So, I'm officially taking requests for testing for those who perhaps want to see which equipment combination fares better in combat - all tests are 1v1 unless stated otherwise. You can be as specific about the equipment as you want, since I've made a quick devmode modification that allows me to be very specific with what gets spawned (e.g. only spawn excellent golden knives), so setup is much easier there.
Of course, I'll only accept requests that are reasonably easy to test in a controlled environment, at least for now. You can also specify any deviations from the standard testing methodology that's shown below (e.g. skill levels to test, cover etc.).
Tests conducted so far & results:
Build 1958:
Melee vs. Megaspider (https://ludeon.com/forums/index.php?topic=42359.msg416772#msg416772) || Longsword/Spear vs. Scyther (https://ludeon.com/forums/index.php?topic=42359.msg416772#msg416772) || Mace/Spear vs. Scyther/Centipede (https://ludeon.com/forums/index.php?topic=42359.msg416788#msg416788) (meta until build 1968) || Sniper/Bolt-Action/Charge Lance vs. Lancer (https://ludeon.com/forums/index.php?topic=42359.msg416841#msg416841) || Level 14 Shooter w/ Masterwork Power Armour + Masterwork Charge Lance/Rifle vs. Centipede (https://ludeon.com/forums/index.php?topic=42359.msg416841#msg416841) || Level 8 w/ Longsword vs. Level 15 w/ Knife (https://ludeon.com/forums/index.php?topic=42359.msg416841#msg416841) || Charge Rifle vs. Machine Pistol w/ different armour types (https://ludeon.com/forums/index.php?topic=42359.msg416867#msg416867) || Knife vs. Bolt-Action/Revolver/Autopistol w/ varying armour and skill combinations (https://ludeon.com/forums/index.php?topic=42359.msg416921#msg416921)
Build 1959:
Level 7 & 14 Melee w/ different Trait and Armour combinations (https://ludeon.com/forums/index.php?topic=42359.msg417040#msg417040) || Excellent Power Armour + Plasteel Longsword (w/ and w/o Joywire) vs. Two Megaspiders (https://ludeon.com/forums/index.php?topic=42359.msg417082#msg417082) (1v2) || Longsword/Spear/Mace vs. Scyther/Centipede/Eachother update (https://ludeon.com/forums/index.php?topic=42359.msg417159#msg417159) || Longword/Spear/Mace scaling over various armour types (https://ludeon.com/forums/index.php?topic=42359.msg417396#msg417396)
Build 1960:
Plasteel Longsword/Spear and Uranium mace scaling over various armour types (https://ludeon.com/forums/index.php?topic=42359.msg417550#msg417550)
Build 1969:
Mace/Longsword/Spear vs. Scyther/Centipede Update #2 (https://ludeon.com/forums/index.php?topic=42359.msg419896#msg419896) || Mace/Longsword/Spear vs. Centipede (https://ludeon.com/forums/index.php?topic=42359.msg419919#msg419919) (2v1 and 3v1)
Standard Testing Methodology
General:
* All pawns are trait-neutral (i.e. no tough people, wimps, brawlers, careful shooters or trigger-happies), and healthy adults
* All pawns will have level 14 shooting and melee skills
* All equipment will be of normal quality
* There are walls between each fighting duo so that nobody else gets hit by any stray shots
* Trials with either combat-relevant inspirations or trauma savantism (unless the pawn is already down) aren't counted, and thus redone
Weapon Testing:
* Armoured testing will be done with flak pants, a flak vest, a steel simple helmet and a flak jacket
Apparel Testing:
* Ranged testing for will be done with assault rifles at a distance of 20 cells
* Sharp melee weapon testing will be done with steel gladii
* Blunt melee weapon testing will be done with steel maces
* The terrain is concrete, so there's no dirt kicking
Results are determined by 10 trials, with each trial being 10 1v1 battles happening simultaneously, so each set of results has a sample size of 100 battles. Scores per trial are simply determined by how many people on each team are left standing. These are logged in an excel spreadsheet, and the final product will look something like this (click to expand):
(https://i.imgur.com/9xVuFn9.png)
This is what the actual test setup looks like:
(https://i.imgur.com/EDb3OMH.png)
Bug testing :D
Megaspider vs pawn with (all poor) plasteel gladius, steel plate armor, simple helmet,
Same, only with (good)
Same with power armor, power helmet
Same with plasteel longsword
Also, since I'm lazy, just a straight up stat comparison of plasteel plate vs power armor, all normal. ::)
You the man, Dan.
Anyway, the rounds:
Melee spear v scyther
Melee longsword v scyther
Melee longsword v melee spear
Unless I am misunderstanding things, will this be viable?
Yep, both sound pretty good to test!
Greep - I'll probably also throw in a run with level 8 melees and plasteel gladii, along with the 'standard' level 20 procedure, just to paint a more realistic picture. Edit: Also, power armour and power armour helmet + plasteel longsword will only be done with good quality since that too would be more realistic, plus otherwise the test run count will get pretty damn high.
And look if Brawler trait maybe made a different.
Ah yeah I was just meaning the four tests, with the first only being poor. Thanks :D
Quote from: Greep on July 08, 2018, 08:42:31 AM
Megaspider vs pawn with (all poor) plasteel gladius, steel plate armor, simple helmet,
Same, only with (good)
Same with power armor, power helmet
Same with plasteel longsword
Here she is!
(https://i.imgur.com/7Q25tt3.png)
I was tempted to skip testing the longsword + power armour since gladius + power armour already had a perfect run, but tested for the fun of it anyway.
As for stats (sharp / blunt / heat):
Plasteel plate armour - 92 / 46 / 78.8
Power armour - 117 / 45 / 54
Both have the same amount of coverage
Quote from: Ser Kitteh on July 08, 2018, 08:44:06 AM
Melee spear v scyther
Melee longsword v scyther
Melee longsword v melee spear
I also threw in some results for level 8 melee with spears since 8 melee is a realistic amount of skill to have. Standard testing methodology was applied:
(https://i.imgur.com/dzOnOox.png)
The spear results are quite interesting, performing significantly better than the longsword against mechanoids. It's probably down to the fact that spears hit internals way more often than longswords... like artificial brains. I'm now intrigued to see what performs better against scythers: spears or maces.
