Home » MODDING HQ 1.13 » v1.13 Coding Talk » Experimental Project 7
|
|
Re: Experimental Project 7[message #353206 is a reply to message #353202]
|
Wed, 18 April 2018 13:35 
|
|
townltu |
 |
Messages:384
Registered:December 2017 Location: here |
|
|
Human (and most other creatures) body is designed to stabilize the head during movement,
in order to keep visual inputs "computeable",
the somewhat higher movement compared to humans torso should be mostly compensated by arms
which are at least partially included in "torso" target zone("shoulder hit")
while the legs have more movements and are harder to hit.
Of course only valid if target is not running directly(more or less) in direction of aiming line.
i.e. if its not about balance but real condition,
option for additional penalty should go for legs of moving taget, increasing with angle.
Makes it harder to stop enemy with aimed shots,
realisic but regarding balance may favour auto fire weapons and skill too much,
dont make something so powerful that it reduces the value of other stuff on same lvl
[Updated on: Wed, 18 April 2018 13:36] Report message to a moderator
|
Master Sergeant
|
|
|
|
|
|
|
Re: Experimental Project 7[message #353280 is a reply to message #332214]
|
Sun, 22 April 2018 21:45 
|
|
crackwise |
  |
Messages:113
Registered:April 2013 |
|
|
Playing now with ja2_7609en+AI_r688 on AR+SDO. I have a couple of observations and requests:
1) Enemy blind-suppression-fire too accurate!
I see that in this AI-version enemy uses blind suppression much more often and very effectively compared to earlier versions. By blind suppression I mean that enemies with automatic weapons return fire at the whereabouts of our mercs without actually having seen them - only by interpreting the sounds of my shots (and muzzle flashes too?). With this version, enemies seem to return blind suppression fire too accurately and would like to ask if it is possible to reduce its accuracy a bit.
2) Visual cues to indicate the approximate location of an enemy necessary: Maybe with a new skill: "Observing"?:
I also use blind suppression fire against enemies I cannot see and I do this by paying attention to the bullet tracers in order to estimate where the enemy is hiding. However, as a player I cannot keep track of all the fires shot at me. This gets very complicated when my mercs are especially getting shot at from different angles. Therefore, I need to have some visual cues indicating me the approximate location of the bullet fired in order for me to compete with the AI-suppression fire. After all, who wins the firefight, wins the actual fight.
I have discovered when I press Shift, it shows a small circle, indicating where the last enemy has shot at me. This is helpful, but shows only one target - and sometimes doesn't show any at all. So I don't know how this thing really functions.
But here is my suggestion: Ideally, what I would like to have is that my merc interprets the direction of enemy shots and I receive visual cues about the approximate enemy location. For instance, those could be indicated with green circles similar to spotting or radio skill circles - the smaller the circle, the more accurately my merc has approximated the location. This feedback could be automatic, or even better, could be made into an active skill as follows: So any merc can be assigned on the tactical to use the skill "Observe". They will be observing and giving feedback about enemy location interpreting enemy blind suppression fire. The more experienced the merc, the better. Deputies and Squadleaders get bonus points, since by profession they know how to guide firefight.
What do you say, is implementing such a thing possible?
[Updated on: Sun, 22 April 2018 21:53] Report message to a moderator
|
Sergeant
|
|
|
|
Re: Experimental Project 7[message #353282 is a reply to message #353280]
|
Sun, 22 April 2018 23:00 
|
|
edmortimer |
  |
Messages:1533
Registered:January 2015 Location: Home Free |
|
|
I second this motion!
Quote:2) Visual cues to indicate the approximate location of an enemy necessary: Maybe with a new skill: "Observing"?:
I also use blind suppression fire against enemies I cannot see and I do this by paying attention to the bullet tracers in order to estimate where the enemy is hiding. However, as a player I cannot keep track of all the fires shot at me. This gets very complicated when my mercs are especially getting shot at from different angles. Therefore, I need to have some visual cues indicating me the approximate location of the bullet fired in order for me to compete with the AI-suppression fire. After all, who wins the firefight, wins the actual fight.
