Matrix Games Forums

ORIGINAL: Joel Billings

snip

Thanks for the additional very interesting info. I'm glad the game continues to get better and look forward to playing it more and seeing how things end up!

At least the preliminary results hint that the AFV losses seem too high in the current patch. I'll try to get some results from an AI vs AI simulation when I have the time to do so.

Judging from the few results seen by myself, and other examples on the forum, battles with low manpower losses but dramatically high AFV losses seem common. How does that add up in the long run? Malyhin's almost 8k AFV lost in 4 weeks seems like a high number to me.

SU total losses during the war were around 83k tanks, 13k SPGs and 37k APC/halftracks, that's a total of 133k during the whole war. 8k during the opening month suggests nearly 100k/year losses, which of course is a high assumption due to the opening pockets creating huge AFV losses too.

ORIGINAL: Nix77

At least the preliminary results hint that the AFV losses seem too high in the current patch. I'll try to get some results from an AI vs AI simulation when I have the time to do so.

Judging from the few results seen by myself, and other examples on the forum, battles with low manpower losses but dramatically high AFV losses seem common. How does that add up in the long run? Malyhin's almost 8k AFV lost in 4 weeks seems like a high number to me.

SU total losses during the war were around 83k tanks, 13k SPGs and 37k APC/halftracks, that's a total of 133k during the whole war. 8k during the opening month suggests nearly 100k/year losses, which of course is a high assumption due to the opening pockets creating huge AFV losses too.

There's an important point here which you and some other people may be overlooking.

From a technical/mathematical perspective, it is actually NOT correct to look at the total number of AFV losses during the war (or over any reasonably long period of time such as a year or half a year) as a metric of whether the game's variables for AFV losses are calibrated correctly. The reason for this is that for pretty much ANY way that the combat is set up within remotely reasonable bounds, over the course of a long period of time, the sole determinant of aggregate losses are production. Losses will always approach production asymptotically over a sufficiently long period of time, at least until you get to the point where production is totally out-stripping losses because the enemy is totally defeated and has barely anything left (i.e. Soviets in 1945). For fundamentally the same reason, the amount of water that flows out of a lake or evaporates will always over a sufficiently long period of time tend to equal the aggregate amount of water that flowed into the lake from its tributaries (and similarly for all sorts of other physical and other systems).

So instead of looking at total losses over a long period of time, the metrics that should be looked at and compared to historical data are the RATIO of AFV's lost per turn as compared to the total amount of AFVs you have at the start of each turn (i.e. combat intensity), or alternatively the total size/level of the AFV stocks that each side has in their OOB over time, which in effect is measuring the same thing as that ratio (source: I do mathematical modeling).

AFV losses, as well as all other losses in the game of men/planes/guns/etc follow a fairly simple mathematical process, which has a stable equilibrium level of amounts of equipment each side will tend towards having, as long as combat intensity is roughly consistent and production doesn't vary dramatically, as long as combat intensity is not at super-low levels/non-existent (in which case the equilibrium could be infinity). The reason for that is that losses are at least partly a positive function of how much equipment you have in your OOB. If you start a turn with 13000 AFVs, other things equal you are going to lose more AFVs than if you start the turn with 3000 AFVs - that is the equilibrating mechanism through which losses adjust over time to match production.


Specifically, the system can be mathematically expressed in the form of two simple difference equations, where OOB stands for the amount of equipment in total in the OOB, P stands for the amount produced each turn, L stands for the amount of equipment lost each turn, and CI stands for combat intensity (or the proportion of equipment that is on average lost each turn). The value of CI will always be bounded between 0 and 1, since it is physically impossible to either lose negative equipment or to lose more equipment than you have.

Equation 1: OOB(t+1) = OOB(t) + P(t) - L(t)
Equation 2: L(t) = CI * OOB(t)

Substituting: OOB(t+1) = OOB(t) + P(t) - CI * OOB(t)

Re-arranging: OOB(t+1) = OOB(t) * [1 - CI] + P(t)

Drop the t's to solve for the equilibrium of the system: OOB(t+1) = OOB(t) * [1 - CI] + P(t)

Equilibrium solution: OOB = P / CI


This means that the equilibrium size of the amount of AFVs (or any other equipment) that you will have is solely determined by production and combat intensity. Any time that production changes (e.g. going to a new year when the game updates production stats) or that combat intensity changes (e.g. a change in the weather leading to fewer attacks in mud/winter), the equilibrium will shift, but as long as production and combat intensity remain roughly constant, the equilibrium will also remain roughly constant.



