DescentBB

Ok, so playing Descent 2 a bit recently brought something to my attention. Namely, the Lou Guards are pitifully weak compared to D1's viral juggernauts the red hulks. It's not just due to their difference in stats, which are superior for the hulk in terms of both endurance and aggression, but due to the fact that the homing missile wielded by them has been significantly nerfed between D1 and D2.

First thing you'll notice is that on direct hit, a robot type homing missile does 26 damage, while in D1 (where I'm assuming the hulks used the same programmed variety as player), they did 43 damage on direct hit. They were also considerably more difficult to evade if memory serves, though to be fair in D1 you'd jump the hulk most of the time and destroy it before it can even fire, or just facetank it under the effects of invulnerability. In D2 though, it doesn't even take afterburners to evade a salvo from the Lou Guard (and the red hulk too, since both use the same weapon).

I've also noticed a discrepancy in power. In the data, a robot homing missile has the strength of 30, while it reliably does only 26 damage to your ship. On the other hand, the player homer from D1 does 43 damage, and in D2 its strength is listed as 40 only. Does that mean a nerf in player homer power between the two games?

Also, what do you think about adjusting the red hulks for D1,5 so that they fire player homers instead of robot homers, to put them in an altogether different threat category than the Lou Guards?

No knowing about the first three paragraphs, but the last one is one of those truly great ideas.

I did some tests and the results confuse me, to say the least - they make no sense at all. First off, the player homing missile is just as good as it was in D1, if not better. In D1 it would take 6 homers to blow up a super hulk, in D2 it takes 5. That would suggest they've actually become stronger in the sequel, of course apart from their turn rate (that's a whole other topic, amirite).

And yet, player homers (as wielded by the super hulk) do 43 damage to the player in D1, and 30 in D2 - even though in the latter their power rating in the data is 40. There must be some damage scaling algorithm at play in D2, because even on Insane, player homers when used by robots only do 4 more damage on direct impact, and have a projectile speed of 90, as opposed to the variable speed of the robot homer, capping off at 80 on Insane (it's 60 on all other difficulty levels).

I have no idea why a missile with the listed power of 40, only does 30 damage to the player. Might be a bug from the nightly versions of D2x-Rebirth, so testing under a stable version might be required, but this gave me a whole new idea:

Maybe we should give the player homer to the Lou Guard instead - it's only marginally more damaging and marginally harder to dodge on Insane than the default robot variety. Of course that would make the robot a bit tricky on Trainee etc, since player weapon values generally don't change throughout the difficulties - but then again the robot so dopey even on Insane that the player is usually able to pick it off before it even gets to fire once, so Trainee players will also probably be fine.

And then, we'd fine-tune the robot homing missile instead, bringing it to values identical to D1 on Insane, and then give it to the red hulk. That way we'd accomplish basically the same thing while sacrificing very little. Either that or we find out why the heck player homers do much less damage than they're stated to do.

EDIT: Just tested this - when fired by the player at a wall point blank, the homers do the same amount of damage in both D1 and D2 - around 31-32 points. That's more than getting hit by a player homer from a robot! There must be some mistake in the code, seriously. I'll test it on a DOSBox installation and see.

Aaaaand nothing changes. Even on a vanilla installation the homers do 30 damage to the player, while they clearly should do more.

Never overestimate Trainee players. Skill generally scales up with difficulty level but it's not necessarily the same in reverse. When you get down to Trainee especially people are probably playing it for a reason.

Ehh so I guess we'll just leave things as they are. We can compensate by stuffing twice as many super hulks as there were in D1 xDDDDDD

I'll also ask around on the DXX boards, plenty of coders there so they might know what's up

I can try to look up the code for the particular issue with homing missiles. At least on weekends, I doubt I will find time for it today or tomorrow.
Sadly I cannot recompile and run it on Windows. But looking on it can be enough. I also wanted to write the robot and weapon data reader for myself (I don't think it will take more than an hour), haxmedt32 still does not work for me (I cannot install it, asks for Tahoma font in system folder), probably it is the problem with installer that works only for Win95. BTW are you using Rebirth 0.58 or this later version (nightly build)?

