[ADV] Advanced Play System - Questions & Commentary

[HexCode]

INTENT & UTILITY
===================

This much "rechristened" topic is dedicated to posts and, hopefully, follow-up discussions directly aimed at identifying and clarifying advanced play system and related technical information.

Although a subject's genesis may be traced back to some [DEV], even [EPH], discussion, the objective here is to come up with reasonably solid findings applicable across the board. So, please, no Mod "historical specificity".

There's no reason to adhere to some strict, sequential coverage of subjects. Not all subjects are equally tractable. However, when posting about a particular subject, it's always helpful to specifically entitle follow-up posts so as to make it easier for participants to follow the particular discussion.

The topic's title starts with "[ADV]". Essentially, the topic is akin to a [DEV] one, the only difference being that, instead of some particular mod, advanced play system and related technical knowledge are being developed, comprising a diverse set of specific subjects.

One can reasonably assume that the readers of and posters writing under the present topic possess reasonably detailed knowledge of PGF's ins and outs. "Novices" are bound to find the discussions herein rather difficult to follow, if not outright... irrelevant.

Note: There's a very good chance that some of this topic's contents will "migrate" to PGF's Online Library in due course.

[Radoye]

I don't recall seeing the AI purposefully creating a bridge - but i did saw it using one to move units across if it was provided by scenario design. That is, until the bridging unit itself is inevitably moved in the typical PGF AI fashion of aimlessly moving units about and chunking them haphazardly around victory hexes.

Although, i must say i wasn't paying too much attention to it - i can imagine a situation where AI stumbles into creating a bridge if there is a river just beside a victory hex or it has no other ways to get where it wants but to move onto a river thus creating a bridge for a single turn.

But then, we do know how to make anchors, do we?

[Lettos]

Elsewhere in this forum, I've formally documented that PGF's AI never disbands any of its units.
There're two related questions which come to mind here, especially in the context of innovative bridging.
1) If PGF's AI happens to find a friendly unit sporting bridging capabilities situated on underlying river terrain, will it ever take advantage of its said capabilities and actually utilize it to allow other friendly units to cross ?

I've tested it now in sandbox scenario. Without any doubts - yes, it will.

2) Depending on the empirical response to preceding question (1), ...

We have not only an empirical response but a practical too ... but (see below)

... will PGF's AI ever purposively situate a friendly bridging unit on underlying river terrain and subsequently utilize it to allow other friendly units to cross ?

Any experiences ?

1) AI priorities in movement of unit classes make almost incredible to use a bridge engineers to cross rivers.
Sequence is: Recons - Tanks - AAircrafts - Artillery - AD - Infantry.
Imagine a "army group" attacking an enemy deployed far behind the river.
1st turn: First of all AI will move recons to river bank. Then will move tanks and tanks will occupy almost all river bank. After tanks AI will move guns on trucks. How the bridge engineers will come to river bank to prepare the bridge on next turn? But even if it occurs and bridge engineers will be placed on river bank (1 against 100 that it is possible) what will happen on next turn?
2nd turn: First Recons, then Tanks, then guns will go on river tiles, and only after it the bridge engineers will start their "job". And only rare infantry units will be able to use this bridge.
3rd turn: Recons, Tanks and Guns will go ahead. Then AI will send bridge engineers in attack using them as infantry unit. And some units will remain on this side of the river. This is actually a war crime.

Conclusion: AI can't understand term "purposively" at all.

2) But if we'll create a bridge engineers not infantry but TANK unit? I'm sure You know about these units in WWII. Yes, they were only a small part of Engineer's battalions. But in PGF these Tank Bridges will go ahead as usual tanks! And there much more chances to see these Tank Bridges units on river tiles first. I've tested this idea now in sandbox - it works great. Only one turn, but it works!

Note: there is only one strange AI limitation using tanks. AI have his own priorities in his decision about which tank unit will move first.
And I'm trying to understand how AI decides which certain tank unit will move first. I've tested in sandbox. 10 different tank units x 3 pcs each of them. Tanks attacking infantry on clear terrain, in same conditions. Certain tanks always go first, second and third. There is some kind of rules but I haven't ideas now about this rule. Only observation of strange fact. And I'm supposing we can in tests find the Tank Bridge Eng unit parameters to give them priority to be moved first. If we really need to have them moved first, of course.


