LRMs, Clustering, and Head Hits

  • We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Oh man, now I really want to tinker with LRM clustering chances just to see what happens. Maybe PPCS and AC/2s will become a bit more useful. But then again, it'll probably just mean even more AC/5s...
 

One problem you've got there is that you're talking about % damage. Once you get your first hit, your center is established. Then we consider things based on the cluster table. I will explain:

So these are cut and pasted from the actual combat constants file:

"ClusterChanceOriginalLocationMultiplier": 8,
"ClusterChanceAdjacentMultiplier": 4,
"ClusterChanceNonadjacentMultiplier": 1,
"ClusterChanceNeverMultiplyHead": true,
"ClusterChanceNeverClusterHead" : true,

From what I've read, OriginalLocationMultiplier means on the centered part established by the first LRM that hits. If a head hit, probably the second LRM to hit establishes the cluster center based on that rule at the bottom there. So the weighting for that center location is multiplied by 8. All adjacent location weights are multiplied by 4. Non-adjacent are left alone. Head weights never get multiplied so the second a center is established, the odds of a head hit almost vanish.

I suspect the 2nd to last rule guarantees no multipliers on the head if it's adjacent, which only happens on the core. So heads, whether they're adjacent or not, always have the same chance to get hit, which drastically reduces odds of taking hits with x4 and x8 happening to all these other weights.

The critical detail is that the first hit that isn't the head, which aren't likely to happen multiple times in a row at base %, establishes the center of the cluster. You could miss on the CS table and hit something different, establishing a cluster center elsewhere. After that, the cluster rules take over and the CS target is irrelevant. So there's also a step in the middle that you can totally throw things off.

The problem with your examples is your thinking in terms of % damage. But how many hits missed altogether at lower accuracy? Your first hit and subsequent hits will be a higher percentage of total damage in those scenarios. And if you only land 3 hits, who cares if you got like 80% damage of total on the core. That doesn't prove anything.

So yes, a called shot with LRMs isn't useless. It probably helps land more hits on the part if your first hit lands there on the CS table, and depending on what part it is, what facing, and what adjacent part values are, but it's still not a great use of morale, IMO. Because, altogether misses included, most of an LRM volley will not hit the thing you're aiming at.

Here's front hit location JSON and yes, it matters where you attack from. This is how it should go if you hit from the front with a normal attack or the first LRM missile to not hit the head:

"HitMechLocationFromFront":
[
{
"k": "Head",
"v": 1
},
{
"k": "CenterTorso",
"v": 16
},
{
"k": "LeftTorso",
"v": 14
},
{
"k": "RightTorso",
"v": 14
},
{
"k": "LeftArm",
"v": 10
},
{
"k": "RightArm",
"v": 10
},
{
"k": "LeftLeg",
"v": 8
},
{
"k": "RightLeg",
"v": 8
}
],

Now, I'm going to shorthand that a bit to demonstrate what it looks like when the first LRM shot lands on the torso:

Head:1
CenterTorso: 128
leftTorso: 56
RightTorso: 56
LeftArm:10
RightArm:10
LeftLeg: 32
RightLeg: 32

%-wise, rounded, that converts to:

Head:0.3%
CenterTorso: 39%
leftTorso: 17%
RightTorso: 17%
LeftArm:3%
RightArm:3%
LeftLeg: 10%
RightLeg: 10%

Now, these odds happen after your first hit that wasn't the head. If you get a head hit initially, your first two hits to establish center shouldn't figure into your damage expectations because the cluster table wasn't in effect yet.

Also remember, called shot is just a regular attack using a skewed table for location IF you hit. I would wager every LRM uses the CS table until there's a hit. Then we use these tables for the centered shots because if we didn't and multiplied the CS table instead, it would be like 8 times your called shot weight. You'd be hitting with almost every shot because the adjacent weights, even at x4 wouldn't have the added multiple of the called shot weight, which must be very big at 90% with mastery on top of that.
 
Alright, I am home and fed. Here are my results.

