Fixing Vehicle Balance with Non-Linear Target Size Scaling

So a little while ago I made this post. Some people seemed to like it and said to make a thread, so here it is:

Fixing Vehicle Balance with Non-Linear Target Size Scaling – v2.0
NLTSS is easier to say, right?

If you don’t want to read this 5 page / 2,082 word post, skip to “Target Scaling” and “A Much More Intricate Formula – NLTSS v2.0.”

Preface
As pointed out by Sander93 in the other thread, this change would require an incredible effort by the balance team; it involves changing the target sizes of every vehicle, changing the close/mid/far accuracies of every AT-type weapon, as well as extensive testing to ensure nothing is broken.

With that said, I still think its worth at least discussing, if not for a possible implementation into CoH2 (maybe one day?), but possibly for “CoH3”, if that ever happens. Additionally, someone else may figure out an easier way to implement this, and can then propose their idea. Either way, it’s a fun thought experiment.

Overview
Right now, many would say CoH2 has a problem with vehicle balance. I think most players will agree that the current state of vehicles and AT is not conducive to ‘medium tank play’; either due to the current prevalence of TD and Heavy use, due to the power of ATGs and AT infantry, or due to other reasons. From discussions in multiple threads, it seems that the main issue is that AT units are far too generalist; most are simply excellent against all ‘classes’ of vehicles - heavies, mediums, light, etc.

The problem is, balancing a unit so that it excels against heavies and premium mediums, but not against regular mediums is exceedingly challenging. For an AT-unit to excel against heavy tanks it needs high penetration values; which means it will almost certainly penetrate mediums 100% of the time. The AT unit also needs to do significant damage, since heavies have high HP pools; this causes problems since mediums can often only survive 3 shots from these TDs, meaning they are forced to retreat after 2 hits – or one salvo, if there are multiple AT units (as is often the case). Additionally, due to salvos being entirely ‘alpha’ damage (reload isn’t considered), Rate of Fire adjustments won’t necessarily address this issue.

This leaves us with only one stat to play with: accuracy – and that is the topic of this post.

The Problem With Accuracy, Target Sizes, and Value Ranges
As it is currently implemented in the game, accuracy and target sizes are directly connected. The formula is (mostly) as follows:

Target Size x Accuracy = Chance to Hit
Due to scatter, moving accuracy, and a few other values, this isn’t the entire formula, but for this discussion, it is accurate enough.

At face value, this formula seems fairly straight forward; if you want something to be easy to hit, make it big – and if you want something to hit things more easily, make its accuracy high. However, this isn’t the full picture; we also need to discuss value ranges.

First, let’s look at target sizes. This number simply describes how “big” a unit is – it (seemingly) does so in whole numbers, and the range is (from my understanding) anywhere from 0 to infinity (there is likely a max value, but I’m unaware of it). Initially, this seems fine; there are an infinite number of whole numbers, so just space out units a lot; a Tiger could have a size of 100, a Kubel a size of 10, and so on. However, this is not the case; in CoH2, vehicles currently range from a size of 14 (Kubel, etc.) to 26 (JT, etc.). This means that are 12 values we can currently assign to vehicles in the game; this is a problem.

Let’s take, for example, the current match-up between an M36 (USF’s ‘heavy’ TD) and the newly adjusted Panzer 4 and Tiger. The M36 has an accuracy of 0.035 at ’60’ range, the Panzer 4 has a target size of 20, and the Tiger has a target size of 26. This gives the M36 a 70% chance to hit the Panzer 4, and a 91% chance to hit the Tiger. This seems fine, but what if we wanted to adjust the M36 to miss against the Panzer 4 more often; for example, a hit rate of 60%? This would mean giving the M36 an accuracy of 0.03; but now it would only hit the Tiger 78% of the time. We could, instead, attempt to lower the Panzer 4’s target size to 17, but this also causes problems. We’ve achieved our 60% chance to hit the Panzer 4 while keeping the chance to hit the Tiger at 91%, but now the Panzer is actually harder to hit than the ‘222’ Scout Car (target size 18).

We’ve run into the issue with value ranges. Having a range of 14-26 leaves us with 12 ‘steps’, and gives us a “middle” target size of 20 (conveniently, the size of most mediums); the biggest unit can only be 30% bigger than a Panzer 4, and the smallest only 30% smaller. This is incredibly restrictive when balancing units, as it means that accuracy and target size can’t really be used to create drastic differences in the chance to hit vehicles of different ‘classes’.

