Bug Report: Swap logic broken - can't get no 75 MM AT guns

[rmonical]

In my test starting with the 42-45 scenario and 7.06, I ran AI verses AI for 40 turns. Then I started running the German side with a primary goal to get the 75 MM AT Production going. Computer continued to run the Soviets. On T-49, 7.07 came out. I stepped through 8 turns to evaluate production. Rifle squads were better. Not a single 75 MM AT gun was produced and those in the active pool were not swapped out even though many divisions had a handful of 37 MM guns (less than 5).

To try to encourage the upgrades, I broke divisions into regiment, moved one regiment off of the line and put that regiment in refit mode. Still no 37 for 75 MM swaps.

It appears to me the only way a unit gets 75 MM AT Guns is if the last 37 MM gun is removed via combat or attrition. Then, I suspect it is a crap shoot as to whether it gets a 75 MM gun or a 37 MM gun via the reinforcement segment.

The other swap element I tracked was the CS tank in the panzer divisions. The Panzer IIIns were building up to a couple of hundred. Most panzer divisions were down to 1 or 2 Panzer IVfs. The swap would not trigger until I took all of the panzer units off the line and put them in refit mode. Then I got one or two per turn.

So, in the case of the produced unit, swaps can eventually be made to fire.

In the case of the A unit, swaps do not appear to fire until some critical mass is achieved in the active pool and then very rarely. In another test in which I added a 75 MM AT Gun factory, I seem to get some swaps, but other things were not working (90K Hiwi in the pool on T35???).

I would be interested in reports from games into '43 on the number of 75 MM AT Guns. Historically, Germany produced 2100 Pak 40s in 42 and 8700 Pak 40s in 43. In addition, there were 2800 Pak 97/38s built in '42. These were French 75s on Pak 38 carriage. I do not think these latter guns are represented in the game.

[TAIL_GUNNER]

Here's something to try...

in an old 1.05 scenario I made..all AT Guns were set to upgrade to themselves...this link was taken out in some later patch.

worth a shot I reckon.

[rmonical]

Up to T-86 AI verses AI mode. Still problems with the 75 MM AT Guns. Less than 500 produced. Also, still problems with rifle squads not upgrading and the active manpower pool is increasing.

[Denniss]

This is also my experience with the 37mm, once reduced to zero and nothing received from pool it will be replaced soon. I actually don't know why there's no 75mm equipment slot created, even at zero strength it may create some production and deliveries. I have this seen working in 44/45 with tank and mech corps (although with tanks, t-34 m42 and /85 mixed in TOE, if unit upgrade to all-/85 during refit, in normal mode the m42 slot will be recreated).

[morvael]

Please run the test again, with added upgrade path from 37 to 75. Anyway, no way juggling the data will fix the problem as a proper fix in the code to production and swap logic.

[Denniss]

37->75mm path would see them replaced too fast, especially on refit (just like MG 34->MG 42). You'd see lots of replacements if you switch the 37mm PaK from AT-Gun to Light AT-Gun as they are most often removed on TOE upgrades. But usage as light AT might have side-effects.

BTW both the german 37mm and 50mm AT-Gun should have their end dates extended (but not their production) as they are in TOE/OB use since 44 or 45. This could be achieved by upgrading them to a non-producible version on the normal end date while extending the end date of the original.

Often the only hope to get more 75mm guns is the 47mm gun as it has an upgrade path to 75mm and is sometimes swapped-in for the 37mm. But I dislike see the 47mm upgraded to 75mm and swapped back to 47mm in the same turn (in the same unit) - upgrade and swap working against each other.

If Pavel can't touch the swap engine because he lacks time to look into the overly complicated function, maybe a change of sequence may help (firing swap earlier than upgrade or moving both to fire earlier than production)

[rmonical]

T-106 of '42 scenario (June 44) AI vs AI and less than 700 75 MM AT guns produced. Nasty little problem. I've been designing and writing software for almost 30 years and I cannot think of an easy algorithm to fix this.

As mentioned, the problem with the third AT gun slot is the old style guns get pulled back too rapidly. Same for using a second rifle squad slot.

The next thing I am going to try is a multi-step OB process to get to the '43 state. A new TOE every 6 weeks or so to get to the '43 division. I will gradually increase the new style elements while decreasing the old style elements. Using this process, I think the extra slots will not result in too many elements getting pulled back.

I think the 7.07 fix may let players get through the rifle squad upgrade if they pay a lot of attention to it from the start of '43. The AT gun problem is serious.

[rmonical]

I modified the '42 scenario to include a 4 step evolution from the 42 to the 43 infantry divisions. I increased the transition elements by 25% with each step every two months. I am currently running the stock 42 scenario to compare the result. One feature of the gradual upgrade is the 75 MM AT Guns get deployed. Another feature is the '39 rifle squad does not get produced after 1/43. In the 7.07 fix, my first test (starting T-49) indicates the '39 squad gets produced throughout 1943. In addition, the 75 MM AT gun apparently does not get deployed in 7.07.

