BATTLETECH - Hit Tables for the AI conducting Rear Torso Shots

[Prussian Havoc]

One line summary of your issue
Hit Tables for the AI conducting Rear Torso Shots

What platform are you running the game on?

Steam

What operating system do you use?
Windows
Which version of the game are you running?
Version 1.8
What DLC do you have installed?
Flashpoint,
Urban Warfare,
Shadowhawk Pack,
Heavy Metal

Please explain your issue is in as much detail as possible.
Prior to Update 1.8, the AI was at least fairly competent when it came to hitting my Rear Torsos when I turned my back toward them.

Since Update 1.8 I’ve been purposely turning my Rear Torsos toward the AI and there appears to me to be a much higher occurrence of Arm and Leg Hits rather than Rear Torso hits.

Has the Hit Tables for the AI been changed?

Unless I am mistaken, this is not a bug just an AI tweak. As such I’ll refrain from attaching a DxDiag or Output Log.

I did post to the forums to see if maybe my experiences are a probability outlier, but the trend continues unabated.
Have you tried verifying your files?
Yes

Steps to reproduce the issue
I constantly turn my Rear Torso towards the AI and it has amazed me how poorly the AI manages to hit my Rear Torsos.

Attach DxDiag (PC), System Info log (Mac), or run sudo lshw -html > specs.html and uploaded specs.html (Linux)

Attach Output Log
Windows: \Users\YourUserName\AppData\LocalLow\Harebrained Schemes\BATTLETECH\output_log.txt
Mac OS: Users >{name} > Library > Logs > Unity > Player.log
Linux: ~/.config/unity3d/Harebrained Schemes/BATTLETECH/Player.log

If the game crashed, please attach the crash file with the time stamp closest to the crash.
Windows: [Steam library folder]\steamapps\common\BATTLETECH\DumpBox
For v.17 and beyond try here: C:\Users\[YourUserName]\AppData\LocalLow\Harebrained Schemes\BATTLETECH
OSX: Macintosh HD > Users > [username] > Library > Application Support > Steam > steamapps > common > BattleTech > DumpBox
Linux: [Steam library folder]/steamapps/common/BATTLETECH/DumpBox

If a save game was loaded at the start of the game session, upload it. Please make sure to upload the save game with the timestamp closest to when you loaded it.
Windows:
Steam: [Steam installation folder]\userdata[SteamID]\637090\remote\C0\SGS1
GOG: 'C:\Users\USER_ID\AppData\Local\GOG.com\Galaxy\Applications\50593543263669699\Storage\Shared\Files\C#\SGS#'
Vanilla:'C:\Users\USER_ID\AppData\LocalLow\Harebrained Schemes\BattleTech\C#\SGS#'
OSX:
Steam: Macintosh HD > Users > [username] > Library > Application Support > Steam > userdata > [SteamID] > 637090 > remote > C#> SGS#
GOG: Macintosh HD > Users > [username] > Library > Application Support > GOG.com > Galaxy > Applications > 50593543263669699 > Storage > Shared > Files > C# > SGS#
Vanilla: Macintosh HD > Users > [username]> Library > Application Support > unity.Harebrained Schemes.BattleTech > C# > SGG#
Linux:
Steam: /home/.steam/steam/userdata/[steamID]/637090/remote/C#/SGS#
GOG: /home/[username]/.config/unity3d/Harebrained Schemes/BATTLETECH/remote/C#/SGS#/
Vanilla: /home/[username]/.config/unity3d/Harebrained Schemes/BATTLETECH/remote/C#/SGS#/

Attach screenshot / video

 

[Prussian Havoc]

I just responded to a forum thread where this tendency was independently noted. Any chance our Good @HBS_RedMenace or @HBS_CaptKerberos can shed some light on this one?

 

[Jade_Rook]

Can you provide any statistics for the hits? X limb hits out of Y shots?

 

[Prussian Havoc]

I am on the road today through Sunday and find the time I can game to be all too infrequent.

I am also hesitant to roll-back my BATTLETECH version in order to capture Skirmish Mode Single-Player metrics as a basis of comparison to what I can find in the current version.


But this is a good cause and one I believe in, so I’ll give it a shot beginning late tonight. :bow:




My best estimation of the delta is what sparked this query and thread, so I’ll detail it here and refer back to it as needed.