Testing Methodology

I used Kintaros shooting at Atlas target dummies in skirmish. The Atlai had no weapons and max armor. The Kintaros were loaded with knockdown SRMs (0 damage, 50 stability damage) and an LRM 100 with no change to damage. I made the absolute minimum changes I need to test this. I would knock down an Atlas with 2 of the Kintaros, record its starting armor values, fire with just 1 LRM 100, recording hit odds and called shot odds as reported by the game, then counted how many hits occurred at each location.

Background Info

Here are the cluster hit tables for the center torso, right side, and right leg. These are just the standard prone hit table, as located in the CombatConstants JSON, multiplied by the cluster hits modifiers (x8 for center of cluster, x4 for adjacent, and x1 for all others, with 0 for head).

prone cluster tables.PNG

Results

Here are the results from 10 shots. I mostly called the CT and 3 of my 4 mechwarriors had Tactics of 6+, so they had the first called shot skill bonus. I did call shots on a couple other locations to get some more info and to avoid coring the Atlai too quickly. Each Atlas carries 320 CT armor + 160 CT structure. After 2 volleys at full damage they were going internal and the third CT cluster finished each one off.

NOTE: Each shot of 100 LRMs was dealing about 200 damage to the CT. Two shots was putting an Atlas at about 100 structure left on the CT. If that seems like a lot, remember that I was firing 100 LRMs in each volley and with full damage each (no cover/brace). The standard complement of an assault LRM boat is 60 LRMs (4x LRM 15 Stalker, 3x LRM 20 Highlander). That is taking the full firepower of 3 assault mechs just to go internal on a max armor Atlas.

There was one instance where I called a shot on the arm and the cluster missed, hitting center torso instead. That Atlas had already taken 2 clusters to the CT and was destroyed. Unfortunately, I did not record all of the damage, but I do know that it would have take at least 60 missiles to destroy the CT and I don't think there were many, if any hits to the arm that I called the shot on. I did get the info for the last mech I destroyed and am only missing the exact number of missiles that hit the CT (but I do have the minimum).

called shot clustering.PNG

They are all fairly reasonably close to their appropriate cluster hits table. The most interesting result was when I did a called shot on the head. Naturally, the cluster didn't land on the head (it was only a 2% chance) so there were no hits to the head. If called shots from LRMs bypassed the clustering mechanics and used the called shot tables, I would expect the hits to be fairly evenly distributed across the mech.

Instead it landed on the right torso. The hits are very clearly weighted in favor of the right side. They are perhaps a little heavier on the CT than I would expect, but some variance is normal. (heh, normal... unintended statistics pun...)

I agree that the cluster hits tables do resemble the called shot tables. They both weight a specific location more than the others. The use of the clustering tables is probably clearest in looking at the instances when the cluster does not land on the called location. It would also be clearer with Called Shot Mastery as that would give 90% odds for the cluster to hit the CT... but only 55-65% of the missiles would hit the CT.

Conclusion
LRMs use the clustering mechanics, even when fired with a called shot. They deal a lot of damage per ton, but it is unfocused and generally bad at destroying fully armored locations.



Side note: I am considering weakening LRMs in my mods. I would give them a 2 point accuracy penalty and am considering lowering the clustering modifiers from 8/4/1 to 6/3/1. I don't think they are that overpowered, but I am considering a tweak.

EDIT: I have added this test to the OP to keep all the info together.
 
Last edited:
Another tweak to consider is limiting how much extra damage and/or instability the various + versions do, and giving them other bonuses to offset the loss of those factors, such as lower heat, better accuracy, lower min range, or something.
 
From what I've read, OriginalLocationMultiplier means on the centered part established by the first LRM that hits. If a head hit, probably the second LRM to hit establishes the cluster center based on that rule at the bottom there.
It looks like you have a pretty solid understanding of how the clustering mechanics work, but I have a couple corrections with regards to head hits.