Graph
Show Spoiler
With the current values, the M36 firing at a Kubel has a 49% chance to hit, firing at a Medium a 70% chance to hit, and at a Heavy a 91% chance to it.

What if we could change that?

Target Size Scaling, and Non-Linear Advantages
This is where the idea of scaling target sizes comes in. If we want more values to work with, we can simply scale the target sizes; for example, if we scaled everything by 2, we now have 24 ‘steps’ to work with, and if we halve the accuracy values of AT units, unit interactions remain the same:

Current Game: 0.035 x 20 = 70% chance to hit
2x Size/Half Acc: 0.0175 x 40 = 70% chance to hit

There is, of course, a problem with that. While the number of values we can work with has doubled, the difference between units hasn’t. A medium tank would now have a size of 40, and a heavy tank would have a size of 52 – which is still only a difference of 30%. This is the core of this idea; scaling unit sizes non-linearly gives us a much larger usable value range to work with.

In my previous post, I suggested scaling units by the following formula:

Size Increase = 2x(old size/26)

What this formula does is scale everything by a non-fixed ratio; for a kubel, with a size of 14, it has its target size increase by only 8%, to a size of 15, for a medium (20) its size increases by 54% to 31, and a heavy (26) is increase by 100% to 52.

Kubel: 2x(14/26) = 2x(0.538) = 1.08
Medium: 2x(20/26) = 2x(0.77) = 1.54
Heavy: 2x(26/26) = 2x(1) = 2.00

Graph
Show Spoiler
If we use our ‘halved’ accuracy value for the M36 (0.0175), we now have new hit chances: A kubel (now size 15) has a 26% chance to be hit, a Medium (now size 31) has a 54% chance to be hit, and a Heavy (now size 52) has a 91% chance to be hit. If we refer back to our ‘live’ values, we have changed our chance to hit by -23%, -16% and 0%.

While we’ve successfully lowered the chance to hit a medium while keeping the chance to hit a heavy the same, we now have the problem (Sander93 pointed out), where very small vehicles are incredibly hard to hit.
This is indeed the biggest problem with this particular scaling formula. However, there are two things we can do to fix this: firstly, we could try scaling accuracy, and secondly, we could use a much more intricate formula.

Accuracy Scaling
With our new target size scaling, we can start to use accuracy values to tune unit performance more finely; this is something I suggested in my original post. Let’s assume we want our chance to hit mediums to be a bit higher than above; for example, 65% - while still using the new unit sizes. To do this, the M36’s accuracy would need to increase to 0.021, from 0.0175.

Graph
Show Spoiler
We now slightly improved hit chances: A kubel has a 31% chance to be hit, a Medium has a 65% chance to be hit, and a Heavy has a 100% chance to be hit. Compared to the current game, that’s -18%, -5% and +9%.

I believe this is a step in the right direction; Mediums are harder to hit and Heavies are easier to hit; but light vehicles are still a problem – but we can do better.

A Much More Intricate Formula – NLTSS v2.0
While I was happy with the general outcome of the formula above, Sander93’s comment is still accurate:

However, there are several implications (the smallest vehicles becoming very hard to hit for example) and it would require a tremendous amount of work and time to finetune everything.

Smaller vehicles become too hard to hit, and it requires an incredible amount of work to implement – arbitrarily changing 3 accuracy values per unit on top of changing the target sizes on every vehicle.

What we need is a formula to change target sizes in a way that smaller vehicles don’t become harder to hit, mediums do become harder to hit, heavies become easier to hit (or stay the same), and to do it in such a way that accuracy values don’t need to be manually fine-tuned (other than halving them). Essentially, we want to keep that 30% difference between small and medium vehicles, but we want to drastically increase the 30% difference between medium and heavy vehicles. This is my formula:

New Size = 2 x ((Old Size/26)^6) x Old Size) x 0.47 + 28

Now, you might be thinking: What is that?

Graph
Show Spoiler
With these new target sizes, the M36 at half its current accuracy (0.0175) now has the following hit chances: Firing at Kubel its 49%, firing at a medium its 56%, and firing at a heavy its 91%. Compared to the current ‘live’ values, we have changed our chance to hit by 0%/-14%/0%.