Quote from: Canute on July 08, 2018, 08:54:14 AM
And look if Brawler trait maybe made a different.
Anything specific to test?
Quote from: XeoNovaDan on July 08, 2018, 10:02:38 AM
Quote from: Canute on July 08, 2018, 08:54:14 AM
And look if Brawler trait maybe made a different.
Anything specific to test?
I just thought about the request people asked before.
And maybe a similar test like the Megaspider, but instead vs a Centipede.
stuff stuff discord
Anyway while waiting for your maces v spears analysis, I shall suggest another:
The issue with sniper rifles is that they take far too long, and the niche of long range is very suspect and I have been suspectful of them ever since I got my hands on one back in A14.
I don't know exactly how you would test damage output because the pros of long range is the pawn is relatively safe. So how about:
pawn w/sniper vs lancer
pawn w/bolt-action vs lancer
pawn w/charge lance vs lancer
Alternatively even pawn vs anybody will do, I am far more interested how consistent the damage output is. It could be a naked tribal for all I care.
ooooooooo.
Charge lance vs centipede compared with charge rifle vs centipede. Masterwork quality on both, full masterwork power armor on both as well. Range for both: max charge rifle range. 14 shooting
Mace vs. spear against mechanoids done, and it looks like we've found the new meta, ladies and gents - for SirKitteh and Canute:
(https://i.imgur.com/sCk0XBz.png)
Although no sane person would dare 1v1 a centipede with a somewhat mediocre melee person in average armour, it actually seems to be viable due to how effective spears are. This is most likely because of the fact that centipedes have very few internal parts, and spears have a tendency to target internals. The most noteworthy internal part which seems to give spears such an edge against mechanoids is the artificial brain.
There was a slight setup change for centipede testing too: I had to force everybody into 2x1s with the centipedes so that the centipede actually stuck to melee rather than attempting to use their ranged weapons. All mechanoids had the 'berserk' mental state so that they were guaranteed to attack.
Also bear in mind that these are only normal quality steel spears too; a higher quality plasteel spear would be even more effective!
For brawlers, it'd basically just be these results but adjusted upwards slightly.
Quote from: Ser Kitteh on July 08, 2018, 10:28:15 AM
pawn w/sniper vs lancer
pawn w/bolt-action vs lancer
pawn w/charge lance vs lancer
Quote from: Greep on July 08, 2018, 10:49:13 AM
Charge lance vs centipede compared with charge rifle vs centipede. Masterwork quality on both, full masterwork power armor on both as well. Range for both: max charge rifle range. 14 shooting
Going to test both of these next. Just figured I'd get this out first since this is really interesting.
Wow, spears the new superior weapon, even against Centipede with their high piercing armor.
So the next colony will be a full mellee colony, and everyone with brawler skill can get recruited but jogger prevered ! :-)
Three more tests onto the list!
Quote from: Ser Kitteh on July 08, 2018, 10:28:15 AM
pawn w/sniper vs lancer
pawn w/bolt-action vs lancer
pawn w/charge lance vs lancer
Results:
(https://i.imgur.com/PFTg4ZL.png)
Only 7 lancers were willing to co-operate for the sniper testing; 6 would fire at 28 cells, and the 7th would fire at the charge lance's maximum range of 36 cells. As a result, there's only a sample size of 70 battles for the sniper testing. Mechanoid ranged testing is definitely harder to do!
As for the bolt-action rifle and charge lance testing, only 8 lancers were co-operating, so they both have sample sizes of 80 battles.
Quote from: Greep on July 08, 2018, 10:49:13 AM
Charge lance vs centipede compared with charge rifle vs centipede. Masterwork quality on both, full masterwork power armor on both as well. Range for both: max charge rifle range. 14 shooting
Results:
(https://i.imgur.com/iIYHsep.png)
Only 8 out of the 10 centipedes were willing to actually fight, so there's only a sample size of 80 battles for minigun and heavy charge blaster testing.
Do note that the inferno cannon testing also didn't have any cover, and there was a 'backboard' that the centipedes could occasionally 'bounce' off of - flawed testing, I know, but didn't realise until too late and didn't feel like re-testing inferno cannons. Inferno cannon testing was definitely the hardest one to do yet since it's one I had to micromanage to make sure that pawns kept as close to the charge rifle's max range as possible (unless their movement was really bad), and so that they didn't spend most of their time extinguishing fires. As a result, inferno cannons only have a sample size of 20 battles.
Having some hard cover (e.g. walls), or having firefoam poppers would dramatically improve these results.
Quote from: Ser Kitteh (RimWorld Discord)i am interested then, how well an 8 melee with a longsword
does against a 15 melee with a knife
Finally, finishing off with an easy one:
(https://i.imgur.com/eydxYoY.png)
This was with armour. Standard methodology.
Well you'd be insane to engage a centipede with an inferno cannon without a shield decoy or turret decoy anyways, I was mainly looking to see how the lance piercing fares. Looks like the rifle is still better which is good :)
Indeed! Also, changing up the methodology slightly: future tests will be run with level 14 pawns if skill levels aren't specified instead of level 20s. This is since level 14 is a much more realistically sustainable skill level, so future tests will be more representative of what'll happen with actual gameplay.
Just done a new set of tests for the Charge Rifle vs. Machine Pistol, following some discussion on another thread that I made:
(https://i.imgur.com/hqglSrV.png)
At 8 cells, the machine pistol generally has a slight advantage over the charge rifle owing to its superior touch accuracy, which soon fades as armour gets factored in - the tables even turn when power armour is involved, but I'm not sure how many people would use machine pistols with power armour anyway.
At 19 cells (machine pistol's max range), the charge rifle always has the lead, which extends as its AP advantage kicks in with higher armour ratings.
Hmm, if this is too hard to set up, no worries:
Joywire + excellent power armor/helmet + plasteel longsword vs 2 megaspiders
compared with
same but without joywire.
Quote from: Greep on July 08, 2018, 04:27:43 PM
Joywire + excellent power armor/helmet + plasteel longsword vs 2 megaspiders
All excellent, right?
This might involve some microing, but the results could be interesting.
- 7 melee, no armor, normal steel knife vs melee bolt-action rifle, no armor
-14 melee, no armor, same as above.