I think making one of the robots (red hulk, or lou guard), fire stronger missiles is a good idea. Why have two robots that are basically the same? BTW you can give the robot guided missiles, like LL L16 boss has, they do 54 damage or so on Insane. Of course you will have to be careful with using the robot, to avoid Bahagad Outbreak issue, where levels were filled to the brim with robots firing 7 merc missiles per volley and many of them in snipe mode (10 missiles per volley).

I'm using the nightlies currently, but the same issue persists under dosbox, so it's not a Rebirth thing. I think I used a dmtools installer way back when, anyway my haxmed32 does work. Keep trying to run it, it's really worth it. Just do what the messages say. That said, it would be interesting to have a stat reader program, especially if it could also read the stats from D1. That way I could easily tweak stuff in descent2.HAM.

As for the super hulk solution, right now I'm just leaning towards leaving the player homers unchanged, and actually buffing the robot homers. That would of course affect both the lou and the hulk and they would indeed remain very similar in function, but the lou would still fire smaller volleys, be more mobile and less durable. I'm afraid there is no way to make these bots much different from each other without excessive tinkering (I want the game to play as vanilla/as D1 as possible, and nothing more). My foremost concern was higher damage in the first place. If both the robots are to do 43 damage instead of 26, then so be it

Yeah, Descent does some additional hard-coded scaling for weapon damage from/to robots, and in D2 I recall hearing there were some further tweaks on certain difficulty settings (Trainee) to make it even easier. I believe missiles and Gauss do double damage to robots or something.

Haven't really looked for myself to see exactly what all the hacks are, though. But it's definitely not all in the HAM.

Goodness gracious, complications everywhere

I looked at the code, some of my findigs are
1) Missile damage is computed as "splash damage", which is weapon power multiplied by 1-"dist to explosion point"/"explosion radius" and explosion, or collision point is
the contact point between two spheres (ship and missile) both have nonzero size. Same formula applied to Gauss shots. That's why real damage from missiles is always lower than base damage, and can also vary (for example I know that green spider can be killed with a guided missile but sometimes it can survive).
There is also a Multiplayer multiplier for weapon strength (D2 only). But this does not affect SP or Coop.
2) There is some vague hack with collision point computation for D1, comments say that it is not computed correctly in all cases, and I doubt I can figure it out only by reading the code. Apart from this hack, damage computation seems to be identical in both games.
3) Are you sure that red hulk fires PLAYER homers in both D1 and D2? I will look at it myself, i thought it uses robot weapons that scale damage with the difficulty level. At least in D1. And I still did not looked up the homer damage in D1/D2 weapon stats.
4) I still did not went into the robot/player weapon issues, apart from different stats they also use different handling code and it also can cause this.

My haxmedt32 is a part of Robot Construction Kit, which require win95 to install, probably. So I cannot use it. I will maybe try to find dmtools but for now I will try to write my own tool that simply dumps the robot/weapon stats into the text file.

Update : rbotedit shows that in D2 both lou guard and red hulk use the same ROBOT type missile, and I cannot view the stats in D1 with it, I even don't know where they are stored, from the code it seems they are in the PIG file. But I found no viewer for it. And I cannot even find dmtools over the web...

But, when you shoot your own missile into the wall right in front of you, conc and homer do same damage to your ship (25 and 32) in D1 and D2. But probably direct hit will do a bit more damage. Anyway it is possible that robot homing missile changed from D1 to D2.

I only assumed that in D1 red hulks used player homers, simply because of the damage. They definitely use robot homers in D2, same as Lou Guards. That's why I'm confused as to why they do 43 damage in 1 and only 26 in 2. I though if the two games' codebases were getting unified for Rebirth, there would be a single algorithm to handle this, but we might have to wait a bit.

As for haxmed32, it really is more convenient to have around, not sure where else to get it though. It did scream for some DLLs at first too, but I installed them and now it works. I could upload my entire dmtools folder for you to get, but not sure if that would work.

When working with descent manager tools always install the dll packs, most of them won't work without them.