Edited and added later:
A) Tested in sandbox: AI will first move tank units with less MVT. Very simple. Very intellectual. Very "unpredictable".
If there is not possible to move these slow units at the beginning of Tank units movement stage AI will not move them during current turn. Even if after faster tanks were moved and necessary hexes has been freed.
IMO: probably Tank Bridge units should have MVT 3 or 4.

B) Nerdyoldman already created Bridge tank in his mode viewtopic.php?f=95&t=174#p4746
"The bridgelayer is good in some scenarios, but SO slow."
It is now clear that the fast tank will not always be the first on the river bank. For example, if the starting position is 7 hexes from the river and the Tank Bridge has MVT=7, then yes, it will come running first. If the group of tank units is located three hexes far from the river, then Valentine with MVT=3 will be the first in right place.}

[Lettos]

Genesis

Recently, # Lettos # has been conducting intensive as well as extensive "experiments" on any discernible patterns underlying how PGF's AI moves its units and utilizes them in attacks.

Next Steps

I propose that other interested parties undertake such "experiments" as well. At some point, some general consensus is bound to emerge. For my part, I'll attempt to, perhaps, identify relevant code segments in PGF's executable.

I can provide a scenario where the Independent Judges only need to press the End Turn button a few times. And have a little popcorn and observation.

Download link:
https://drive.google.com/file/d/1o3rEkc ... sp=sharing
in zip archive:

Scenario TEST 1 Replacements. Enjoy a replacement sequence on 2nd turn when Allied will receive 20000PP. It is possible to tune injections to 1-2-3 thousands per turn and slowly check which units will be reinforced first. But I also did a replacement process a little bit visual in real-time

Scenario TEST 2 Tank MVT.
Deployed 4 recons with MVT=6,7,9,10.
Deployed tanks with MVT=2,3,4,5,6,7.
And Bridge T-28 tank with MVT=3 to check Tank Bridge too.

Set Allied = AI Intermediate
Don't necessary to move AXIS units. Only press End turn button to complete 3-4 turns.

[Lettos]

3) May (or not) conduct all per Unit Class moves not intended to attack prior to carrying out any actual attacks. It's not clear to me yet.

It can't. Tested. Very simple algorithm. AI takes units one by one and moves either with an attack at the end of the movement or without an attack.

5) Always observes the following priorities as they apply to the following three Unit Air Classes: Tactical Bomber (1), Level Bomber (2) and Fighter (3). The movement priority assigned to the Air Transport Class isn't clear to me yet.

Tested. Air transport have a priority of loaded unit class.

[Lettos]

Further empirical "investigation" is certainly warranted here.

Defense and Offense by results in sandbox experiment with Ground and Air units.
Tested in sandbox with 3 VH Cities and 5 usual Cities and one airfield. All Cities are "empty".

Chapter I. "Unlimited PP"
- "Unlimited"(40000) PP assigned on 1st turn. -

AI actions in turn are combined from two independent parts performed in this order:

1st part (it is common for Defense and Offense Scenarios).
Movement and sending replacements to existing units.
AI perform it in priority's sequence:

1. Tac Bombers
2. Level Bombers
3. Fighters
4. Recons
5. Tanks
6. Anti-aircrafts
7. Guns
8. AD
9. Infantry
10. ATG

! Until all higher class units are not fully reinforced not any replacements will be sended to next lower class !


2nd part. Purchasing new units.

Offense

1st turn.
2-1 step. Purchasing around VH Cities. It's mandatory for AI to purchase all mentioned units before go to step 2-2. Purchasing sequence is following:
AI Advanced: INF, INF, GUN (the most expensive). Infantry units set for all VH is mixed (i.e. 1 cheapest, 1 Bridge Eng, 1 HW Inf randomly). Only in very rare cases AI purchased Paratroopers. Unit's costs in experiment: Bridge Engineers = 144, Para = 132, HW Inf = 84, Inf = 48.
AI Intermediate: INF (mixed), GUN (the most expensive).
AI Basic: INF (one cheapest unit).