-7 melee, full flak (normal), same as above (bolt-action using flak)
-14 melee, full flak (normal), same (bolt using flak)
-7 melee, power armor (normal), same as above (vs bolt-action with power armor).
-14 melee, same.
All tests involve the bolt-action wielder to shoot as the melee pawn charges in (another test with them not shooting could be interesting).
Objective: how effective are long-range firearms as melee alternatives given the melee combat changes, and how armor affects those changes.
Quote from: Razzoriel on July 08, 2018, 05:32:32 PM
- 7 melee, no armor, normal steel knife vs melee bolt-action rifle, no armor
-14 melee, no armor, same as above.
-7 melee, full flak (normal), same as above (bolt-action using flak)
-14 melee, full flak (normal), same (bolt using flak)
-7 melee, power armor (normal), same as above (vs bolt-action with power armor).
-14 melee, same.
My setup should easily handle this type of testing! Do you also want the level 7 melee runs to also have level 7 shooters, and 14 for 14?
Quote from: XeoNovaDan on July 08, 2018, 05:41:23 PM
Quote from: Razzoriel on July 08, 2018, 05:32:32 PM
- 7 melee, no armor, normal steel knife vs melee bolt-action rifle, no armor
-14 melee, no armor, same as above.
-7 melee, full flak (normal), same as above (bolt-action using flak)
-14 melee, full flak (normal), same (bolt using flak)
-7 melee, power armor (normal), same as above (vs bolt-action with power armor).
-14 melee, same.
My setup should easily handle this type of testing! Do you also want the level 7 melee runs to also have level 7 shooters, and 14 for 14?
Yes, please. Id also test with revolvers and finally a melee-heavy gun. Werent miniguns equipped with a 10-damage melee attack? These three tests could potentially say a lot about melee, the steel knife being the absolute basic melee weapon. Given how the new meta is shaping up to be the spear, would be quite interesting with that, too. Not sure how long this would take, so baby steps for now.
TMK most, if not all ranged weapons have had their melee DPS normalised to ~4, which is inferior to fists. Revolver and bolt-action seem reasonable to test. 8 sets of trials in total, but most variation can be done through some very quick save editing which is easy enough. Probably gonna be my last one for the day since it's 10:55 PM here at the time of posting this!
Might also throw a non-staggering weapon in there (autopistol) too, to see ranged stagger's actual effect in results.
edit: sorry, 12 trials. Ah well, if I could do that with bearskin vs flak jacket, I'm sure I can do it again!
New testing request: Nimble vs Brawler vs Tough for melee effectiveness.
- 7 vs 14 melee, both naked, both steel knife, normal.
- 7 vs 7 melee, nimble vs no traits, naked, normal steel knife
- 7 vs 14 melee, nimble vs no traits, naked, normal steel knife
- 7 vs 7 melee, nimble vs brawler, naked, normal steel knife
- 7 vs 14 melee, nimble vs brawler, naked, normal steel knife
- 14 vs 7 melee, nimble vs brawler, naked, normal steel knife
And another two sets of the above, one set with both using flak, other set using power armor. Will ask for interactions with Tough later.
I'm betting tough > nimble > brawler
Inb4 centipede guts next patch
Quote from: Razzoriel on July 08, 2018, 05:32:32 PM
- 7 melee, no armor, normal steel knife vs melee bolt-action rifle, no armor
-14 melee, no armor, same as above.
-7 melee, full flak (normal), same as above (bolt-action using flak)
-14 melee, full flak (normal), same (bolt using flak)
-7 melee, power armor (normal), same as above (vs bolt-action with power armor).
-14 melee, same.
Done and dusted, with the autopistol thrown on top to see how the shooters fare without ranged stagger on their side. The biggest difficulty was probably how to actually automate this in the first place, and I came to the conclusion that forcing all melee pawns to go manhunter did the trick :P
Save editing with weapon and skill level changes also made this a much quicker process.
Knife vs. Bolt-Action Rifle:
(https://i.imgur.com/Y9Hwqr0.png)
It seems that this set of results is actually the set that least favours melee. I was honestly expecting the revolver's fire rate to have it pull ahead, but it seems that the bolt-action rifle's raw damage and range actually proved superior. They both have ranged stagger, but I guess a bolty shot to the leg is going to stand out that bit more.
Also, the fact that the gap is wider at higher skill levels, probably since the shooter will get more hits in, meaning that the brawler's going to be pretty weak, slow and inaccurate once they actually reach the shooter.
Knife vs. Revolver:
(https://i.imgur.com/OjuuvoT.png)
Similar results to the bolt-action, except with a narrower delta. The flak results are interesting though, with level 7 flak melee actually beating level 7 flak gunner. I did do some extra (admittedly unrecorded) testing to check that this wasn't a streak of luck, and this does indeed actually seem to be a legitimate set of results. Strange.
Knife vs. Autopistol:
(https://i.imgur.com/kljvwWV.png)
Unsurprisingly, owing to the autopistol's lack of ability to cause ranged stagger, it's almost 50-50 as to which side wins, with the only exception being with power armour. Skill level has relatively little effect here, with the deltas between levels 7 and 14 basically being within margin of error.
I'm done for the day; it's 1:42 AM here in the UK and I'm tired. Will do more testing later on!
Quote from: XeoNovaDan on July 08, 2018, 05:16:12 PM
Quote from: Greep on July 08, 2018, 04:27:43 PM
Joywire + excellent power armor/helmet + plasteel longsword vs 2 megaspiders
All excellent, right?
Yup, all excellent. I'm assuming most will survive two megaspiders that way, if not, you can tweak the melee skill/quality so it works out.
Very cool info!
Eventually it might be worth making a mod that automates this kind of testing.
Also FYI, weapons are still being rebalanced.
The spear debate might be improved if the higher your Melee the more likely you are to spear an organ.
Of course that's an issue if raiders had the same rule applied as well.
Quote from: XeoNovaDan on July 08, 2018, 11:21:07 AM
Mace vs. spear against mechanoids done, and it looks like we've found the new meta, ladies and gents - for SirKitteh and Canute:
(https://i.imgur.com/sCk0XBz.png)
Although no sane person would dare 1v1 a centipede with a somewhat mediocre melee person in average armour, it actually seems to be viable due to how effective spears are. This is most likely because of the fact that centipedes have very few internal parts, and spears have a tendency to target internals. The most noteworthy internal part which seems to give spears such an edge against mechanoids is the artificial brain.