I have discovered when I press Shift, it shows a small circle, indicating where the last enemy has shot at me. This is helpful, but shows only one target - and sometimes doesn't show any at all. So I don't know how this thing really functions.
But here is my suggestion: Ideally, what I would like to have is that my merc interprets the direction of enemy shots and I receive visual cues about the approximate enemy location. For instance, those could be indicated with green circles similar to spotting or radio skill circles - the smaller the circle, the more accurately my merc has approximated the location. This feedback could be automatic, or even better, could be made into an active skill as follows: So any merc can be assigned on the tactical to use the skill "Observe". They will be observing and giving feedback about enemy location interpreting enemy blind suppression fire. The more experienced the merc, the better. Deputies and Squadleaders get bonus points, since by profession they know how to guide firefight.
What do you say, is implementing such a thing possible?
Report message to a moderator
|
|
|
|
Re: Experimental Project 7[message #353286 is a reply to message #353282]
|
Mon, 23 April 2018 02:58 
|
|
townltu |
 |
Messages:384
Registered:December 2017 Location: here |
|
|
Lowering accuracy of enemies blind reply fire may be a bad idea,
i think its intended to force the player to frequent moves, literally every time you shoot.
If reply fire exactly hits the spot, mercs who moved 2 tiles will be rather safe,
while cth penalties or random modifier for spot of noise heard by enemies will result in more bullets hitting merc who moved.
However i would rather prefer an approach which does not more or less force player
to also fire at unseen enmies and always move after a shot,
it favours specific player behaviour and as consequence makes athlete trait op
while high AP weapon users suffer, even more with dynamic aiming costs.
Imagine we get feedback on all noises, 1st "bait" shot results in people shooting all over the sector,
finally team X wins after taking heavy losses, and all happened without any visual contact.
Unfortunately i have not even a layer8 idea how to achieve better enemy tactics,
but i think a random/event_related switch of tactics would probably contribute.
btw a big increase of gravity appeared to work here, as long as weapon range was low enough,
while the reply fire may not care about unseen penalty in cthconstants.ini
Report message to a moderator
|
Master Sergeant
|
|
|
Re: Experimental Project 7[message #353287 is a reply to message #353286]
|
Mon, 23 April 2018 03:41 
|
|
edmortimer |
  |
Messages:1533
Registered:January 2015 Location: Home Free |
|
|
Quote:Imagine we get feedback on all noises, 1st "bait" shot results in people shooting all over the sector,
finally team X wins after taking heavy losses, and all happened without any visual contact.
My point, and I think the original point, was that while the computer can accurately keep track of every noise, every direction of shot, every seen opponent - we cannot because there is too much happening to too many characters, and the screen is moving rapidly from one area to the next. A counter-suggestion would be to dumb down the AI so it cannot remember every noise, every etc. - but we don't want to dumb it down.
Knowing the general direction of enemy fire doesn't necessarily mean shooting blindly in that direction - it can also mean popping smoke, or finding cover from that direction.
I think we are both asking for a little help in remembering which direction (not exact positioning) shots came from regarding a specific PC who would ordinarily know that info if that PC was AI-controlled.
Report message to a moderator
|
|
|
|
|
|
|
|
Re: Experimental Project 7[message #353297 is a reply to message #353296]
|
Mon, 23 April 2018 16:01 
|
|
Deleted. |
 |
Messages:2650
Registered:December 2012 Location: Russian Federation |
|
|
crackwise wrote on Mon, 23 April 2018 17:59I don't know if this is from the AI mod or from something else but another thing I have noticed is that the enemy camouflage seems to be overpoweringly effective!
I have experienced such cases: I hit an enemy with stun grenade, he goes unconscious and falls prone. Then my merc goes to knife him at the spot he has fallen but the stunned enemy has disappeared! I cannot move the merc back or forward since there is an invisible thing blocking the path. I wait a couple of turns and then the enemy gets up and kills my merc! The guy was apparently there all the time, but because he wears camo, he disappears when he falls prone and I can not detect him even if I am almost on top of him zero tiles away. This is very strange, do you know what the reason could be?