Xfing wrote: ↑Sun Dec 10, 2017 2:59 pm I only assumed that in D1 red hulks used player homers, simply because of the damage. They definitely use robot homers in D2, same as Lou Guards. That's why I'm confused as to why they do 43 damage in 1 and only 26 in 2. I though if the two games' codebases were getting unified for Rebirth, there would be a single algorithm to handle this, but we might have to wait a bit.

As for haxmed32, it really is more convenient to have around, not sure where else to get it though. It did scream for some DLLs at first too, but I installed them and now it works. I could upload my entire dmtools folder for you to get, but not sure if that would work.

Yes, please upload dmtools if you can, and any other packages it uses if possible, I don't see why it should not work on my system if it works on yours. BTW does it work with D1 stats, too? And do you know where the stats are stored in D1? BTW D1 red hulk homers are ROBOT version (scale damage with difficulty), I tested it in L10 red hulk on various difficulties, on Insane one missile does 40 damage, Ace - 25, Hotshot - 16. Also varies a bit even on direct hit (+-2-3 points on Insane). And your own homer does the same damage to you on Hotshot to Insane. So yes, it seems you can just bump up the ROBOT homer damage for your mission, and probably give the Lou guard player version which does less damage. But remember it will make it much harder for low difficulty players.

EDIT : on Rebirth code unification : the code I've seen looks pretty "unified" to me, the bulk of the code being the same for D1/D2, with some conditional compilation sections for D1 and D2, that are controlled by single compile time switch. So probably they mean just further polishing it, unfortunately every developer knows polishing software takes infinite time but we like it. Still there are differences in data structures and AI algorithms, and some others, most of them probably cannot be unified because of backward compartibility requirements for mission data, saved games, etc. And of course weapon and robot data can be (and is) different for both games...

Yeah I'd give it a download to see if it worked too. Never got the installer to work on any of my systems, but, that doesn't necessarily mean the programs wouldn't.

As far as I know, robot data in D1 is stored in descent.pig. Haxmed32 doesn't let you access that, it only works with descent2.ham. The only way to read D1 robot data is to see the source code (most probably). Haxmed32 itself is rather limited, it doesn't support creating new item and robot classes within descent2.ham, you can only edit existing positions. That means you can't create another category of robot homers to give one to the red hulk and one to the lou guard - they still have to share the only existing one.

Here's my installation of dmtools, it should have everything you need bar the DLLs. Those can be easily obtained online and the applications tell you where to put them, so just take it slow and get all the dlls you need.
https://www.sendspace.com/file/b2sfzd

Thanks but it still does not work, first it asked msvbm50.dll, I downloaded it, then it gave "stif3 not properly installed" error, running as admin as you suggested in the old thread does not help, BTW the message is entirely correct... Maybe smth is still missing, or I should install the StIf, too?
About new items and robots, I don't think it possible because they are fixed-size arrays in the code - you can only modify existing entries... Still it would be interesting to make a simple reader/writer program for D2 ham and also D1 pig files. Two diferent formats (both documented) make it twice as difficult, however.

Yeah, you need STIF too, but it appears to be included in the package. Maybe use the REG files to add registry entries? They're found in the Shell subdirectory.

Did some calculations and tests today. For one, I compared 43 to 26, which gave me a 1,6538 factor by which one is bigger than the other. If you multiply 30 by this factor, you get a number that neatly rounds up to 50. That means that the robot homer power on Insane was set to 50 in D1. I tested D1 on Ace then, and I consistently received 26 damage from homer strikes, same as in D2 on insane. That means the progression went 30-50, as opposed to 25-30 that you have in D2's files. The jump is admittedly pretty steep, so I can't blame the designers for wanting to reduce it - playing D1 with no deaths relied specifically on never being hit by a single homing missile from a hulk, let alone a volley. Reducing the damage to 30 on Insane let the player absorb an entire 3-missile volley without it spelling instant doom for them.