There was a slight setup change for centipede testing too: I had to force everybody into 2x1s with the centipedes so that the centipede actually stuck to melee rather than attempting to use their ranged weapons. All mechanoids had the 'berserk' mental state so that they were guaranteed to attack.
Also bear in mind that these are only normal quality steel spears too; a higher quality plasteel spear would be even more effective!
For brawlers, it'd basically just be these results but adjusted upwards slightly.
I'm not convinced frankly. Your conclusion is based on spears being more likely to target internals and.. I know of no mechanical reason for this to be the case.
The conclusion that spears are better on average than maces vs centipedes is literally based on unlikely hit locations. You can find these hit chances in game - The chance to hit the head is 15%, artificial brain 0.75%.
If the spears don't have this assumed trait then your conclusions could just be the result of ridiculously unlikely variance across a sample size.
If you do the math on maces vs spears it's very obvious that the difference in average dps of the mace is much higher owing primarily to the difference in relative armor penetration values.
As mechanoids cannot be incapacitated through pain, they must be incapacitated by reducing their Movement or Conciousness(data processing) to ~0%.
ALL centipedes are more likely to go down to unlikely head hits irrespective of the weapon used owing to the fact that damage to the body rings is so relatively small owing to their massive hitpoint pools, it's the nature of their effective-toughness. A hit to the brain is worth ~20x more than a hit to a body ring.
For your conclusion to be correct the spear
has to be more likely to hit the artificial brain.
Can you actually support that with a piece of code that alters the likelihood of internals being damaged on a weapon by weapon basis? Or even a by-damage-type basis?
I'd be keen to find out - because frankly after playing with a melee colony for ~7 years I'm of the opinion that maces absolutely decimate spears even post attack frequency re-factor, especially vs centipedes.
Quote from: Boboid on July 09, 2018, 03:51:39 AM
...Can you actually support that with a piece of code that alters the likelihood of internals being damaged on a weapon by weapon basis? Or even a by-damage-type basis...
<DamageDef>
<defName>Stab</defName>
<label>stab</label>
<workerClass>DamageWorker_Stab</workerClass>
<externalViolence>true</externalViolence>
<deathMessage>{0} has been stabbed to death.</deathMessage>
<hediff>Stab</hediff>
<hediffSolid>Crack</hediffSolid>
<impactSoundType>Slice</impactSoundType>
<armorCategory>Sharp</armorCategory>
<stabChanceOfForcedInternal>0.6</stabChanceOfForcedInternal>
<stabPierceBonus>2.5</stabPierceBonus>
</DamageDef>Done. This in Damages_MeleeWeapon.xml. See 'stabChanceOfForcedInternal' two lines up from the bottom. A spear mostly stabs, so it has a ~40-50% chance of hitting the artificial brain with each attack since that's the only 'internal' part mechanoids have.
And to back this XML up with some decompiled code (using dnSpy):
protected override BodyPartRecord ChooseHitPart(DamageInfo dinfo, Pawn pawn)
{
return pawn.health.hediffSet.GetRandomNotMissingPart(dinfo.Def, dinfo.Height, (Rand.Value >= this.def.stabChanceOfForcedInternal) ? dinfo.Depth : BodyPartDepth.Inside);
}This is from Verse.DamageWorker_Stab.
If you're still skeptical, I can even do some recordings, as video evidence that this is real.
Quote from: Tynan on July 09, 2018, 12:59:04 AM
Very cool info!
Eventually it might be worth making a mod that automates this kind of testing.
Also FYI, weapons are still being rebalanced.
Yeah, automation would definitely be even easier. I don't think my C# skills are that point yet though. :P
For now, a harmony transpiler, a really basic thought workerClass that gives 1000 mood to everyone and some save editing's all that this is going off of!
Quote from: XeoNovaDan on July 09, 2018, 03:57:35 AM
Quote from: Boboid on July 09, 2018, 03:51:39 AM
...Can you actually support that with a piece of code that alters the likelihood of internals being damaged on a weapon by weapon basis? Or even a by-damage-type basis...
<DamageDef>
<defName>Stab</defName>
<label>stab</label>
<workerClass>DamageWorker_Stab</workerClass>
<externalViolence>true</externalViolence>
<deathMessage>{0} has been stabbed to death.</deathMessage>
<hediff>Stab</hediff>
<hediffSolid>Crack</hediffSolid>
<impactSoundType>Slice</impactSoundType>
<armorCategory>Sharp</armorCategory>
<stabChanceOfForcedInternal>0.6</stabChanceOfForcedInternal>
<stabPierceBonus>2.5</stabPierceBonus>
</DamageDef>
Done. This in Damages_MeleeWeapon.xml. See 'stabChanceOfForcedInternal' two lines up from the bottom. A spear mostly stabs, so it has a ~40-50% chance of hitting the artificial brain with each attack since that's the only 'internal' part mechanoids have.
And to back this XML up with some decompiled code (using dnSpy):
protected override BodyPartRecord ChooseHitPart(DamageInfo dinfo, Pawn pawn)
{
return pawn.health.hediffSet.GetRandomNotMissingPart(dinfo.Def, dinfo.Height, (Rand.Value >= this.def.stabChanceOfForcedInternal) ? dinfo.Depth : BodyPartDepth.Inside);
}
This is from Verse.DamageWorker_Stab.
If you're still skeptical, I can even do some recordings, as video evidence that this is real.
Very interesting! And no I don't require video evidence - I can interpret that code just fine :P
Do you happen to know where I can find the information regarding the % chance of any given attack type being used? Spears for example have 3 attack types in addition to the fists/dust throwing that pawns do irrespective of their armed status, but I haven't got a clue how prevalent each attack type is.
Quote from: Boboid on July 09, 2018, 04:20:15 AM
Do you happen to know where I can find the information regarding the % chance of any given attack type being used? Spears for example have 3 attack types in addition to the fists/dust throwing that pawns do irrespective of their armed status, but I haven't got a clue how prevalent each attack type is.
I don't know if there's actually any logging for this; I went through and skimread all of the dev mode logging options and there doesn't appear to be any, but I have attached a melee DPS calculator spreadsheet that I never got around to finishing though. I've confirmed that the calculation side of it is fully functional and accurate for B18, but some weapon values will need tweaking for 1.0, and then there's the issue with it then not factoring in those terrain tools you mentioned.