Apart from this example, camouflaged enemies seem to have 100% camouflage, so they see me almost always before I see them. Is there a way to reduce or limit enemy camouflage values?
Never experienced that. What camo settings are you using? Maybe you have a save with reproducible bug? (just before the bug appears)
Camo works for enemy just the same as for merc, there's no difference in the code. Some soldiers in Ja2+AI have camo value to allow them hiding better, but it's no different from player having high camo.
[Updated on: Mon, 23 April 2018 16:03]
Left this community.Report message to a moderator
|
|
|
|
Re: Experimental Project 7[message #353300 is a reply to message #353295]
|
Mon, 23 April 2018 17:27 
|
|
townltu |
 |
Messages:384
Registered:December 2017 Location: here |
|
|
sevenfm wrote on Mon, 23 April 2018 14:08...
I was thinking about two possibilities:
1. Game is randomly saved each turn with 1/10 probability in iron man mode, so you cannot save/load every turn but if something goes wrong you don't lose the whole battle.
2. Morale is reduced for mercs on each load on small amount and restores after several turns - you can save\load any time, but if you start saveloading a lot your mercs will not like it and they will start losing combat effectiveness because of morale drop.
I dont like 1. as it could overwrite a correct save with a bug that happened after the last save,
besides who plays JA2 in real IM mode although the game produces a substancial amount of gamebreaking stuff?
Usually i dont like stuff which limits player, also when on the modding side and even if its about exploits,
but 2. sounds great to me, it has little effect on occasional reload due to misclick etc,
but penalizes reload_until_hit & when_will_the_rng_gimme_what_i_want behaviour.
However if required effort is high, it may not be worth it,
as file date, system time or a value in save are not really reliable references.
[Updated on: Mon, 23 April 2018 17:28] Report message to a moderator
|
Master Sergeant
|
|
|
|
|
Re: Experimental Project 7[message #353304 is a reply to message #353301]
|
Mon, 23 April 2018 17:57 
|
|
townltu |
 |
Messages:384
Registered:December 2017 Location: here |
|
|
Massive number of save&reload usually indicates difficulty for player to thik about and/or use different approaches,
however it just came to my mind that 2. may have impact on e.g. playtesting battles in new sector
where frequent reloads are part of the job.
Report message to a moderator
|
Master Sergeant
|
|
|
|
Re: Experimental Project 7[message #353370 is a reply to message #353305]
|
Sun, 06 May 2018 14:16 
|
|
CareBear |
 |
Messages:137
Registered:April 2016 |
|
|
I've got runtime error with something about utf-8. Next time i will try to write down the error log.
Ok happened again, same exact error. Here's log message
Runtime Error
File: \src\Core\UFS_STRING.cpp
Line: 205
Function: ufs:string:as_utf16
invalid utf8 character 'a'=139
Game runs fine without your mod
[Updated on: Sun, 06 May 2018 14:25] Report message to a moderator
|
Sergeant
|
|
|
|
|
|
|
|
|
|
|
Re: Experimental Project 7[message #353663 is a reply to message #353619]
|
Sat, 02 June 2018 12:02 
|
|
Deleted. |
 |
Messages:2650
Registered:December 2012 Location: Russian Federation |
|
|
Ja2+AI r701:
- removed options: UNCOVER_IN_COMBAT, ALERT_ON_DEAD_LIMIT (mercs are not automatically uncovered in combat, instead suspicion counter us used, officers can raise alert if certain percent of soldier in the enemy team is dead, so for silent killing all soldiers, kill officers first)
- NCTH: increased movement penalty
- AI: movement mode tweaks
- Red AI: allow helping friends on roofs and in buildings
- Red AI: min cth = 25% for sniper shot or 2% if soldier is under attack (no more unaimed sniper shots unless AI soldier is under attack)
- bLastID fix (bLastID is incorrectly used in many places in the 1.13 code, resulting in various bugs with last merc in team)
- ClosestReachableFriendInTrouble: don't allow helping from a building to outside, don't allow helping if too many enemies nearby
- AI movement mode tweaks: swatting/crawling on roof or in a building