On the other hand though, D2 had levels with so many open spaces that dodging homing missiles even without the afterburner is rather trivial. In D2 your only hope were stealthy attacks with homers of your own from behind corners, stunlocking the hulks while cloaked, or being invulnerable. In the second game you can just jump the Lou Guards with Gauss and dispatch them before they can even fire, and dodge their volleys with trivial ease to such a degree that with more play you'll start to disregard them as top tier threats, something you'd never do when it comes to D1's red hulks. For that reason I think bringing back robot homers for both these robots to their D1 values would definitely be a good balance decision, and that's what I'll do for the D1,5 HAM. Time to make players fear homing missile shooting robots again, like they would in the past!

EDIT: I've discerned that the values for D1 were 15-15-20-30-50. In D2 they're listed as 10-15-20-25-30. To preserve D2's extreme Trainee handicap, I'll keep the newer Trainee value, so their final shape will be 10-15-20-30-50. Of course I'll do the same thing for concussion missiles for consistency. I'll leave Mercuries alone because there was no D1 precedent to judge them against for one thing, and their traveling speed makes them troublesome enough as it is. What I would like to do but cannot is also set the missiles' flying speeds to the same as they had in D1 - I suspect they were a bit faster on Insane after all. Without access to actual values from D1 it would be guesswork though. Of course I may be mistaken and the missile speeds may not have been altered in D2 at all.

EDIT 2: Robot concussions in D1 are 10-15-20-30-50. In D2 they are 8-12-18-23-30. I'll be modifying them to 8-12-20-30-50.

Not succeeded with haxmedit still, no, reg files in STIF folder are just file associations and don't help a lot, probably I should modify the ini files (obviously there should be some kind of installer for that but it seems to be missing), I tried to do it by hand (to make sure the Descent/Descent2 folders are correct at least), still the same error...
But I hadn't too much time for that. As well as for coding my own D2/D1 data reader. Will get to it on New Year holidays, surely. BTW be careful with bumping up concussions in D2, green spiders firing 6x50 damage conc would be deadly.

AlexanderBorisov wrote: ↑Wed Dec 20, 2017 8:29 pm Not succeeded with haxmedit still, no, reg files in STIF folder are just file associations and don't help a lot, probably I should modify the ini files (obviously there should be some kind of installer for that but it seems to be missing), I tried to do it by hand (to make sure the Descent/Descent2 folders are correct at least), still the same error...
But I hadn't too much time for that. As well as for coding my own D2/D1 data reader. Will get to it on New Year holidays, surely. BTW be careful with bumping up concussions in D2, green spiders firing 6x50 damage conc would be deadly.

Didn't we have the same with missile platformers in D1 though? Well ok, I checked and missile platformers only fire 5 in a volley on Insane. But their firing wait is only 1 second, whereas the spider's is 1,25 secs, which still gives the platformer slightly higher damage over time (not that it should matter, since taking a single volley is game over anyway ) Also, the platformer is 1,5x tougher than the spider. The spider's mobility may be an issue, but that's up to the designer not to put any in tight corridors I guess.

As for Haxmed32, make sure to run it as admin. I get an error myself when I try to open it with no admin privileges. If not, I think there might be an installer in that big archive I got from someone here. "germany1.download.descent-network.com.7z". It has so much different stuff in it, though to be fair, I didn't install DMTools from anything found inside it, i had it from some other source. I'll look harder, maybe there'll be some installer.

https://www.sendspace.com/file/x9h3mw

Here's my folder with DMTools installer packages. I think Haxmed32 might have been installed by one of these. Feel free to try them out.

Most likely the admin issue is because it tries to write to files in protected locations (probably the root directory of the drive). That was totally allowed in Windows 9x, but when Vista brought in UAC it started clamping down on the practice.

I have UAC turned off, and of course always work under admin (Win 7 64). Only if it needs "SYSTEM" account which only can be used through hacks...
Wait a bit, I'll try the new installers however...
EDIT : in fact, you need an installer in DMTools to set correct paths. I first run "dp2003.exe", of course said "no" when it asked me to overwrite some system DLLs with earlier versions; then vpview13, it installed stif, then run haxmedt32 (it asked me the path to Descent). Everything works now, big thanks, Xfing!