I don't think there's been any significant changes in terms of standard DPS calculations either, so the formulas should still be accurate. The one exception would be the armoured formula due to the armour system's revamp though, but the new system should be quite easy to put into a formula.
[attachment deleted due to age]
Hrm bummer. It would be useful to know how often a pawn is swinging with their fists instead of bean-filled-whack-bonk.
Would make my analysis of melee altering prosthetics a bit more accurate.
Oh well!
Test on sir! You're doin' god's work. Or at least statistically relevant testing, which is basically the same.
TIME FOR REPEAT DAN
Melee spear v scyther
Melee longsword v scyther
Melee longsword v melee spear
Quote from: Razzoriel on July 08, 2018, 05:58:42 PM
- 7 vs 14 melee, both naked, both steel knife, normal.
- 7 vs 7 melee, nimble vs no traits, naked, normal steel knife
- 7 vs 14 melee, nimble vs no traits, naked, normal steel knife
- 7 vs 7 melee, nimble vs brawler, naked, normal steel knife
- 7 vs 14 melee, nimble vs brawler, naked, normal steel knife
- 14 vs 7 melee, nimble vs brawler, naked, normal steel knife
Whew, this one took a while to test! 180 trials (1,800 battles!) in total, maybe given a little since I did 20 before build 1959 dropped, and I had to scrap a few because of trauma savantism (which I should probably disclose; it's always been part of the methodology to scrap trauma savant runs unless the trauma savant was already in pain shock).
Also, since this is such a large amount of data, I've decided to present this all in a stacked column chart:
(https://i.imgur.com/Rtc4vWg.png)
Not the prettiest since I tried following a tutorial on doing the grouping but Excel wasn't playing ball, so I did what any sane person would do and manually do the group labels using GIMP. Not perfect, but it at least seems legible enough!
For those that would prefer to see just the tables though, they are also available.
Naked results:
(https://i.imgur.com/GqVTHrj.png)
Flak results:
(https://i.imgur.com/4NnYfXV.png)
Power armour results:
(https://i.imgur.com/r2BmqXW.png)
Next up:
Quote from: Greep on July 08, 2018, 04:27:43 PM
Joywire + excellent power armor/helmet + plasteel longsword vs 2 megaspiders
compared with
same but without joywire.
Not really a 1v1, but I guess I'll do 1v2s as well. Seems easy enough.
Quote from: Ser Kitteh on July 09, 2018, 08:45:52 AM
TIME FOR REPEAT DAN
Melee spear v scyther
Melee longsword v scyther
Melee longsword v melee spear
I doubt the meta's changed but sure, I guess it'll be interesting to see how things have changed with the CD rebalancing! :D
We're planning to make a debug output that shows all the melee attack tools/modes for each weapons, race, and hediff. And their selection weights.
Selection weights are basically not tuned because we don't have this. I think they favor low-damage things too much.
E.g. if you have a 20 damage attack and a 10 damage attack, currently it uses the 10 damage 33% of the time (selection chance proportional to damage). Should be more like 10%.
Ah right, that's very interesting.
I imagine that causes some odd interactions too.. I should probably do the math on this but - Adding the first power claw is probably really, really efficient compared to the second.
There's what... Four? Standard unarmed melee attacks? Both fists, biting, throwing dust? Or is throwing dust handled separately?
Fist, fist, bite, headbutt.
Dirt kick actually comes from the terrain. You can't do it on e.g. tile floor.
So melee verbs can come from races, weapons, hediffs, and terrain! A dude with a beer and a power claw on dirt can headbutt you, bite you, punch you, claw you, smash you with his beer, or kick dirt in your face.
Quote from: Greep on July 08, 2018, 04:27:43 PM
Hmm, if this is too hard to set up, no worries:
Joywire + excellent power armor/helmet + plasteel longsword vs 2 megaspiders
compared with
same but without joywire.
Done:
(https://i.imgur.com/aXPDv01.png)
And in table form:
(https://i.imgur.com/1K76Gq1.png)
Quote from: Ser Kitteh on July 09, 2018, 08:45:52 AM
TIME FOR REPEAT DAN
Melee spear v scyther
Melee longsword v scyther
Melee longsword v melee spear
Next set of testing to do. Gonna also throw mace vs. scyther/centi results in like last time, but all with level 14 melee pawns (standard methodology), and it'll be graphed.
[Edit: Type derp fix]
Quote from: Tynan on July 09, 2018, 10:30:08 AM
We're planning to make a debug output that shows all the melee attack tools/modes for each weapons, race, and hediff. And their selection weights.
Selection weights are basically not tuned because we don't have this. I think they favor low-damage things too much.
E.g. if you have a 20 damage attack and a 10 damage attack, currently it uses the 10 damage 33% of the time (selection chance proportional to damage). Should be more like 10%.
What about if the 10 damage attack had a higher DPS than the 20 damage attack? Say a 1s CD vs. a 3s CD on the 20 damage? Probably not too relevant to vanilla weapons, but I reckon there are probably some strange modded weapons that'd do something similar. I guess it'd then just be a matter of adjusting the tool commonality anyway.
Quote
A dude with a beer and a power claw on dirt can headbutt you, bite you, punch you, claw you, smash you with his beer, or kick dirt in your face.
What a talented individual! I can barely juggle.
Totally forgot about headbutts, hrm I wonder what value dirt kicking is assigned when calculating its weight.
And I wonder what influence the lack of dirt has on combat. So many questions!
Should I be paving the world when using melee weapons? Is it more or less efficient vs gun users if they have dirt but I don't? What about centipedes!? My mind races with impractical stupidity!
My devolution into insanity aside. I look forward to mucking about with melee colonies every time there's a change - They offer such a stark and interesting contrast. And the general viability of melee makes combined arms that much more engaging.
I will say that knowing the weight of each melee attack is reasonably important information for a player to have access to - although I'll admit that it could easily take someone out of the experience to see that their pawns headbutt enemies 12% of the time or what have you. Although I guess they're already looking at the gritty, behind-the-curtain damage calculations anyway.
Anyway the min-maxer in me would appreciate it, who knows about everyone else.
