BATTLETECH

BATTLETECH

View Stats:
BATTLETECH > General Discussions > Topic Details
your.sheepy Jun 13, 2018 @ 1:14am
Why Our Heads Keep Being Hit - a tale of 58000+ SRMs
Update - Headsoht bug fixed in 1.1.1, partly thanks to this thread!
Before that, game version 1.1 has removed clustering effect of called shot.

The mod I made before to fix the head hit bug is still effective on fixing vehicle called shot, clustering effect, and more:
https://www.nexusmods.com/battletech/mods/242/


---------------

You may have heard them many times. I've said them here myself. Don't call shot the head. Hit rate is adjusted. But I haven't had the data to back it up.
(-‸ლ)

Well, now I do. I modded SRM to do 1 damage x 500 shots, modded pilots and precisions to change %, and fired them in Skirmish against AI.
( -_・)︻デ═一 - - - - - - - - ≧〔゜,,,,゜〕≦

When you simply spray from the front, head hit chance is ~2.7%. (12200 samples)
When you call the head at 2%, actual rate is ~1.3%. (11800 samples)
When you call the head at 5%, actual rate is ~2.3%. (7300 samples)
When you call the head at 8%, actual rate is ~3.8%. (7000 samples)
When you call the head at 18%, actual rate is ~8.6%. (7700 samples)
When you call the center torso at 82%, actual rate is ~94.0%. (4600 samples)

So, what does that mean and why you may want to flip the table? (╯°□°)╯︵ ┻━┻

.Called headshot chance is around half that of displayed, at least in skirmish.
.Plain frontal shot actually deals more headshots than normal and bonus called shot.
.Instead, both called shots direct shots towards the torsos, away from head and limbs.
.The to-hit "correction" and miss breaker code on reddit match the total hit% very closely.
.Roll adjustments does not match actual called shot chance; three of the four real chances are lower than adjusted %, and the gap widens the higher we go.

Declaration: Many missiles, crabs, and trees were hurt in these experiments, in horrible ways that I dare not describe. May their souls rest in peace.
≧〔X,,,,X〕≦ ≧〔X,,,,X〕≦ ≧〔X,,,,X〕≦

==========

For frontal shot (non-called), 7500 SRMs were fired at 95% hit chance.

Total hits = 7297 (97.3%, expected 97.2% after adjustment and miss breaker)
Head hits = 199 (2.7%, game says 1%)
CT hits = 1380 (18.9%, game says 20%)
LT/RT hits = 1251x2 (17.2%, game says 17%)
LA/RA hits = 932x2 (12.8%, game says 12%)
LL/RR hits = 676x2 (9.3%, game says 10%)


For frontal shot (non-called) made by NPC, 9000 SRMs were fired at various hit chance.

Total hits = 4925 (54.7%, from 18.8% to 86.2%)
Head hits = 129 (2.6%, should be 1%)
CT hits = 961 (19.5%, should be 20%)
LT/RT hits = 872x2 (17.7%, should be 17%)
LA/RA hits = 591x2 (12.0%, should be 12%)
LL/RR hits = 454x2 (9.2%, should be 10%)


For 2% headshot, 12500 SRMs were fired at 90% hit chance.

Total hits = 11819 (94.6%, expected 94.5% after adjustment and miss breaker)
Head hits = 157 (1.3%, game says 2%, or 1.3% after adjustment)
CT hits = 3538 (30%, game says 20%)
LT/RT hits = 3105x2 (26.3%, game says 17%)
LA/RA hits = 536x2 (4.5%, game says 12%)
LL/RR hits = 421x2 (3.6%, game says 10%)


For 5% headshot, 8000 SRMs were fired at 80% hit chance.

Total hits = 7056 (88.2%, expected 88.3% after adjustment and miss breaker)
Head hits = 163 (2.3%, game says 5%, or 2.9% after adjustment)
CT hits = 2102 (29.8%, game says 19%)
LT/RT hits = 1846x2 (26.2%, game says 17%)
LA/RA hits = 313x2 (4.4%, game says 12%)
LL/RR hits = 237x2 (3.3%, game says 10%)


For 8% headshot, 8000 SRMs were fired at 85% hit chance.

8% headshot can be achieved by modding pilots to have called shot master but without called shot bonus. It is the best middle point I can find between 18% and 5%.

Total hits = 7363 (92.0%, expected 91.6% after adjustment and miss breaker)
Head hits = 278 (3.8%, game says 8%, or 4.5% after adjustment)
CT hits = 2095 (28.5%, game says 18%)
LT/RT hits = 1873x2 (25.4%, game says 16%)
LA/RA hits = 356x2 (4.8%, game says 11%)
LL/RR hits = 266x2 (3.6%, game says 9%)


For 18% headshot, 8000 SRMs were fired at 95% hit chance.

Total hits = 7768 (97.1%, expected 97.2% after adjustment and miss breaker)
Head hits = 671 (8.6%, game says 18%, or 10.8% after adjustment)
CT hits = 2147 (27.6%, game says 16%)
LT/RT hits = 1866x2 (24.0%, game says 14%)
LA/RA hits = 345x2 (4.4%, game says 10%)
LL/RR hits = 264x2 (3.4%, game says 8%)


For 18% CT shot, 5500 SRMs were fired at 75% hit chance.