I play a lot of BATTLETECH... just North of 6,000 hours of combined BATTLETECH and BATTLETECH Beta. The great preponderance of this was Multiplayer BATTLETECH, a good bit of which is spent in Lobby.

This focus on BATTLETECH Skirmish Mode gives me a truly excellent feel for all of the BATTLETECH Skirmish Mode MechWarriors.

BATTLETECH Skirmish Mode Fury is a fun and interesting game mechanic but it does not accrue as fast as one accrues Morale in Campaign or Career Mode for example.

Firing into a Rear Arc gives one a certain spread of damage locations. It is my contention that the current build of BATTLETECH has a markably greater spread of damage locations than we found at Launch / the early builds.

Upward of a third more shots are hitting Locations other than the three Rear Torsos in the current build as opposed to the Launch build - 34% is my best estimation.

It is this I will attempt to prove/disprove.

My sample size will be statistically insignificant, I just don’t have the time for the thousands of iterations it would require to be statistically significant.

But I will give it a good faith effort and allow AndrewT and our good friends at Paradox and HBS to take it from there. Of course it just might be that @HBS_RedMenace might have some incite to Hit Location Distribution under HEAVY METAL / Update 1.8...


More to follow as the weekend continues and likely into next week... : )

 

[Jade_Rook]

Limb hits from the back should be at 29%. Here are the weights:

Arms: 4 each
Legs: 5 each
Side Torso: 14 each
Center Torso: 16
Total: 62
Total Limbs: 18
Chance of a limb hit: 18/62 = 29%

I don't know if they changed the hit location tables in 1.8, but I don't think they did.

If the long run probability is 34%, then something is off, but it would take a decent chunk of data to confirm that.

 

[Prussian Havoc]

Unforeseen and some foreseen events have slowed me down in my ongoing data collection efforts.

I’ve looked through the very limited iterations I’ve already accomplished and there is just simply not data to be useful yet.

And with LSU playing Clemson in the NCAAFB Championship at the moment, odds are I’ll get no data colection tonight.


More to follow in the days/week ahead.

 

[Prussian Havoc]

INITIAL FINDINGS
First batch of data was flawed as overtime my ability to tract back this or that damaged limb to this or that of my Weapons, degraded significantly. Counting Strikes became even more difficult and slightly muddled as one or more of my Flamers cause Enemy to overheat, or my large damage weapons destroyed locations and had damage carry over onto other sections. Enemy return fire proved problematic too, as I lost weapons and even Mechs because I was prioritizing Backstabs over my holding cover or building Evasion Chevrons, etc. Importantly, Enemy Mechs were destroyed all too soon given the volume of fire directed against ill-armored Rear Torsos. Plus the number of weapons (thus the number of possible hits) proved so low that data collection was occurring too slowly for my tastes and time I have to contribute here.


REVISED COURSE OF ACTION
Rather than use Stock Mechs, I need purpose-built AI Targets and purpose-built Attackers. I’ll use max-Armored Stalkers (with a single Medium Laser) as my AI Targets. And I’ll use max-Armored Grasshoppers armed solely with Machine Guns as my Attackers. This should both speed up data collection as the number of 2-points of damage hit locations will scale up nicely, improving data validity as the number of hits can quickly be deduced by simply dividing total damage per section by 2. I will use Lunar Biome Maps so as to limit Cover and neither Mech should be in danger of overheating.



All in all this is proving more fun that I thought, but taking more time than I ever thought it would. Given finite gaming time, this effort is not my main priority but it is important to me and will continue along the lines I describe above.

All questions, comments and constructive criticisms are welcome.

And as ever, good luck and good BATTLETECH gaming! : )

 

[Jade_Rook]

Yeah, this kind of testing takes some very deliberate approaches to get strong evidence. Thanks for working on this!

 

[Prussian Havoc]

Yeah, this kind of testing takes some very deliberate approaches to get strong evidence. Thanks for working on this!

Just a day or so ago I began Modded BATTLETECH gameplay.

So as to not taint my data collection with the presence and influences of Mods, I’ll go ahead and provide what Vanilla BATTLETECH Findings I’ve managed to collect thus far.

First off, my finding served to highlight the presence of “streaky Hit Table Abnormalities.” It was just such a streak of limb hits that cued me into posting the OP Bug Report in the first place,

If I remove these Streaks, the remaining Data hangs within a standard margin of error (3-4% points one way or the other) of @Jade_Rook’s 29% he offered above,

The problem is self-evident when the Data from the Streaky Hit Table Abnormalities are factored into the mix. With this data, the Limb Hit rate rises by 5-7% for final tally of 34-36% Limb Hits.