So, how do you take this mathematical understanding and apply it to balance losses and combat intensity in the game to ensure it matches history? Take another look at the graph of the total size of the Soviet AFV OOB in my StB game. Notice that my AFV stock started at around 13,000 (which I presume is a roughly historical value based on the game designer's research) and then immediately starts going down. But as it is going down, the rate by which it goes down gradually decreases and it starts to stabilize. It looks to me like probably the equilibrium value of Soviet AFVs, given the production and combat intensity, was maybe something in the range of 7000 or so. So I started the scenario with an AFV stock that was far from equilibrium, and then over time it started approaching equilibrium.



Now compare that to this:



The way that "reagants" go down gradually and then stabilize looks quite similar, and that is not a coincidence, it is a reflection of the game's system approaching a (temporary) equilibrium similar to how the chemical system is doing so.

At the start of the war in 1941, each side had AFV stocks that began far from their equilibrium values. As a result, e.g. the Soviet AFV stock in 1941 quickly dropped by many thousands. Then it started to go back up as production picked up and as the combat intensity declined a bit (a lower proportion of each side's total AFVs was lost per week after the initial huge battles like Brody and Raseinai etc). By the time the StB scenario begins, the war had been going for long enough that it is reasonable to suppose that the AFV stock was at something approximately in the general neighborhood of its current historical equilibrium level.

So, the fact that my AFV stock was going down from 13,000 (roughly the true historical level) down towards 7,000 or so tells us that the historical equilibrium of Soviet AFVs during the winter of 1942-43 was quite a bit higher than the game's equilibrium level of Soviet AFVs. The only possible explanations for that are either that the game has production wrong (we can pretty safely rule that out) or that combat intensity with AFVs in my StB game was quite a bit higher than historical AFV combat intensity. And there are two possible explanations for why combat intensity for Soviet AFVs has been higher in my StB game than historical - either AFV losses were simply too high as compared to what would be historically accurate (in which case they would be even more too high in this new patch), or else I was playing more aggressively than the Soviets did historically and being more aggressive in using my AFVs and getting them into combat. The true explanation is probably some of both, to be honest.

So to accurately calibrate AFV losses, as well as plane losses etc including the much discussed operational losses, what you need to do is to try to figure out exactly how aggressively equipment was being used in combat historically (i.e. were the historical Soviet more or less aggressive than me), reproduce that same level of historical aggressiveness in the game, and then compare the equilibrium OOB levels that the game tends towards to known historical rough equilibrium levels, based on any periods of time when your research gives you approximately accurate statistics on the true historical OOB sizes of various equipment types.

But what you don't do is compare overall historical losses to overall losses in the game. The game will basically always reproduce historical losses, for the simple reason that the game has historical production and that losses are an increasing function of your OOB. So that metric would essentially ALWAYS tell you that the game is correctly calibrated, even when it is not, because that is the wrong variable to look at and it does not have any relation to an equilibrium value, as a technical/mathematical matter!

Here's a bit more analysis to show what you can tease out from understanding this. Look again at the graph showing my AFVs. On the first turn, my AFV OOB dropped from 12,829 to 12,346. That means that I lost 483 AFVs, plus I lost an amount equal to my production of new AFVs. I am not exactly sure what production per turn is, but I think it was something like 500 AFVs per day or so produced, so let's just assume it was that. If so, then I lost something like 483 + 500 = 983 AFVs on the first turn. Since I had 12,829 to start with, that means that I lost 983 out of my 12,346 AFVs, or about 7.7% of the total. In other words, "combat intensity" for my AFVs was 7.7%, or 0.077 that turn.

If you do that same calculation across all the turns, you get this:



So on average I was losing about 7.5% of my AFVs each turn, with some variation, but overall seems to have been staying fairly consistent, which generally makes sense.

If production of AFVs was about 500 per turn, and combat intensity was about 0.075, we can plug those two numbers in to the equilibrium formula to calculate the equilibrium OOB size of AFVs:

OOB = P / CI

OOB = 500 / 0.075 = 6666.67

So the equilibrium the game was going to was around 6,667, as compared to a true historical approximately equilibrium level of 13,000 or so. Since combat intensity is mathematically linear with the OOB, that means it was about 95% too high, from whatever combination of me being more aggressive than historical and from the combat model simply resulting in too many AFV losses. Depending on how much you think the explanation was me being too aggressive, that tells you how much you need to change average AFV losses in the combat system to reproduce roughly historical results in the game.