A Stab's tendency to target internal organs would also be kind of handy to know in-game.. as would a description of whatever this part of the cut mechanic is doing :P
-<cutExtraTargetsCurve>
-<points>
<li>0, 0</li>
<li>0.6, 1</li>
<li>0.9, 2</li>
<li>1, 3</li>
</points>
</cutExtraTargetsCurve>
<cutCleaveBonus>1.4</cutCleaveBonus>Quote from: XeoNovaDan on July 09, 2018, 11:25:09 AM
What about if the 10 damage attack had a higher DPS than the 20 damage attack? Say a 1s CD vs. a 3s CD on the 20 damage? Probably not too relevant to vanilla weapons, but I reckon there are probably some strange modded weapons that'd do something similar. I guess it'd then just be a matter of adjusting the tool commonality anyway.
Edit: Interesting question. Really highlights how heavily the attack weighting affects the overall melee system.
Quote from: XeoNovaDan on July 09, 2018, 11:25:09 AM
What about if the 10 damage attack had a higher DPS than the 20 damage attack? Say a 1s CD vs. a 3s CD on the 20 damage? Probably not too relevant to vanilla weapons, but I reckon there are probably some strange modded weapons that'd do something similar. I guess it'd then just be a matter of adjusting the tool commonality anyway.
Yes. I'm modding all polearms to have a 3-damage, 1s attack which involves striking with the butt of the pole. This is very relevant.
Quote from: Ser Kitteh on July 09, 2018, 08:45:52 AM
TIME FOR REPEAT DAN
Melee spear v scyther
Melee longsword v scyther
Melee longsword v melee spear
And re-ran for build 1959!
Chart for Longsword, Spear and Mace vs. Scyther and Centipede:
(https://i.imgur.com/MxzKD8z.png)
Chart for Longsword vs. Spear, Longsword vs. Mace and Spear vs. Mace:
(https://i.imgur.com/t5y9vBx.png)
If anybody is wondering why these results differ quite a bit from the last run (other than the build update), it's because all of these runs were tested with level 14 melee pawns across the board, rather than some level 8 and some level 20.
Just like last time, everybody's also wearing flak pants, flak vests, flak jackets and steel simple helmets.
As always, I'm also showing the tables which give a per-trial score breakdown too. These are sliced differently because the slicing had to be adapted to cut down the graph count.
Battles against Scyther:
(https://i.imgur.com/ktNu4sE.png)
Battles against Centipede:
(https://i.imgur.com/Fi6TRbl.png)
Weapons against eachother:
(https://i.imgur.com/47uXfBk.png)
Quote from: Boboid on July 09, 2018, 11:29:26 AM
...a description of whatever this part of the cut mechanic is doing :P
-<cutExtraTargetsCurve>
-<points>
<li>0, 0</li>
<li>0.6, 1</li>
<li>0.9, 2</li>
<li>1, 3</li>
</points>
</cutExtraTargetsCurve>
<cutCleaveBonus>1.4</cutCleaveBonus>
I looked into the code for that, and this is my understanding of what those particular bits of XML affect in terms of mechanics:
cutExtraTargetsCurve - This appears to be a percentage-based chance curve which determines how many extra parts a successfully-landed cutting attack will actually hit. The game randomly generates a float between 0 and 1, and based on what that number and curve are, the cut will land x amount of extra hits (e.g. if 0.75 is rolled, hit either 1 or 2 extra parts, both with equal probability).
cutCleaveBonus - Looks like some sort of damage amount retained/given determined by how many parts the cutting attack hits. When a cut hits multiple parts, its damage amount per part is actually reduced, but if cutCleaveBonus is greater than 0, the weapon will effectively do more damage than normal.
The formula that determines how much damage each affected part takes is this:
finalDamage = baseDamage * (1 + cutCleaveBonus) / (affectedPartCount + cutCleaveBonus )So say that an attack from a longsword hits and does 23 damage (normal quality steel), and you got really lucky and that strike actually hits 3 extra parts. The cutCleaveBonus is 1.4, so the damage amount will be determined by this:
finalDamage = 23 * (1 + 1.4) / (4 + 1.4)So each affected part takes around 10.22 damage, meaning that the affected pawn can take as much as roughly 41 damage - ouch! Compare that to the original amount of 23, and that really is quite a significant jump. For reference, it takes 64 points of damage to incapacitate a healthy adult human from pain with most attack types (including cuts).
CutCleaveBonus is a multiplier.
A 10-damage attack that hits two body parts with 0 cutCleaveBonus deals 5 damage to each part.
A 10-damage attack that hits two body parts with 0.5 cutCleaveBonus deals 7.5 damage to each part.
Interesting, I was expecting joywire to make a lot more of a difference, but I guess it just matters more for shooting since that's per-tile accuracy. Nice!
I'll be honest, I'm somewhat surprised at the the almost half/half results for the spear/longsword/mace thing. I'm not sure how to balance it myself, but it's clear that maces should be buffed in a way. Considering that longswords' Thing is cutting, I don't personally mind it being a pawn v pawn thing.
I've also come to equip every melee fighter with a spear. It has gone, very very well!
Alright, the big thing:
Powerclaw vs scyther
Powerclaw vs centipide
Powerclaw vs spear
Not sure about dual power claws or single power claws either. You do you!
I'm not super surprised that most of the melee weapons came out reasonably equal in pawn v pawn battles frankly.
Maces don't have the raw dps but their ability to Stun more consistently is a huge deal.
Just from personal experience I've noticed that a single stun is often enough to end a battle right there and then simply because the stunned pawn will take so much damage without being able to fight back. Humans are so flimsy that the stun time relative to combat time is reasonably long.
I'd be interested to see how the weapons compare in pawn v pawn fights when they're wearing power armor or plate armor. Especially now that Uranium maces have so much armor penetration relative to the blunt armor values of plate/power armor.
The fact that longswords and spears so consistently win against scythers/centipedes is kind of bonkers. Those stabs really are kicking the crap out of them. It's fairly unintuitive frankly.
It's such a finicky thing to balance though, a tiny buff can push a weapon over a threshold where it generates a positive feedback loop. Damaging the enemy so badly that it can't fight back effectively.
Time is obviously a huge deal as well - how long it takes to murder something is often very important, and obviously how dangerous that thing is during that time. Spears obviously sniping important mech parts is shortening fights considerably.