The sample is fewer than others, but given the very high CT rate, the CT % is actually more accurate.

Total hits = 4639 (84.3%, expected 84.6% after adjustment and miss breaker)
Head hits = 0 (0%, game says 0%)
CT hits = 4362 (94.0%, game says 82%, or 89.7% after adjustment)
LT/RT hits = 108x2 (2.3%, game says 4%)
LA/RA hits = 16x2 (0.3%, game says 3%)
LL/RR hits = 14x2 (0.4%, game says 2%)

Link of the detailed log of each firing is here:
https://gamefaqs.gamespot.com/pc/205058-battletech/faqs/75955/probability-tests

==========

P.S. I've tested more than SRMs... won't compile those here but this is what I found.

.Not in the data, but pretty sure each weapon is limited to one crit, skipping locations destroyed by that weapon. (i.e. only survived location can crit)
.Odd piloting does not seem to do what they claim. (4500 samples, need more testing)
.Called shot weight applies to first LRM that hit, then LRM clustering weights overrides called shot. (10000 samples)
.Elevation does not have obvious effect on CT called shot distribution.

Also, I have done enough SRM testing. No further tests expected.
Last edited by your.sheepy; Jun 27, 2018 @ 7:20pm
< >
Showing 1-15 of 71 comments
Madkine Jun 13, 2018 @ 1:18am 
Very impressive, looking forward to seeing more data.
Mistfox Jun 13, 2018 @ 1:20am 
Holy crap, you're way more hardworking than I ever will be. lol. Thanks.
Senkoau Jun 13, 2018 @ 2:12am 
Since no one else has said it yet I sadly feel I have to.

You don't understand RNG, its working as intended, the enemy has the same hit % you do.

I think that covers all the bases.
Madkine Jun 13, 2018 @ 2:32am 
Originally posted by Senkoau:
Since no one else has said it yet I sadly feel I have to.

You don't understand RNG, its working as intended, the enemy has the same hit % you do.

I think that covers all the bases.

And clearly you understand nothing about statistics, so your statement covers none of the bases.
Stonedog Jun 13, 2018 @ 2:40am 
Wow, some pretty shocking numbers in there. If this data holds up to your further testing , hell if it holds up at 75% accuracy for that matter, it explains a lot of things. Like all the people posting about downed mechs and called shots never doing anything near what they should be . Someones math was a bit off somewhere , been apparent to all but the most blinded of us playing this game, but damn.

You take in the fact that mid game and all on you are almost always look at the AI having a 3-1 odds advantage and that magnafies the problem to a staggering extent.

Nice work man !
Zyrrashijn Jun 13, 2018 @ 2:47am 
Thing is, on a called shot with missiles, the %-chance only applies to the first (single!) missile. So the rest of you 500 missile volley is no more likely to hit the head than an uncalled shot.
Stonedog Jun 13, 2018 @ 3:17am 
Originally posted by Zyrrashijn:
Thing is, on a called shot with missiles, the %-chance only applies to the first (single!) missile. So the rest of you 500 missile volley is no more likely to hit the head than an uncalled shot.


Right, and since the chance appears to be even higher for uncalled shots , well that is frightening.Even as he staed, 50% going into ct on called head shots. To me that says a lot of things. Like next time I want to hit a CT on a down meched I am calling head shot.

Yet real implication is when you take out missle ( Missles just let you get a larger sample size much faster, I am asuuming that was our OP intent ) and substitute weapon X , outside of the modifiers applied to the code for various weapon types the actual formula used for non melee combat should be the same for every weapon. so if uncalled shots are hitting the head across the board at 2.7% vs the 1% its supposed to be, thats a seriously high rate %wise.

Putting Missles back in this Idea, no damn wonder everyone loads them up. your giving me a weapon that has a chance to hit at a much higher than it should and letting me hurll 2 or 10 or 30 at once.

Doesnt matter though how you want to put it, that needs fixing as soon as humanly possible if the data holds up, and I see no reason to doubt the OP or his data as it rings true with all Ive seen in game and on this, and other Forums.

I mean did no one inhouse talk about this in beta ? Did no one bring it up or notice ? OR as I suspect were they silenced by the mass of Fanbois and worshipers that all major brand games tend to have ?
Senkoau Jun 13, 2018 @ 3:23am 
Originally posted by Madkine:
Originally posted by Senkoau:
Since no one else has said it yet I sadly feel I have to.

You don't understand RNG, its working as intended, the enemy has the same hit % you do.

I think that covers all the bases.

And clearly you understand nothing about statistics, so your statement covers none of the bases.

I was posting the standard response to this kind of thread as a joke, apparently it fell flat.
your.sheepy Jun 13, 2018 @ 3:27am 
Originally posted by Stonedog:
If this data holds up to your further testing, hell if it holds up at 75% accuracy for that matter, it explains a lot of things.
If you mean accuracy adjustments, yes, pretty sure they apply. I tested 65% and 40% with LRM, and the total hit matches like SRM.