- spotting: full bonus only when shooting at the same tile (half spotting bonus for first shot, full bonus for second shot at the same tile to make spotting less cheating)
- allow changing orders in town at day time, allow changing orders to ONGUARD for stationary/snipers on roofs when under attack (less restrictive AI roaming)
- SeekRange: increased seek range at day time (as a result, more defending AI soldiers will attack player team)
- AICheckSpotTooFar: increased max roaming range for day/night
- set COVER_SYSTEM_MOVEMENT_EFFECTIVENESS = 0 to prevent problems with AI (movement penalty for cover makes calculating cover at destination point unpredictable)
- r8564 fix: lua function CheckMiscFlags1 checks the wrong flags (by Flugente)
- r8562: If a gun's currently used ammotype has <numberOfBullets> > 1, display the gun damage in EDB in 'bullet damage X bullets fired' format, taking <multipleBulletDamageMultiplier> / <multipleBulletDamageDivisor> into account. (by Flugente)
- improve fix for ignoring invisible opponents when showing path (the fix still doesnt work correctly for prone opponent/merc)
- r8566: NPCs with profile > 170 have missing approach values, making it impossible to fulfill opinion checks (by Flugente)
- soldier tooltips update: show bUnderFire, ubPreviousAttackerID
- fix BG_PERC_PRICES_GUNS, BG_PERC_PRICES backgrounds
Most important changes:
- big penalty for aiming at invisible target
- less restrictive roaming limits for AI
- sniper will only shoot with good CTH
- harder to shoot at moving target in NCTH
- spotter bonus reduced for first shot
- various fixes
[Updated on: Sat, 02 June 2018 18:57]
Left this community.Report message to a moderator
|
|
|
|
Re: Experimental Project 7[message #353678 is a reply to message #353663]
|
Mon, 04 June 2018 15:30 
|
|
Deleted. |
 |
Messages:2650
Registered:December 2012 Location: Russian Federation |
|
|
Ja2+AI r708:
- no instant merc death feature improved (mercs will die more often, it's easier to kill dying merc now)
- restored normal jam chance (was greatly lowered compared to stock 1.13)
- use Flo discount only if backgrounds disabled
- r8557 fix: sound for drinking water played twice (by Flugente)
- r8553 fix: bad radar/overhead map colours for creature team. (by Flugente)
- r8559 fix: creature attack variables are not always cleaned up properly (by Flugente)
- r8544 fix: If 'forced turn mode' is active, sometimes enemy groups already present in an enemy-occupied sector are interpreted as pending reinforcements, causing them to spawn at the sector edges instead of inside the sector. (by Flugente)
[Updated on: Mon, 04 June 2018 15:44]
Left this community.Report message to a moderator
|
|
|
|
|
|
|
|
|
Re: Experimental Project 7[message #353799 is a reply to message #353798]
|
Fri, 15 June 2018 20:14 
|
|
R-TEAM |
Messages:3
Registered:January 2015 |
|
|
It is only happen on the start of an new char - then maybe list it as know bug, and the user should create/start an new game with the english version, after this he can switch back to the german version ...
The root of the problem is maybe in german chars (ä,ö,ü,ß...) ?
Regards
Report message to a moderator
|
Civilian
|
|
|
Re: Experimental Project 7[message #353800 is a reply to message #353799]
|
Fri, 15 June 2018 20:20 
|
|
Deleted. |
 |
Messages:2650
Registered:December 2012 Location: Russian Federation |
|
|
R-TEAM wrote on Fri, 15 June 2018 22:14It is only happen on the start of an new char - then maybe list it as know bug, and the user should create/start an new game with the english version, after this he can switch back to the german version ...
The root of the problem is maybe in german chars (ä,ö,ü,ß...) ?
Regards
Just make IMP in stock 1.13 r7609 exe, then import it into Ja2+AI by typing name at IMP screen. No need to switch language versions.
The problem is definitely related to German localisation, but it's hard for me to debug non-English versions of the game.
[Updated on: Fri, 15 June 2018 20:20]
Left this community.Report message to a moderator
|
|
|
|
Goto Forum:
Current Time: Fri Apr 18 10:47:08 GMT+3 2025
Total time taken to generate the page: 0.18614 seconds
|