Obviously if you gib an enemy immediately and consistently when it enters melee combat it's 0% dangerous :P
Quote from: Boboid on July 10, 2018, 07:06:39 AM
...
Yeah, indeed. I think I've already beaten you to the punch there since I've just finished some testing regarding how the longsword, spear and mace scale over different levels of armour (all steel unfortunately, but this should still be an indication). I also threw in a bonus set of runs to see how they fare against level 7 melee pawns with assault rifles, but said pawns also being high on go-juice (basically your average elite mercenary), but that testing was still done purely in melee combat. No plate armour was involved though since plate armour is already on par with flak gear when it comes to melee combat.
In fact, here are the results (bonus runs are on a separate chart):
(https://i.imgur.com/OWV1JNR.png)
One thing that's definitely interesting is that the spear scales better than the longsword over increasing levels of armour, but this is also unsurprising in another sense since the spear does have superior AP.
Bonus run results:
(https://i.imgur.com/JZQ5tO5.png)
And, as always, the tables are available too, but sliced differently.
Naked test results:
(https://i.imgur.com/wzcswdH.png)
Flak test results (taken from a previous test, still build 1959 though):
(https://i.imgur.com/pTudPRJ.png)
Power armour test results:
(https://i.imgur.com/qwpC4Zb.png)
Bonus test results:
(https://i.imgur.com/d412ynP.png)
Quote from: Ser Kitteh on July 10, 2018, 06:52:33 AM
Alright, the big thing:
Powerclaw vs scyther
Powerclaw vs centipide
Powerclaw vs spear
Not sure about dual power claws or single power claws either. You do you!
I'll do both single and dual.
Quote from: Razzoriel on July 09, 2018, 03:09:16 PM
CutCleaveBonus is a multiplier.
A 10-damage attack that hits two body parts with 0 cutCleaveBonus deals 5 damage to each part.
A 10-damage attack that hits two body parts with 0.5 cutCleaveBonus deals 7.5 damage to each part.
Do you know maybe if weapon's with that bonus can hit multiple target's (infront) too ?
Then i would see at last a good advance from longsword vs spear.
And i think Xeo should use mace's out of uranium. Special when he use plasteel for longsword/spear.
And maybe test how club's out of jade are compared to them. Since i can't play the unstable i can't check that myself.
Quote from: Canute on July 10, 2018, 09:09:48 AM
...
Nope, longswords can't hit multiple targets at once; only multiple body parts on a single target. All tests have been conducted with steel weapons since that's the standard methodology; no plasteel involved other than in requests that specifically mention the use of plasteel weapons.
I could do a variation of the tests I've done with the weapon materials you've mentioned though.
I'd be fairly interested to know how plasteel sharp weaponry compares to uranium maces - Given that the AP scales with damage I think that uranium maces might actually have the edge they're supposed to vs heavily armoured targets.
Plasteel only increasing sharp damage by 1.1 compared to uranium's 1.5 for blunt is a fairly big deal.
I imagine that vs lightly armoured targets Plasteel sharp weaponry will be even more dominant due to its attack speed. Although who knows what breakpoints that comparison will bring up, perhaps the maces might now be capable of gibbing a vital part that they were previously merely wounding.
The steel v steel tests in power armor look almost right, except for spears. God damn spears :P Were they wearing power armor helmets?
Quote from: Boboid on July 10, 2018, 09:33:16 AM
...
Yep. Power armour and power armour helmets. If you mean by how spears scale much worse against people than they do against mechanoids, that'd simply be because humans have many more internal body parts than mechs.
Looks like this test's going to be a high priority one! Right now though, I'm just getting stuck in with modding, so the testing will start later.
I mostly meant that Spears seem to be over-performing vs heavily armoured targets when compared to maces.
Also I hate to be the bringer of bad news but:
Quote
Melee attacks which deal higher damage are now much more likely to be chosen in melee combat.
So.. that's probably invalidated most of the melee data :'(
Edit: If Tynan happens to stick his head back into here: Any chance we could know what the new attack weighting system is?
Quote from: Boboid on July 10, 2018, 09:57:58 AM
Edit: If Tynan happens to stick his head back into here: Any chance we could know what the new attack weighting system is?
I think it's just weight=power^2 (it used to be weight=power). But I haven't reviewed it. Ison implemented it but I still need to go over it. It'll probably change some more, I have a task to review more aspects of the melee system.
Quote from: Tynan on July 10, 2018, 10:11:50 AM
Quote from: Boboid on July 10, 2018, 09:57:58 AM
Edit: If Tynan happens to stick his head back into here: Any chance we could know what the new attack weighting system is?
I think it's just weight=power^2 (it used to be weight=power). But I haven't reviewed it. Ison implemented it but I still need to go over it. It'll probably change some more, I have a task to review more aspects of the melee system.
Is Commonality still there so we can make certain attacks more common despite their power?
Quote from: Boboid on July 10, 2018, 09:33:16 AM
I'd be fairly interested to know how plasteel sharp weaponry compares to uranium maces - Given that the AP scales with damage I think that uranium maces might actually have the edge they're supposed to vs heavily armoured targets.
Plasteel only increasing sharp damage by 1.1 compared to uranium's 1.5 for blunt is a fairly big deal.
I imagine that vs lightly armoured targets Plasteel sharp weaponry will be even more dominant due to its attack speed. Although who knows what breakpoints that comparison will bring up, perhaps the maces might now be capable of gibbing a vital part that they were previously merely wounding.
Slacked a little today since I was instead occupied by modding things, but I did manage to get the testing done!
Chart:
(https://i.imgur.com/o22eDY6.png)
It's odd that this didn't scale in the same way that the other chart did - possibly down to margin of error? Considering that melee's still very subject to change though, I don't think the lack of super thorough skew-checking is really relevant.
Once it's confirmed that melee's implementation is final or isn't expected to get any further changes, I'll probably do a batch of runs with a 300 sample size per equipment combination rather than the standard 100. 2,700 tests would be a lot to do, but I've been getting somewhat more efficient with this when I actually get around to doing it.
Also interesting is that plasteel spear wielders actually have a higher DPS than the longsword wielders, yet the longswords outperform the spears until power armour. This is probably due to the fact that longswords are simply better at dishing out pain, which compensates for their DPS deficit (9.78 for a level 14 pawn wielding a longsword vs. 10.45 with a spear).