Additionally, the formula is scalable; the central exponent (the 6), can be increased or decreased to create a steeper or more shallow curve without changing the chance to hit the smallest and biggest vehicles. We can also stretch the height of the curve using the multiplier (0.47); this allows us to increase the chance to hit larger vehicles, with little change to mediums, and no changes to small vehicles.

As an example, let’s say we want small vehicles to stay the same as they are now, we want mediums to be 10% harder to hit, and we want heavies to be 5% easier to hit:

New Size = 2 x ((Old Size/26)^4) x Old Size) x 0.53 + 27

Graph
Show Spoiler
Now an M36 with 0.0175 accuracy has a 49% chance to hit a light vehicle, 60% to hit a medium, and 96% to hit a heavy – precisely what we wanted.

Of course, this also works with other unit matchups; let’s take a half-accuracy Pak40 (0.02 from 0.04), and apply the scaling formula again. The chance to hit a UC goes from 60% to 58%, the chance to hit a medium goes from 80% to 68%, and the chance to hit a heavy remains at 100%.

On paper, at least, this seems to work.

Conclusion

While it would be a fair bit of work to implement, I think that the above changes could solve the problem of Medium tanks being less viable in competitive play. By increasing the range of values we can use for target sizes, and halving the accuracy of all AT weapons, we can use a formula to systemically adjust target sizes to produce the desired hit chances.

Addendum
While writing this, it was pointed out that this would likely cause a lot of issues with SP units. If decimal values are permitted for target sizes (as infantry uses), then the proposed changes could be “halved”; this would allow the accuracies to remain exactly as they are now, and it shouldn’t impact SP too much, while still having the desired changes in MP. However, I don’t know if this is possible.

The decimal idea is also a potentially good route to take (if its possible for vehicles). Either way I think the goal is the same

Wish someone had thought of this earlier! Iirc Vipper used to mention something about the scale of sizes in the whole game being a bit wonky

The detail of your idea is remarkable dude, bravo

While improving game mechanics or, like in this case, the accuracy and target size system for vehicles is a good thing, I honestly never found it problematic. The game involves a lot of RNG management and choosing to fire on the move, or to stop and fire and manipulating with the distance of the engagement, they all depend on your and your enemy's decisions and the outcomes are easy to predict. With most vehicles being fairly similar in size and easy to hit if played right, it also feels intuitive and I am hesitant to changes in this field.

Meanwhile a significantly less flexible system that brings any teamgame meta to TD/Panther/Heavy spam is related mostly to damage breakpoints and penetration. For example, 1 ATG needs 3 reloads (4 shots) to kill an average medium tank, but 2 ATGs need 1 reload each (2 shots each) to kill the same tank. This is 3 times lower time-to-kill and is a bad design, as all it takes is just to build 1 unit extra, meanwhile your wipe potential increases 3 fold. Additional problem is related to penetration and some armoured units being horrbile versus other be it P4 vs Comet or Cromwell vs Panther. While I don't think mediums should win vs premium mediums, they could definitely be less useless against them. Current armour flanking system is a gimmick with map design dictating the engagements and occasional close range tank phasing. Lago suggested quite some time ago a point-blank system for tanks which would allow mediums to score penetrating hits more often if the engagement conditions are favourable (for example a close proximity fight between allied medium and Panther in Lienne Forest, where tank destroyers or ATG have no range advantage or sight). Suggested point-blank system is a band-aid and is by no means perfect (potential tank gameplay facehug dumb down), but it's already better than current system.

If the Medium/Premium-Medium/TD dependancy is improved and Mediums are less useless in Premium engagements, TD could see a change to be a support unit with lower rate of fire and damage breakpoint picked such to synergize with Mediums but by itself having high overkill vs Mediums and with always high penetration. This would leave them vulnerable even to P4 attacks if they were left alone. This would also make them preferred companion for Mediums as TD would require less micro than Medium itself and would be a solid force multiplier.

Not sure if any of you agree with me on this topic, or if any changes this big could even make it at this stage of game life, but we can always theorycraft and potentially provide some decent ideas for potential future system of coh3, ofc in another thread, as this one is related to vehicle accuracy and target size.