Yay, glad it worked out! Now just remember to use that trick I described earlier to enable descent2.ham editing.

Ok, I saved it for later, now I am ok with only viewing it. BTW it still does not show some info like ship and weapon sizes (that are used in collision computation for example), also we still don't know how to view D1 data so writing a reader/writer program is still useful...
Some interesting info on weapons - actually mercury missile (player) is the same speed (400) as robot gauss (player Gauss is faster and stronger), and Vulcan (which does not have robot version), is 500 speed - a bit faster... And ROBOT Gauss is said to do 16 damage where in fact it does 4 or less, that means 4/16 = 1 - (player ship size / damage_radius), where damage radius for ROBOT gauss = 7, so player ship size should be around 5. Would be interesting to check it. Anyway interesting thing that I observed that if you move fast sometimes ROBOT gauss can do 1 or even zero damage even if you see the explosion near you! At least in Rebirth. Never happens with Player gauss (sometimes given to bosses), which has 10 damage radius (700 speed and 25 strength).
And, in fact, Mega flash missile fired by those Vertigo MAX and Obsidian IDR has the same damage radius as a Mega missile! And is 105 speed, faster than robot homer (80 speed on Insane). And your flash missiles have the same blast radius (80) and a bit higher speed (110) in fact.

Yeah, the flash missile blast radius is often overlooked, but quite an important detail in head-to-head games because you can often use it to deal damage around corners.

In fact you ship size is ~4.7, it is "rad" property of polymodel. Another interesting thing, your missiles deal twice as much damage to robots. For example you kill 270 HP Lou guard with 3 guided (damage 55 each). That is because it does both direct hit damage + splash damage (it is clear from the code) to robots, very strange! Where it does only splash damage to human player... BTW direct damage from Gauss to bosses is reduced (2x on Insane) via special code hack! About dealing damage around corners, at least there is a check that splash damage is applied only when there is direct sight between missile hit point and the target.
Edit - I also found why missiles and Gauss do varying damage. In fact collision (between you and the projectile) is detected on the line connecting your position on current frame and the next frame, and for computing distance to collision point, your current position is used! Which implies different distance to explosion point (can be lagr or smaller than your ship's radius). This way you can get zero damage from ROBOT gauss if you are moving fast, or plus/minus 10% damage variation for homing missiles.

Some more findings on red hulks and their homers in D1 vs D2. I looked at homing missile algorithm in D1 and D2 and found it to be the same. Not even affected by any physics in fact (that used mass, drag, thrust, etc). Every frame homing missile direction (velocity vector) is rotated halfway toward the target, while keeping it's length. NEWHOMER flag in Rebirth makes this running at fixed FPS (which controls the difficulty of dodging the missile in fact). Yes, that's far from how real homing missiles work in real life... And your only chance to avoid the missile is to get out of the acquisition cone which is defined by max angle between missile velocity and vector to the target. Which can only happen when the missile is very close to you and you are moving quickly in perpendicular direction. This angle in D1 is around 40 deg, in D2 twice less. That's why homers in D1 are twice harder to dodge. But I checked (again played a bit with the red hulk in the start of FS L10) that even in D1 homers are perfectly possible to dodge, if you trichord past them at the right moment. But start too early and missile hits your back!

Two more things that make D1 red hulks different (and harder!) even from the same robot in D2. 1) It cannot be stunlocked in D1, the robot can both rotate and fire when you attack it! Where in D2, it cannot do anything when you got it with Vulcan or Plasma. 2) D1 version seems to have much faster turn rate, in D2 you could get at the back of the robot easily if you manage to avoid it's first volley of missiles, and I tried circling D1 version at the closest distance possible, he manages to keep up with you even when you are firing at it! And in rare case even manages to fire a missile.
Those two things make the "dodge and kill" tactics against Red Hulk in D1 unpractical - very few places apart from L10 where you can do it...