Tables as always...
Naked:
(https://i.imgur.com/1L52WPn.png)
Flak:
(https://i.imgur.com/HItBE3q.png)
Power Armour:
(https://i.imgur.com/81pxu9T.png)
Huh. So with more armor, now maces are much better? Understandably due to mechanoid's few organs and spears' pierce strike, they're still preferable to longsword/mace. I actually don't mind using specific weapons for specific enemies. You're still gonna down a tribal/bug/mechboi with a plasteel melee weapon.
XeoNovaDan, can you explain or share the devmode modification that allows you to spawn pawns with specified equipment, etc?
Quote from: topace3000 on July 12, 2018, 03:55:30 AM
XeoNovaDan, can you explain or share the devmode modification that allows you to spawn pawns with specified equipment, etc?
It's just a quick harmony patch. Alas it doesn't quite allow for spawning pawns with specific equipment, but it does allow for spawning weapons and apparel with specific quality and stuff when using settings like 'try place near', or 'spawn apparel'. You'd configure the quality and stuff through mod settings.
I've attached it if you also want to play around with it, with the source code included.
[attachment deleted due to age]
You still taking requests?
I'd like to see Thrumbo horn vs some misc stuff. Cents/Sythes/people.
I got one on a trader really early on in my tribal game and my melee pawn is wrecking house with it. Most hits on normal raiders are one shot done kinda stuff. It would be nice to see just how much more effective than spears and stuff it really is or if its actually less so against mechs for example.
I have somewhat burnt out on 1v1 testing over the past week, so I haven't really done any testing over the past week or so. However, I did decide to revisit the Melee vs. Mechanoid topic since mechanoids now have a bunch of internals and the 'stabbing' damage type has been nerfed, as well as the reworked tool weights since I last tested.
No charts for this one since I'm not motivated to make them, but tables are still here since they're easy to just put into image form. That being said, roll on the results.
Melee vs. Scyther
Build 1959:
(https://i.imgur.com/ktNu4sE.png)
Build 1969:
(https://i.imgur.com/hvay84e.png)
Melee vs. Centipede
Build 1959:
(https://i.imgur.com/Fi6TRbl.png)
Build 1969:
(https://i.imgur.com/p09HXHQ.png)
Note: I did forget to normalise the quality for centipede weapons for the 1969 testing, and only noticed this after gathering two sets of results; 7 centipedes have normal quality weapons and the other 3 have poor quality weapons. As a result, the results are technically slightly skewed in favour of the people.
Interesting data, thanks for this.
For something like the centipede it might be more meaningful to test 2 on 1 or 3 on 1. It's designed to be too powerful to beat 1 on 1.
Good catch. Motivation has picked up somewhat, at least for now, so I'll probably test both! ^^
Edit: Also changing things up a bit and actually removed the ranged verbs from the centipede weapons, so they'll be 100% engaged in melee combat rather than often attempting to engage in ranged combat even though the setup never allows them to actually use their weapons at range. Results could be considered an absolute worst case scenario in that sense
Alright, results with the 2v1s and 3v1s are in. Since I did normalise the quality this time around and also cut the ranged verbs for this testing, these results naturally won't be directly comparable to the 1v1 ones.
3v1 also had a smaller sample size of just 70 battles compared to the standard 100 for 2v1, since I only have 20 pawns in the setup and didn't feel like 'binning' too many new pawns, so I just spawned in a 21st who was trait-neutral, melee-capable and healthy.
2v1 Results:
(https://i.imgur.com/8MQ0QZR.png)
3v1 Results:
(https://i.imgur.com/7b4yKMy.png)
Edit: I mistakenly put 6-1 for the final trial with the longsword in the 3v1; only just noticed this as I was going to get the link to link to this from OP. Final trial was actually 7-0
It's somewhat surprising that maces still output consistently worse results, but less so at the same time since mace tools have a lower selection weight due to the exponential commonality with damage.
My first instinct was to assume that the stun nerf had a significant impact on maces vs centipedes. It was of course justified - Stunning humans was almost always a death sentence. But now that I think about it centipedes were already hard to stun. I assume it's down to their heads having ridiculously large amounts of health.
Adding additional internal hit locations might have reduced the chances of stunning a bit as well.
Uranium vs plasteel will hopefully even up the playing field considerably. At the point where maces are essentially totally ignoring armor they ought to pull ahead. I'm not totally confident though.
Although, to be perfectly frank from a practical gameplay perspective Uranium is a real pain in the ass to obtain until reasonably late game. You'll always have access to some plasteel from mechs and it's fairly common in exotic goods traders which you can request.. Uranium not so much. Fortunately demand for it is comparatively low since maces require less material to construct than spears or swords but.. still.
It's a weird game state to be in where only uranium maces were really worth making, and you can't make them until reasonably late most of the time.
A uranium v. plasteel retest is certainly something to consider, indeed. Uranium maces will probably perform much better than their steel counterparts from the fact that their tools are more likely to be selected alone, and then the raw DPS bonus on top of that.
I think stun only has a minor effect. Probably plasteel sharp weapons have enough AP to reduce the effective sharps to small values where the superlinear scaling makes it insigificant, and outweighed by the higher overall DPS. Another possibility is that even with the extra organs, a significant fraction of deaths still comes from destruction of vital parts, which I think can now come from the reactor as well as the brain.
Quote from: XeoNovaDan on July 19, 2018, 05:57:00 PM
A uranium v. plasteel retest is certainly something to consider, indeed. Uranium maces will probably perform much better than their steel counterparts from the fact that their tools are more likely to be selected alone, and then the raw DPS bonus on top of that.
Hmm.. instinct makes me think that plasteel melee weps will be quicker to wield due to their light weight (& 'sharpness') while uranium will do more dmg per hit but be significantly slower.. does this make sense from a gameplay PoV & does it carry to 'real' GP combat?
How would the two wep mats compare vs (say) a centipede, with significant sharp but low blunt resistance, & vs a fully-armoured pawn (in assorted armours)?
Apologies if I've missed stats on previous combats where you've tested this; this cruddy laptop & net connection means I can't check tables/graphs as readily as I would normally :(