I've noticed Germans lose a bucket of AT guns in turn, when Infantry Divisions attack Soviet Armour / Mechanized - in the first turns of the game already.

ORIGINAL: Beethoven1

ORIGINAL: Nix77

At least the preliminary results hint that the AFV losses seem too high in the current patch. I'll try to get some results from an AI vs AI simulation when I have the time to do so.

Judging from the few results seen by myself, and other examples on the forum, battles with low manpower losses but dramatically high AFV losses seem common. How does that add up in the long run? Malyhin's almost 8k AFV lost in 4 weeks seems like a high number to me.

SU total losses during the war were around 83k tanks, 13k SPGs and 37k APC/halftracks, that's a total of 133k during the whole war. 8k during the opening month suggests nearly 100k/year losses, which of course is a high assumption due to the opening pockets creating huge AFV losses too.

There's an important point here which you and some other people may be overlooking.

From a technical/mathematical perspective, it is actually NOT correct to look at the total number of AFV losses during the war ... Losses will always approach production asymptotically over a sufficiently long period of time, ...

fully agree, one reason why the air war is so often off kilter in player tests (vs the historical record) is the opposite effect. Players (esp Soviets) tend to horde their assets and since you can outproduce your on map capacity (ie you don't have the air groups to use up what you have), then losses end up relatively low.

with armour, its much harder to do this, you have much less control over assignment to formation, in the main formations are harder to rotate and there is a lot of secondary damage going on. So yes, once any initial rebalancing happens, losses follow production simply as you can deploy and use pretty much all that production.

But it will becomes circular if one try to measure CI (Combat Intensity) by looking at their AFV losses... if in final, it's for readjusting the AFV loss rate, no?

Don't forget that if the other side has not moved yet you could be fighting against a side that is over command limit which could have an adverse effect on the results. Just wanted to throw that out there.

Excellent arguments by Beethoven1 above, there are of course multiple variables when inspecting AFV losses as whole.

I guess my point mainly was that in relation to previous loss results, the new results seem quite dramatic. I'm not sure if they really are, I've been on WitE-vacation for quite some time

When comparing them, I would be assuming the production rate is comparably constant, and "combat intensity" would be the same for testing purposes. This is difficult to achieve in HvH games, but one reasonably reliable way to test this is to run multiple AI vs AI games and see where the AFV losses settle under each patch.

quick confession, the last (game) year of my vs Soviet AI AAR was played with a variety of beta patches that had these changes in.

So, broad view, it made a huge difference but I think broadly a good one. I put up a load of battle reports in the beta forum so that Joel et al could track what was happening and essentially German tanks did well at range. Which is what was intended - now that didn't just mean more Soviet tanks lost it also meant less Soviet tanks coming into range to damage German tanks.

But if they got into range, I could lose 60-80 tanks in a bad battle, so my feeling is its not protecting German armour as such, just means you get first shot.

What it meant in terms of game play as 1943 went into 1944 was I could still use the Pzrs offensively (just) and was still trying for pockets. As my infantry fell apart, the Pzr/PzrGr formations became more static but could still blunt the Soviets even late game.

Now that was an AI game and the AI chucked away a lot of its armour in early-mid 43, so it may well be that a human Soviet player will concentrate better and overwhelm any Pzr based defense.

But it did feel better - and, to me, more realistic

And the logical result of such losses in the 5th week of battles!

ORIGINAL: loki100

quick confession, the last (game) year of my vs Soviet AI AAR was played with a variety of beta patches that had these changes in.

So, broad view, it made a huge difference but I think broadly a good one. I put up a load of battle reports in the beta forum so that Joel et al could track what was happening and essentially German tanks did well at range. Which is what was intended - now that didn't just mean more Soviet tanks lost it also meant less Soviet tanks coming into range to damage German tanks.

But if they got into range, I could lose 60-80 tanks in a bad battle, so my feeling is its not protecting German armour as such, just means you get first shot.

What it meant in terms of game play as 1943 went into 1944 was I could still use the Pzrs offensively (just) and was still trying for pockets. As my infantry fell apart, the Pzr/PzrGr formations became more static but could still blunt the Soviets even late game.