The net of this approach is a bunch of '43 squads get deployed in 1943. Not as many as are shown in the '43-45 scenario. That (with the early deployment of 18000 panzerfausts) and the AT guns will greatly improve the Axis AT capability over what I suspect most games experience. Using this approach, there were still 1000 fewer 75 MM AT guns than the '43-45 scenario. This is an offset to the early panserfaust deployment. I'm not sure how much.

I tried to figure out a good way to get the M42s and reduced headcount separately from the panzerfausts, but it does multiply the complexity of the upgrade process.

I think the 75 MM AT guns are important.

One problem with the wonderful detail in WITE is the difficulty in introducing new weapon types in a completely historical manner. Since I don't understand the combat system, I cannot assess the impact of these design decisions.

[Denniss]

Could you run another test with only a single modification to these 43 OBs - add a single 37mm AT-Gun?
This should force these guns out via oversize reduction and should also keep the 75mm gun slot

[rmonical]

ORIGINAL: Denniss

Could you run another test with only a single modification to these 43 OBs - add a single 37mm AT-Gun?
This should force these guns out via oversize reduction and should also keep the 75mm gun slot

Tried it already - that is exactly what happens. I did not publish screenshots because the results were so bad. The old type is pulled back far faster than the new type is produced. Also, we had the ubiquitous 37 MM filling both slots in at least one case.

[morvael]

Unfortunately, I can only repeat my last post. It's the algorithms that needs to be changed, not just data. If not for WitE then at least for WitW.

There should be separate upgrade paths for production (what does factory produces next) and field deployment (what does unit is equipped with instead of the old item). Infantry squads should upgrade in place, by paying only for difference in weapons. Different models of equipment should be allowed to coexist, even if division had a lot of them (I'm looking at you max-32-elements TOE and OB limit). Production and swap algorithms should be improved, to prevent starving of units if there are armaments and men in the pool, while at the same time allowing older and newer equipment to coexist peacefully (prevent all-or-nothing swaps/upgrades).

[rmonical]

It's the algorithms that needs to be changed, not just data.

I think one option may need to be some manual control of ground unit upgrades similar to the air unit upgrade options.

[skitchieP]

Been following these issues for a while and finally decided to chime in.

This same issue seems to also be affecting the Panzerschreck. It should begin production in 1/43, but I dont see any in your ss. And I don't have any in a 42-45 Human v. AI game I have been playing. The Panzerschreck shares the same spot as the other Anti-tank Rifle whose production goes all the way through 12/44. The ATRs are still being built just fine through 44 so we dont see the disaster where numbers go to 0, but its still blocking the swap to a superior weapon.

[skitchieP]

From a quick look over at the ToE tables in game it looks like new slots were created for the 75mm gun in the '43 InfDiv and for the Panzerschreck in the '44 Div. I ran my 42 GC out until Sept. '43 when the '44 ToE kicks in. As soon as I got to Sept my divs flipped to the new ToE and were instantly full of PSchrecks. I went from 0 to 500 in the first week and up to 1100 the next week. Obviously your not getting the PSchrecks when they should be produced on 1/43, but they do get there when the ToE allows. The same thing should be happening in Nov '42 when the '43 ToE kicks in, but it doesn't.

It seems like the 2 ATG slots are blocking new types coming in. Could you possibly rerun the three ATG slot test starting the last week of Oct '42. That way the next turn the ToE's should flip and A produce everything they need. BTW it looks like the Soviet divisions do this. They finally switch to a three ATG slot model in Dec '44, so we probably never see this happen.

[Denniss]

I'll do some test as well but the modification may be different than yours - keep the single 37mm ATG at the same spot like in the previous OB and placing the 75mm ATG between the 50mm and whatever follows next.
If that doesn't work then the problem is not only element swapping or TOE upgrade function, there should at least an empty slot for the 75mm ATG be created but it may get caught in some cleanup process and removed in the same logistic phase.

[rmonical]

I ran the stock '42 scenario AI-AI through T-54 (7/3/43). As predicted, more than half of the rifle squads produced in '43 were the '39 squads. This is wasted arms points for the Germans compared to the incremental upgrade scheme in which all of the squads produced in 43 were the '43 squads. Also as expected, the 75 MM AT guns are produced and deployed in microscopic numbers and the AI does not understand how to get the swap logic to fire so a few Panzer IIIns might make it into the game. I'll run it into 44 now to see the full upgrade cycle.