75% = 84.0% single hit chance, or 84.6% on LRM20 because of miss breaker.
(miss breaker has a diminishing effect; LRM500 has practically same chance as 20.)

I haven't tested on prone mechs and have no plan to test it, because I think we have enough data points here. Testing does take time.


Originally posted by Zyrrashijn:
Thing is, on a called shot with missiles, the %-chance only applies to the first (single!) missile. So the rest of you 500 missile volley is no more likely to hit the head than an uncalled shot.
Well, LRM is totally different, that much is true. You only get one chance to hit the head per salvo. The rest never hit if the first didn't. Even when the first did, the rest have much lower chance to hit because of clustering. In my data that happened twice after 49 firings, or 8 missile hits after 15400 LRMs if you want to know the total.

But please don't quote "Sheepy says 8/15400", because that's not how LRM stats works. Fortunately SRM does not have this clustering behaviour; I've hurt enough crab heads to know.


Originally posted by Madkine:
And clearly you (Senkoau) understand nothing about statistics, so your statement covers none of the bases.
Senkoau does have a point; I didn't state it in the body because I haven't tested that. But what my title mean is IF the enemy use the same hit rates, their normal attacks have a higher head hit chance than our called shots. Which fits my experience.

I'll add that to my test queue. I only need to resist my instinct to keep moving, and try to stand still out of cover. Looks like I'll finally get my lost to AI achievement.


Originally posted by Stonedog:
I mean did no one inhouse talk about this in beta ? Did no one bring it up or notice ? OR as I suspect were they silenced by the mass of Fanbois and worshipers that all major brand games tend to have ?
Hit chance has been talked and tested, mostly on the official forum. I have their efforts to thanks. But I think I may be the first to test at this scale post-beta?
Last edited by your.sheepy; Jun 13, 2018 @ 3:40am
TehJumpingJawa Jun 13, 2018 @ 3:41am 
Great, so the numbers upon which our strategy is based (in a strategy game!), are a lie.

Love to hear the justification for this.

Is there any way to test whether these hidden modifiers are applied to the AI too?
Last edited by TehJumpingJawa; Jun 13, 2018 @ 3:43am
Stonedog Jun 13, 2018 @ 3:47am 
Originally posted by Senkoau:
Originally posted by Madkine:

And clearly you understand nothing about statistics, so your statement covers none of the bases.

I was posting the standard response to this kind of thread as a joke, apparently it fell flat.
I thought you were being sarcastic and that was what you were doing, but I wasnt going to jump in the middle , and be wrong and face the both of you flaming me LOL !
your.sheepy Jun 13, 2018 @ 3:57am 
Originally posted by TehJumpingJawa:
Great, so the numbers upon which our strategy is based (in a strategy game!), are a lie.

Love to hear the justification for this.

Is there any way to test whether these hidden modifiers are applied to the AI too?
Planning to test that. I think it'll take me two nights. Got baby to milk and part time to do.

I can understand the desire to fudge overall to-hit. I've coded an IRC dicebot, and although I can verify its randomness (unlike many other bots), players still complain. Human is indeed very bad at evaluating chances. So I added a miss breaker and crit breaker and same roll breaker, and the complains are (mostly) gone. I believe I did the right thing.

But the head hit chances, well, yeah it is worse than I expected. I want some explanations too. Should be fun if someone copy or link my post to the paradox forum. (・ω・)
Last edited by your.sheepy; Jun 13, 2018 @ 4:04am
Senkoau Jun 13, 2018 @ 4:03am 
Originally posted by Stonedog:
Originally posted by Senkoau:

I was posting the standard response to this kind of thread as a joke, apparently it fell flat.
I thought you were being sarcastic and that was what you were doing, but I wasnt going to jump in the middle , and be wrong and face the both of you flaming me LOL !

Well at least someone got a laugh out of it.

On topic I really hate that last one your.sheepy where a called shot of 18% only hits around 8% of the time since it means we're effectively losing over half our chance to hit. Which would mean instead of 1 in 5 on average hitting for an injury its only 1 in 10 and not that much better than their normal spread chance. Not that the rest is much better when they could be hitting us 2 -1 in an even battle with those values much less when the odds are stacked in their favour like they normally are on normal random shots.
Stonedog Jun 13, 2018 @ 4:07am 
Originally posted by TheWorld:
I deleted it out of respect for your polite request, you can delete the above too.

Thank you my friend, it was a honest request and I am glad you took it that way. I was not tryoing to interfere but only want this post to stay up as long as possible.

I am grateful,

Stone
Zyrrashijn Jun 13, 2018 @ 4:08am 
I fully understand that the thread is about SRMs not LRMs, thanks for your always kind notification, TheWorld.
But to my knowledge, both missile types behave the same in regards of hit localization. Am I wrong there?
< >
Showing 1-15 of 71 comments
Per page: 15 30 50

BATTLETECH > General Discussions > Topic Details
Date Posted: Jun 13, 2018 @ 1:14am
Posts: 71