Bug Report - swap logic broken Pz38(t)E

[rmonical]

I am going to put this in another thread from the prior swap bug thread because the Pz38(t)E is produced and the other thread is focused on A units.

I created a test '41-45 scenario to see if adding a extra CS tank slot to the '42 Panzer divisions helped with the AFV swap problem. Then I ran AI-vs-AI. As recommended, I set the German values to 110 until Feb-42. As expected, there were no problems with the PzIIIns getting deployed with the extra slot. Swap was not required.

I interrupted my '41 test game T70 to step through some manual turns and get the medium tank swaps in because the AI is bad at putting units in refit mode. As of T79, I have a set of units that is not taking reinforcements. In one case - 9.Panzer, the weakest Pz division, it did not receive a single AFV from T73 to T79. I kept improving its situation finally moving more than 10 hexes from an enemy unit on t77. This division does have Pz 38(t)E in both Medium tank slots. The other divisions each have at least one Pz38(t)E slot. The 13th Pz had two Pz38(t)E slots T73 and by T79 it has a PzIIIj L42.

This process started with a bunch of medium tanks in the active pool, they were raqpidly farmed out once the Pz divisions were in refit mode. Since about T75 the number of medium tanks in the pool has been increasing unless a swap fires. And to add insult to injury, 4.Panzer started with 15 PzIIIj L/42 & 9 Pz IIIm on T73 and had the out of production Pz38(t)E swapped in for the in production PzIIIms. Even with 9.Panzer the only in refit mode on T79, the swap did not fire for either of the medium tank slots - each with 6 Pz38(t)Es. This graphic shows the starting deployments and the evolution of the active pool while I was manually stepping through the turns.

[rmonical]

This graphic shows the situation as of T80. The Pz38s are in 13 divisions and 14 slots. The PzIVf2 has gone from 4 to 1 slot with only 21 tanks deployed and 58 sitting in the pool. Swaps are not firing.

[rmonical]

Stepped through to T85 and got a single swap. Just to reiterate, all of the divisions have been in refit more than 10 turns. Off the line. The 9th, which got a swap was 10 hexes off.

What makes this doubly irritating is the Pz38s in the pool are not all allocated out as reinforcement as are the newer types even though there is essentially infinite demand.

I'm going to rewind to T70 and restart AI vs AI.

[morvael]

Thanks for finding example after example of problems caused by lack of proper and timely swapping. Maybe if you persist, they will finally try to fix the code, instead of just altering the data a bit. For rifle squads it was doable and worked, but with guns and tanks it's harder to do, due to more types, with overlapping production times, multiple slots in TOEs, and (in case of non-autobuild elements) the need to control how do the factories upgrade.

[TAIL_GUNNER]

Interesting finds...nice job.

For a mod I did awhile ago, I changed the Pz35(t) and both types of Pz38(t) Ground Element Type to Infantry Tank and Cavalry Tank. This had both the desired affect of keeping these tanks in their proper divisions, and dumping them all out when a TOE upgrade occurred to replace with regular German tanks.

What would be real nice is the ability to create custom Ground Element Types in the Editor....or even to copy and rename current ones.

[rmonical]

So, I threw away all of the manual swap I tried to accomplish in late 42 and ran AI-vs-AI until T114. This was the panzer situation.

[rmonical]

I put the 9.Pz and two others in refit. The swaps worked pretty well with the oddity that PzIIIms were swapped into three divisions at quantity 0. It appears that the nice batch of PzIVf2s are forever banned from deployment. The Pz38s that were swapped out were allocated back out to the remaining Pz38 equipped units. In the case of 7.Pz Panther Ds ended up in all three (?) slots. A couple more turns on refit and the swap engine swapped in PzIIIms into the second slot.

[morvael]

I'm less concerned about problems the AI has. Of course 2by3 declares that games vs AI are important for them as they account for 80% games played. But sometimes it's hard to teach the AI to do such sophisticated things like garrissoning, refitting in the rear etc. For me the problem is mainly that even human player can't help if there is no manual swap option and the alghoritms do not work properly. For both PBEM players and AI players, swaps should be fixed...

[Denniss]

The IVf2 may have been deployed but then upgraded to IVg.
The 38(t)E overdue for swapping is indeed a problem as it also affects exports to Hun/Rom.

[rmonical]

By T118 the slots were reasonable. The AI's decision to not use the f2s is a frustrating mystery. The PzIVg L/43s and Pz38s are sitting in the pool because the units so equipped are up to strength on medium tanks.

So the major issue with the swap cycle is the engine created more slots for the obsolete PzIIIm even though there was still a decent number of current PzIVhs in the active pool. At the time the PzIIIms were swapped in at quantity 0, 3 of the 5 units so equipped were under strength medium tanks. On T118, the swap engine put PzIIIms into 7.Pz in preference to either of the obsolete PzIV types with a decent number in the active pool.

So I see these issues here for swapping produced element.
1. Unit must be in refit mode which means the AI never supports swaps.
2. The selection logic for what to swap in will sometimes put the same element type in multiple slots.
3. The selection logic sometimes selects an obsolete type in preference to the current type even when the size of the active pool is comparable.
4. The stinginess of the reinforcement engine means that obsolete element types are available to swap even though they are needed by the units already using that type.
5. The required pool size to induce a swap is too high. This swap exercise worked reasonably well because there were over 500 PzIVh and Panthers in the pool.

[rmonical]

The IVf2 may have been deployed but then upgraded to IVg.

The f2s were fully deployed in June 42. They could as easily have been "upgraded" to Pz38s by the swap engine.

[rmonical]

I went back to T66 to eliminate weather effect. The first thing I wanted to do was get the Pz38s out of the active pool so they would not get swapped into new slots. So I put all of the units with Pz38(t)E in refit mode. On T67 the active pool went from 105 to 92. On T68 the active pool went down to 60 as Pz38s were swapped into 3 slots on divisions that already had Pz38s. AAAARRRRRGGGGGHHHHHH!.

So the engine is stingier on allocating Pz38s as reinforcements than it on allocating them for swaps. Go figure.

[rmonical]

My current human-v-human game is on T-58. It is much bloodier than the AI vs AI games I tested. It looks there may never be enough Pz38(t)Es in the active pool to trigger the problems I see here.
I got the PzIIIn swap on the one unit I placed in refit mode.