Lago suggested quite some time ago a point-blank system for tanks which would allow mediums to score penetrating hits more often if the engagement conditions are favourable (for example a close proximity fight between allied medium and Panther in Lienne Forest, where tank destroyers or ATG have no range advantage or sight). Suggested point-blank system is a band-aid and is by no means perfect (potential tank gameplay facehug dumb down), but it's already better than current system.

I liked that idea as well. I assume sander or another modder gave some feedback in that thread, but i can't remember what it was

I think both are too large a change at this late in the game, at least they already helped mediums a little in this last patch

Great OP, I love the point of it, having a better adjustment through acc and target size.
I sincerely would prefer to adjust the curve in such a way that acc values reflect their intended performance. I.e. the highest acc would be 0.01 and the biggest TS would be 100 for a 100% accurate shot, always.

I admit that I don't care so much for SP and it's my own fault. At this point we're should debate to make s separate branch for MP than for SP, similar to what the COD series did. Obviously because of balance and active updates.

Could we post here all the acc/ts discrepancies and possible fixes too?

I think its unlikely that any big changes happen at this point, for better or worse.

The dramatic increase in penetration at point blank range has been suggested by multiple people now and is probably the only one that has a real shot of being implemented, and would basically redefine Heavies as long range specialists.

I'm not sure its the ideal solution, as diving with mediums becomes more and more dubious the longer the game goes, but it would for the first time give a clear difference in the way mediums and heavies are used.

Seems like you're over-complicating the problem. Between armor-penetration, and vehicle size-accuracy you can create 4 different classes of vehicles e.g. high armor low size, high armor high size, low armor small size, and low armor high size. You can also throw in the variables of HP, dmg reduction, ROF, and damage. This is more than enough to express unique values for 20-30 something vehicles.

Could more effort be invested into evaluating the variables defining each vehicle? Sure, but adding in even more variables is just going to magnify the effort into balancing things, not help.

The game can't really support units that do well against mediums, but not against heavies. Not in a theoretical sense, but in a fun sense. Do we really want to be building a PaK 40 for LV, have it become obsolete 5 minutes later, then build a PaK 41 for mediums and PaK 43 for heavies?

COH2 excels because of it's elegance IMO. It solves complex problems in the simplest way possible. There are tons upon tons of games out there that have 20 different stats and 1,000 different units, and no one really plays them.

With all that said I do prefer CoH1's vehicle system over CoH2, and I hope that if we ever get a CoH3 it's not WWII yet again, because we'll see even more poor decisions for the sake of making the game "new".

The suggestion I have made year ago seem allot easier to implement imo.

https://www.coh2.org/topic/68297/super-heavies-td-overhaul-mod

Instead of messing up with all value of target size and accuracy one only changes super heavies.

The idea is that Super heavies get super size.

Unit that meant to hard counter super heavies get switchable rounds that are separately balance to deal with them with low accuracy high penetration and damage ROF accordingly.

Another point worth mentioning is that ROF for most vehicles is not effected by range. If one make ROF changes with range one can make some unit gain more advantages from closing in.

The suggestion I have made year ago seem allot easier to implement imo.

https://www.coh2.org/topic/68297/super-heavies-td-overhaul-mod

Instead of messing up with all value of target size and accuracy one only changes super heavies.

The idea is that Super heavies get super size.

Unit that meant to hard counter super heavies get switchable rounds that are separately balance to deal with them with low accuracy high penetration and damage ROF accordingly.

Another point worth mentioning is that ROF for most vehicles is not effected by range. If one make ROF changes with range one can make some unit gain more advantages from closing in.

I had similar ideas to you and the OP as well and spent a considerable amount of time thinking about the tank combat in this game.

COH for the most part is loosely based on Rock/Paper/Scissors approach to balancing. For example Machine Gun counters Infantry. Mortar counters Machine Gun. Infantry owns mortars. Like I said loosely based on Rock Paper Scissors. Now when it comes to Tanks, at least at the end game you have only Tank Destroyers and Heavy Tanks.

The reason for this is because just like the interaction between LMG Troops and Close Range Troops, it is generally better to have LMG Troops in an open field than SMG Troops. Sturmpioneers for example if they manage to get close they can do some serious damage, if they are far they will usually lose.

Medium Tanks are basically in a similar position as SMG Troops. They are not as effective as Tank Destroyers Long Range but even if they get close (assuming they can get close) They are usually almost dead at that point and or aren't much more effective than if they had stayed in long range.