Now that was an AI game and the AI chucked away a lot of its armour in early-mid 43, so it may well be that a human Soviet player will concentrate better and overwhelm any Pzr based defense.

But it did feel better - and, to me, more realistic

Good points and I think whatever we need to see months of game combat in differing times of the War to get an overall perspective.

I agree with you though that IF the quality of the men/women and machine matter now then that's a good thing. We all know and can see the loss ratio of German AFV vs Soviet AFV in the War. It should at least lie here as the basis for any WITE 2 game.

ORIGINAL: loki100

But it did feel better - and, to me, more realistic

I agree, the change seems good from balance POV, but it feels that the AFVs get sometimes committed too aggressively in some of the battles. In these battles, usually the attacker CV skyrockets to hundreds of thousands, and defender drops to 0, and AFVs suffers major losses.

I'm not sure what exactly the "disrupted" column in ground losses details exactly means, but in these catastrophic battles, the number of disrupted elements seems to be over the actual number of the elements?

a given element can be disrupted many times in a particular battle, so its quite feasible for the shown number to exceed the number of elements.

Just that once its distupted its more vulnerable to damage/destroy if hit again and also (of course) it no longer counts in the cv calculation.

I don't think beyond that it makes any difference if it had a single disruption or 20.

I'd suggest maybe more informative to look at a clash where the Soviet formation had T34/KV1 elements - may give a better feel for how this will work out as 1841 progresses

ORIGINAL: loki100

a given element can be disrupted many times in a particular battle, so its quite feasible for the shown number to exceed the number of elements.

Just that once its distupted its more vulnerable to damage/destroy if hit again and also (of course) it no longer counts in the cv calculation.

I don't think beyond that it makes any difference if it had a single disruption or 20.

Ok, the disruption number makes sense then.

ORIGINAL: loki100
I'd suggest maybe more informative to look at a clash where the Soviet formation had T34/KV1 elements - may give a better feel for how this will work out as 1841 progresses

Not sure about the 1841 Opium Wars...

Jokes aside, it's true that the results I'm checking are from early stages, and the older Soviet tanks are surely playing a role in these catastrophic results. Will need to test some later results with T-34 or KV-1 involved.

Here's another perhaps a bit alarming result, with a very decent morale 50 Rifle Division supported by a weaker Tank Division getting absolutely murdered by panzers with artillery support... maybe Kostenko was following Order #227 by the letter?

May need to update those loss rates lol.

Some of the screen-shots posted does not really seem that bad from Soviet's perspective.

Even the 2-1 odds attack against the German 7th Pz Division, on TURN 2.

As the Germans, I would be horrified. 178 AFVs out of 253 destroyed (even at the cost of 679 AFVs) is a nightmare for the Germans.

The 7th Pz Division is effectively, emasculated for the rest of Barbarossa.

Where as the Soviets fart more than 679 AFVs at a time. On TURN 2, those are most likely crap Soviet AFVs anyways.

Stalin would be dancing in the Kremlin after receiving this report.

ORIGINAL: Sauron_II

Some of the screen-shots posted does not really seem that bad from Soviet's perspective.

Even the 2-1 odds attack against the German 7th Pz Division, on TURN 2.

As the Germans, I would be horrified. 178 AFVs out of 253 destroyed (even at the cost of 679 AFVs) is a nightmare for the Germans.

The 7th Pz Division is effectively, emasculated for the rest of Barbarossa.

Where as the Soviets fart more than 679 AFVs at a time. On TURN 2, those are most likely crap Soviet AFVs anyways.

Stalin would be dancing in the Kremlin after receiving this report.

The AFV exchange ratio seems OK to me (4:1 for Germans in that 7thPzD battle), but the total amount of AFV destroyed is the crippling part, especially for the Germans in this case. The Soviets for sure had the upper hand in that battle with 5 divisions facing one GE PzDiv, and the 2:1 victory is well deserved, but the AFV losses got out of hand in that battle too, like in many other examples.

Just guessing here, but it could be that the AFVs are now firing & getting fired at too many times in single battle?

I wouldn't focus too much on the results of one battle screenshot, both due to variability and the fact that the full pre-battle situation is not available to us. It's best to play through multiple turns and gauge the overall pattern you're seeing, then report back to us. Keep in mind also that there are historical examples in the early war of significant tank losses on the German side, just not generally without higher losses on the Soviet side which we do see here.

Most German tank losses in the first months of Barbarossa were recoverable, how many of these losses are?