First, if the first missile does hit the head, then the cluster is centered on the head. The torso locations get the x4 adjacent location modifiers and limbs are x1. What the "head never multiplied" line actually does is make it so that the head gets a x1 multiplied instead of the normal x8 multiplier for the center of the cluster. This actually makes it less likely that subsequent missiles in the volley will actually hit the head.

Second, if the cluster is not centered on the head, the chance of hitting the head is zero. There is absolutely no chance of getting a head hit if the cluster is centered on a torso, arm, or leg. I've tested this a lot. My results are in the OP.
 
Another tweak to consider is limiting how much extra damage and/or instability the various + versions do, and giving them other bonuses to offset the loss of those factors, such as lower heat, better accuracy, lower min range, or something.
Agreed. I would top the +damage and +stability at 1 point each. A bonus to accuracy would be preferred. For range I might give an increased optimal range without changing the minimum range. Maybe even increase the max range too (it is not as if we fire at max range very often anyways ;)). Reduced heat is a good idea for + variants for just about all weapons.

I would also like to have variable cluster hits multipliers. Currently there is only one set of multipliers in the CombatConstants file. If we had more, then Artemis LRMs could have tighter clustering than regular LRMs and we could have +clustering bonuses.
 
It looks like you have a pretty solid understanding of how the clustering mechanics work, but I have a couple corrections with regards to head hits.

First, if the first missile does hit the head, then the cluster is centered on the head. The torso locations get the x4 adjacent location modifiers and limbs are x1. What the "head never multiplied" line actually does is make it so that the head gets a x1 multiplied instead of the normal x8 multiplier for the center of the cluster. This actually makes it less likely that subsequent missiles in the volley will actually hit the head.

Second, if the cluster is not centered on the head, the chance of hitting the head is zero. There is absolutely no chance of getting a head hit if the cluster is centered on a torso, arm, or leg. I've tested this a lot. My results are in the OP.

Ah, thx. But weird. Why not just make it impossible on all but the first hit. I can't think of any ways to exploit the rule if cluster damage from core to head were possible. Decent chance if there was just a leg left I guess but who cares at that point.

I'm wondering if you happen to know: when a part is gone, does its weight go away or does damage just get applied automatically to the first inward part that's intact as excess damage does when you destroy a part?
 
I thank you to have taken the time to make all their test and long explanation. I made other test and they show some called shoot at 80% where only 50% of missiles go at the target. I can now only recognize that LRM called shoot use the cluster mechanic.
 
Ah, thx. But weird. Why not just make it impossible on all but the first hit. I can't think of any ways to exploit the rule if cluster damage from core to head were possible. Decent chance if there was just a leg left I guess but who cares at that point.

I'm wondering if you happen to know: when a part is gone, does its weight go away or does damage just get applied automatically to the first inward part that's intact as excess damage does when you destroy a part?
I'm pretty sure the hit chance for destroyed sections remains and it just gets automatically transferred in. That is how it worked in TT and I see no reason why they would have changed it. I am pretty sure I have seen red damage numbers (showing internal damage) when the enemy had destroyed locations, but all existing locations still had armor. I think that happens when a shot hits a location which is already gone, which would be internal, but then transfers the damage.

As for the quirks of clustering around head hits... this could have been just what they could get to work properly. Code isn't always easy. At one point during the beta there were complaints about how rare critical hits were. It was found that crits weren't rerolling empty locations, so you could have a crit which doesn't actually hit anything. The devs made some adjustments and said that crits should now reroll locations. The players tested it and found that it still wasn't rerolling locations, but other adjustments the devs had made increased the number of crits to a spot that players were happy with and there was a call to not change it. I don't know if they changed that in the final game or if it just got left as an unintended feature.

The point of the story is that code is messy and it may not be easy to get it to do what the programmers want.

I thank you to have taken the time to make all their test and long explanation. I made other test and they show some called shoot at 80% where only 50% of missiles go at the target. I can now only recognize that LRM called shoot use the cluster mechanic.
When in doubt, test it. I am happy to help people understand how the game works.