Interesting observations! I didn't know the acquisition cone was reduced by half in D2. But that's indeed what my experience reflects too, that's why homers in D2 are so pitifully easy to dodge. I would argue with the stunlocking part though - the effect does diminish on every subsequent difficulty level, but Vulcan always gets the job done if memory serves me right. Plasma doesn't, but quad laser 4s also do let you stunlock a red hulk due to their mass, which knocks them around - that's why lasers are best to use when the mech is in an alcove. Just verified it too - Vulcan does stop insane red hulks from firing in D1.

I asked on the Rebirth boards whether restoring the 40 degree cone will be one of the elements of the codebase unification, or at least toggleable. Sure would be nice to be able to make the life even harder for yourself in D2, wouldn't it? Haha

In my version D1X 0.58.1 even continuous Vulcan fire does not prevent Red hulk from repeatedly firing missiles at you when it sees you. Even when you are cloaked! It does decrease it's turn rate significantly however, so if you attack it from the back you have a chance to kill it before it completes the turn and sees you. EDIT : same in the last 0.60 weelky version, too... And for some reason in the latest D1X version cheats don't work. Should report a bug, probably... Can you record a demo of the stunlocked red hulk if possible - seems interesting... Mine is here, it managed to fire 4!!! times while I was killing it. But it is not less than 2, always. Seen your last post later than submitted mine, checked the quads too (0.58, where cheats still work) - robot firest once or rarely twice, I think this is because the quads are more powerful and kill it faster, not because of the stunlock. I am also happy they did this change to D2!
https://www.sendspace.com/file/0ccvh6
Also I found that you don't even need to trichord to dodge the homing missiles in D1, just have right movement angle to them. And LOL in the 0.60 you can simply fly for minutes with bi-chording and never get hit at all!!
https://www.sendspace.com/file/5nrg9i
Seems like another bug to report, it was not possible in earlier versions, where it should be done like this (trichording only, and with bichording the missiles will eventually get you):
https://www.sendspace.com/file/0hvej7

I'll check it in dosbox then and see what happens in the vanilla game.

EDIT: Indeed, not even vulcan can entirely stop red hulks from firing in d1. Which of course is no good, since even one missile of theirs takes a huge chunk out of your shields. In some cases a hulk won't fire even one missile while under fire until destroyed, but most of them will get off at least half a volley. I'm glad they made this particular change in D2 though.

Quad lasers are the most reliable stunning tool in D1.

I remember the D1 strategy guide saying the best strategy was vulcan and being prepared to take one hit. I've always preferred either plasma + ducking around corners or charging them from behind with spreadfire and missiles.

My preferred strategy was to hit them with both a primary and a secondary at the same time. Even if the stun lock isn't perfect and they can eventually overcome it, they don't have the HP to survive long enough to get any shots in vs any primary other than low level lasers if you unload a couple concs at the same time.

Worst surprise I know of with red hulks is in some situations they can fire missiles behind themselves. Idle hulks can fold up in D1, nothing worse than trying to catch one off guard by opening a door behind a red hulk and having it still fire a whole barrage at you.

fun fact: in the playstation version of d1 the red hulks can shoot at you no matter what direction they are facing

Ok, well that one part is definitely not something we'll be tweaking. I'll settle for 50 damage on Insane and D1's 40 degree acquisition cone, not sure how to achieve the latter though.

Xfing wrote: ↑Sun Dec 31, 2017 5:08 pm Ok, well that one part is definitely not something we'll be tweaking. I'll settle for 50 damage on Insane and D1's 40 degree acquisition cone, not sure how to achieve the latter though.

Probably setting the acquisition cone isn't possibel without code modifications (it is hardcoded value, but different for D1/D2), and I am even happy thet you cannot set the wider cone - missiles turning 90 deg instantly and hitting you from the back definitely look like smth OP. About robots firing missiles from their back, yes I've seen it too.

But the 20 degree cone is way too easy to dodge, like all you need to do is to accelerate and move to the side, not even trichording is necessary, let alone the afterburner.

Maybe a 30 degree compromise? (If possible, of course)

LightWolf wrote: ↑Mon Jan 01, 2018 4:11 pm Maybe a 30 degree compromise? (If possible, of course)

Why do you care? You never test the levels you populate anyway. Waves of super hulks in a single room are tough even with the current cone

Fair point