready plants the charge, it's simply too high of a risk to hope that they all happen to stumble out of the building in time when quite often they do not except for the smallest and straightest of buildings. Sometimes cache positioning in the really big buildings can be an inevitable death sentence for the ai even if you plant the charge yourself, because depending on the interior layout and how long or cluttered the exit path is even a single ai still won't make it out in time. Hopefully devs will quickly release a pathfinding fix, ironically enough I do feel much more confident with the Ready and much more vulnerable without him. There have been many posts emphasizing how terrible the ai is and although I really like the implementation of the Fireteam ai, it is objectively in a arguably janky and barebones state, plus by now I am used to their behaviors and the way in which they must be managed, whereas someone coming over from RoN ai is going to find the current state to be substantially less complex and complete, quite possibly even confusing and frustrating. Sometimes they do exactly what they should, sometimes they do the direct opposite of what they should, and sometimes they don't do anything at all. I would assume a full Human coop Fireteam is extremely capable in the most demanding of situations but since I don't play the coop online and after the pre-Fireteam solo play struggle I am just glad to have a semi-functional team at all and am greatly looking forward to future updates.