So then the solution I came up with is, Allow Heavy Tanks to flat out Counter Tank Destroyers but then allow for Mediums to Counter Heavy Tanks. So lets take a T-34 for example.

T-34 Penetration
near value="120"
mid value="100"
far value="80"

The Near Penetration is not that much different than the far penetration. If we were to take the Near Penetration and made it say 240 then we could have a similar unit interaction to infantry where:

Heavy Tank = Essentially a Machine Gun
Tank Destroyer = LMG Troops
Medium Tanks = SMG Troops


So the purpose of this is to give mediums a clear defined role instead of acting like a shock unit like the T-70 where it has a small window of usefulness before the Tank Destroyers and Heavy Tanks hit the field and promote combined arms. This is why I had proposed that each faction have access to heavy tanks as part of the core army instead of via doctrine so something like this could be implemented. Also from a historical and realistic standpoint allied tanks had to get close to penetrate German Armor as the guns were not as effective from long distances. It would make sense to give Mediums an SMG Type of role in tank combat.

Main difference here is that vehicles move allot faster and it would be allot les punishing to simply rush the enemy frontally.

My suggestions was to create another type of "flanker" vehicle. The characteristic should probably be:
high mobility
fast turret rotation
high ear armor
and instead of high near penetration higher near ROF so that it would work better when flanking to straight forward driving on the enemy vehicle.
Possibly one could also add a higher engine damage threshold from snares.

A thread of that value needs to be upped. Just a comment for the UPs

While I appreciate the effort and thought that has gone into this idea, I don't think that it is the best way to go at the end of CoH2s life cycle.

TL;DR version:
I don't think that this higher resolution by 1-2% will provide enough benefit to legitimate all the issues it will likely cause. This proposal would have been perfect 6 years ago where there was still enough power behind CoH2 to make big changes and lay a good foundation, but now I think it is too risky.


For the sake of being, I will sum up what your approach gives us:
It widens the target size differences from 12 to ~30 depending on the approach and formula. The latter approaches, which would be preferred by you from what I can tell, widen the differences to "only" ~25. Overall, we get a higher "resolution" of target sizes by a factor of 2 to 2,5. While we currently need to fine-tune with about 4-5% natural hit chance, this would allow us to fine tune with ~2% natural hit chance. At this point I will already point out that the actual hit chance is lower than that. From my simulations between a T34/76 and an PanzerIV, I saw that about half of the scatter shots still hit (standing shoot out; take this with a grain of salt as I do not have the actual in-game hit box sizes of the tanks so I took the real world size for calculation). Assuming this is true, this would bring the actual hit chance down from 2-2,5% (current target sizes) to ~1% (OP's proposal) for each target size changed.


The two biggest concerns I have:

- While you can normalize the natural hit chances and compare them, the actual hit chances will be different due to scatter hits. as described above. I am not sure how "stable" these conversions are to rework all units without introducing artifacts, as there are quite a lot of different accuracy profiles and target sizes.
Of course this is also the case with the current implementation, but reworking all target sizes and accuracies will lead to multiple patches of rebalancing and for the first patches we probably have to hope that we will reach the current level of balance again.
The other option would be a complete rescaling of scatter as well, but this would then affect anti infantry performance and just cause additional unwanted problems.
So to make the first point short: While your idea brings a lot of benefits, I am not sure if there will be enough support by the community, balance team and Relic to go through what will probably be multiple iterations of rebalancing a big rooster of units.

- The other issue is that I don't know how large the benefit is to rebalance units compared to the current model.
The game provides three layers of RNG to fine tune combat: accuracy, scatter and penetration. Game play wise, it does not matter if the shot missed or bounced.
If the issue with small vehicles is that they are hit too often by TDs, we can adjust this with mainly target size. If they are too good vs heavies we can adjust mainly the armor. The effect of target size reductions of small vehicles are also more difficult to calculate, but for this case we do not need to go through multiple cycles of patching beforehand. And also your model does not guarantee that we do not need to rebalance armor or something else afterwards, it just allows us to do this more precisely. And to be honest even while I do not doubt the overall qualification of the balance team, I don't think that any of them will be able to make a difference between the impact of 1% or 2% higher hit chance. I don't think anyone is able to say "No, these -3% definitively won't cut it, but -4% is the perfect spot for unit XYZ" with any guarantee. The previous -3% coupled with -5 armor or so might have effectively done the same.