[rmonical]

Continuing on to 6/17/44 of the stock '42-45 scenario. The Germans have:
122K manpower in the active pool. Over 1 million armaments points.
Type 95 motorized rifle squad (not in the 7.07 fix) still dominates the panzer divisions and is far below TOE.
There are 686 PzIV H, 263 Panther D, 950 Panther H in the active pool even though the Panzer division are well below TOE. This is because Pz38s are still in 5 divisions, PzIII L/60s are in 8, Pz IIIms are in 11. Panthers are in a grand total of 2 divisions - I guess because they arrived with them: 5th SS and 25th Pz.

The PzIIIns never made it into the game (500 in the pool). Even though many divisions are down to a single StugIII, there are over 100 Hetzers and 240 StugIVs in the active pool. Not a single one has been deployed.

There are over 900 Lt Tank destroyers in the pool. The Marder III was barely deployed and the Panzerjager38(t)H was not deployed at all. Going back to Jan 44 (T-81) it is clear that the reason is because the Marder IIs are locking up the slots even though they are at 10% of TOE.

All told, the Germans have almost 5000 AFV in the active pool and 6800 on the map. On the Soviet side, they have over 16K AFV on the map and 5600 Su76s dominating the pool of some 18K AFVs. I was surprised to see almost all of the lend lease types made it into the game. I guess the Soviet AI has greater luxury to place units in refit mode and exercise the swap logic.

The fact that the German AI cannot get the swap logic to fire on key element types confirms for me the swap logic is broken.

[morvael]

Swap logic may be harsh, but not broken. It's the AI that may be broken, being unable to work with swap logic in it's current form, which damages AI Axis a lot.

[rmonical]

Swap logic may be harsh, but not broken.

It is broken for A types in my mind until someone shows me a game where the 75 MM AT Gun is properly deployed without a special TOE like I created. We already confirmed it was broken for rifle squads - hence the 7.07 fix and the need to add the type 95 to 96 upgrade path to the next update.

It is overly harsh to the point of impossible for produced types. It is certainly ridiculous for the presence of 2 Marder IIs in a division to prevent the issue of other 43 needed elements that are sitting in the pool.

[rmonical]

I think this algorithm is a reasonable and hopefully easy fix for the produced type swap logic:

1. if the element type in the slot is obsolete AND
the element's TOE % is 25% below the unit average TOE % AND
there is a not obsolete type in the pool sufficient to bring the unit up to target TOE
then
execute the swap as an override to all other rules. Bring in the same number of elements of the not obsolete type as are swapped out.
2. else if the element type in the slot is NOT obsolete AND
the element's TOE % is 45% below the unit average TOE % AND
there is any type in the pool sufficient to bring the unit up to target TOE
execute the swap as an override. Bring in the same number of elements of the swapped in type as are swapped out.
end if

Discussion:
By forcing swaps, the obsolete low quantity types are forced into the pool where they can be allocated out again. They will gradually get concentrated and be attrited out. Realistically, they would usually have been kept in the unit until they were dead, but the engine works by slots.

Swapping in the newer type at the low quantity that is swapped out means that the replacement system will take over to fill out the unit. The idea here is that the swap overrides proposed will occasionally allow an unrealistic swap. In this case, the replacements will be constrained. This is not bad in my mind because so many perfectly reasonable swaps are not allowed by the engine.

The second clause is specifically designed to address extraordinarily low production rates such as the Marder II. On T54 of the test, 75 had been produced and 134 were in play spread across 24 units. Those 134 elements were locking up the slots for 1025 Lt Tank Destroyers. At that time there were over 500 Marder IIIs and Panzerjager 38s in the pool with one of the Panzerjager 38s types in 4 units. As indicated above, the Lt Tank Destroyer pool eventually grew to over 900.

One problem with any override of this type is potential churn. If the reinforcement system does not fill out the unit before the next swap cycle, then the element may get swapped by the else clause. For example, by June 42, all of the CS types are well below so as the PzIIIns ramp up, they will get swapped in by clause 1. So in one case, the division would go from two PzIVfs to twop PzIIIns. If somehow this happens several times and a bunch of PzIVfs build up in the pool. The next turn, the reinforcement process empties out the rest of the PzIIIns before getting to the unit with only 2. If (in the CS tank case) there are still 20 PzIVfs are in the pool when this unit is checked, then clause 2 would fire the swap again.

I think this will be a rare case. In the case of CS Tanks, the swap override is possible when PzIIIn builds above 20. The first (or possibly second) override takes it below 20 and no more swaps will fire. Then in the next reinforcement segment, the units will attempt to draw the remaining PzIIIns to fill out their TOE. Only the first few units that got the swap override will get the PzIIIns unitl they are up to TOE. Then more overrides might fire. So I think the logic will keep the active pool empty of PzIIIns and churn on rare occasions.

This proposal does nothing for the swap problems with A type units because there is rarely anything in the pool to swap in.