An example or two of such streaks is very useful at this point.

Streak Example 1.) Pristine AI Mech facing away (presenting its Rear Torso toward) a pair of my Hunchback-4P’s. I fire 6-Medium Lasers from each PartyBack, Results were 2-hits to the CT Rear and 1 to the LT Rear (no penetration of the Armor.) 2 Medium Lasers miss the target completely. 7 Medium Laser shots hit their Target’s Limbs despite being fired directly into the Rear Arc. The Armor is stripped from one Leg and one Arm (Left Leg, Right Arm) and as I recall a crit to an arm-mounted weapon is achieved.

Streak Example 2.) I jump behind an AI Hunchback with my trusty Firestarter, looking to take out either its AC20 Rear Torso and damage destroy its Autocannon, or better yet, pierce the relatively thin Rear Armor over its Left Side Torso Rear and ignite and explosion of its AC20 ammo. The AI Hunchback had not moved much, maybe IIRC a single Evasion Chevron. I have 60% chance to hit with the Medium Lasers and upwards of 80% of better on the MGs and Flamers. I fire. Both Lasers hit, one an Arm, one a Leg. The CT Rear loses 6-points of Armor and another 6-points from the Right Side Torso. The remaining Hits... maybe 10, are spread across the Hunchbacks Limbs with the Leg struck by one of the Medium Lasers taking the brunt of this damage. Knowing this was the manner of situation I was looking for, I sacrifice my Firestarter and have my Gladiator (Coolant Venting Bulwarker) engage in Mitch-style Light Mech Melee of a much larger and vastly more dangerous opponent while that opponent still has its back turned. The Melee strike does hit the CT Rear, but again at the ratio of maybe 3 to 1, the MG and Flamers hit the Limbs.


Yes, these were my two best examples of Streaks, but their data and the data of lesser streaks are just as valid and together serve to move the needle toward the mid-30’s percentage-wise.

Plus in hindsight, such streaks proved mightily frustrating and were clearly responsible for my motivations to start this bug report in the first place.



@Jade_Rook, that is the best I can do. :bow:

I am having a blast with my current Mods, but since they include @AmechWarrior’s Better AI and a large number of others, I find that any data collected now would be of extremely low value in an inquiry of Vanilla BATTLETECH Rear Hit Tables.

I don’t quite know what @AndrewT can make of all this and while I believe he is one of the hardest working folks at Paradox (witness all his many 10’s of thousands of Bug Tread posts! (Thank you btw Good AndrewT :bow: ) I don’t think I have provide anything he can make use of.

Rather it is @HBS_RedMenace or @HBS_PnkElephnt who might be able to shed some light on if an intentional change, unintentional change or Bug might have caused such Streaks of Limb Hits from the Rear Arc?

Again I first noticed this behavior in the couple weeks prior to this thread’s OP of 05DEC2019, which places the start alongside HEAVY METAL’s and Update 1.8’s Release into the Wild.

Thoughts?

 

[Jade_Rook]

Thank you for the information Prussian! Personally, I would say that there is most likely not an issue with the location determination. Streaks happen with random events. They are one of the things which catch our attention and stand out, making it harder to understand the underlying probabilities, but they are normal, expected even.

At some point I might test this myself, but life has me rather busy right now.

 

[Prussian Havoc]

Thank you for the information Prussian! Personally, I would say that there is most likely not an issue with the location determination. Streaks happen with random events. They are one of the things which catch our attention and stand out, making it harder to understand the underlying probabilities, but they are normal, expected even.

At some point I might test this myself, but life has me rather busy right now.

Yeah, i would recommend you not waste your time testing it. It really took away from the fun of gaming having to stop and parse out all individual weapons hits based on resultant damage and back-engineering what weapons hit what location.

But I do disagree to a degree... Streaks do happen, but so then do long periods of no hits to Limbs. These Streaks and Dearths should have evened out. They did not. There was still a statically significant increase to Rear Torso Hits between Build 1.7 to Build 1.8


And thank you for suggesting this, I have not taken my Quantitative Business Analysis and dusted it off in way too many years. And since I was accepted to an MBA Program this week, I am likely to need all the practice I can get before this Fall! : )