Note:
Advanced and Intermediate AI deployed GUN in City and INF unit close to city, in one of two hexes on line between City and Enemy VH. AI not moved these units during next turns.
Basic AI deployed INF in one of two hexes on line between City and Enemy VH and on next turn moved INF unit into city.

2-2 step. Purchasing around/and in VH Cities and usual Cities. Tanks.
The most expensive tank will be purchased. For deployment AI will first choose City closest to enemy.
AI Advanced: In all free hexes until all hexes are occupied.
AI Intermediate: 2 tanks.
AI Basic: 1 tank.

AI will repeat Step 2-2 will on all next turns if PP are available.
All purchased tanks on next turn will be moved towards somewhere in Enemy VH direction even enemy units are not in spotting range. No any accumulation of forces. Looks as tactics "Move to enemy. Find. If found - don't think, don't wait for help but kick him!" (by the way this tactic is very similar to red Army tank counter attacks in 1941-42)

Defense

AI purchases in modes:
Advanced and Intermediate: INF units mixed set, AI often purchased Paratroopers (for Defense - sic!), GUN and AD are the most expensive.
Basic: cheapest INF and GUN units only.

1st turn.
2-1 step. Purchasing around VH Cities. It's mandatory for AI to purchase all mentioned units before go to step 2-2. Purchasing sequence is following:
AI Advanced: INF, INF, GUN, AD, INF, AD, GUN (All hexes occupied with full set of 7 units = 3 INF, 2 AD, 2 GUNS).
AI Intermediate: INF, INF, GUN, AD, INF. (3 INF, 1 AD, 1 GUN, 2 free hexes). Very rarely(1 case from 20-30) in one VH from three available: 3 INF, AD, GUN, INF.
AI Basic: INF, INF, GUN.

2-2 step. Purchasing around usual Cities*(See NOTE below!). INF and GUN units:
AI Advanced: INF, INF, GUN. (GUN in City)
AI Intermediate: INF, INF. (Both units placed outside the city)
AI Basic: no purchases.

2-3 step. Purchasing around usual Cities. Tanks.
The cheapest tank only will be purchased.
AI Advanced: in all free hexes
AI Intermediate: 2 tanks (outside City closest to Enemy)
AI Basic: 1 tank (outside City closest to Enemy)
All purchased tanks on next turn will be moved backwards somewhere to own VH City.

* NOTE about quantity of usual Cities where AI will purchase these units.
I've experimented with following sets:
3 VH Cities + 5 Usual Cities = 8
3 VH Cities + 13 UC = 16 and
3 VH Cities + 21 UC = 24.
On Step 2-2 AI will not add INF and GUN units to all empty UC but only to some of them. There is not strongly defined proportion between total city quantity and cities where AI deployed INF and GUN on Step 2-2.
For example, from 8 cities AI deployed units in 6, only in 9 from 16 and in 13 from 24. Other cities in Advanced mode were protected by cheapest tanks only, without any infantry and guns. AI in Intermediate and Basic modes never protected the rest cities by any units. So total percentage of cities protected by INF and GUN units is somewhere between 54-67%.

Chapter II will be dedicated to "Limited PP" and will come soon.

[Lettos]

Chapter II. Limited PP.

Same sandbox DEFENSE 10 turn scenario. AI task is to defend 3 VH Cities + 5 usual cities + 1 airfield. Testing AI actions after 9th turns. Additionally subjective observations included.
-- Info about used unit's costs:
UK 39INF = 48 , HW INF = 84, PARA = 132, Bridge Eng = 144,
25Pdr GUN = 132, 6'' = 180
AD 20mm = 72, 40mm = 180, 3'' = 240
Matilda I = 48 (cheapest tank), Cruiser I = 108. The most expensive tank in Class is Matilda II = 156.

Below are units purchased by AI after 9 turns.

50 PP/turn
ADV: 9 39INF
INT: 9 39INF
BAS: 6 39INF + 3 Matilda I

100 PP/turn
ADV: 4 HW INF, 2 25Pdr GUN, 3 AD 20mm
INT: 8 39INF, 2 25Pdr GUN, 3 AD 20mm
BAS: 6 39INF, 3 25Pdr GUN, 3 Matilda I

200 PP/turn
ADV: 9 HW INF, 3 25Pdr GUN, 9 AD 20mm. 21 from 22 units deployed in 3 VH Cities and 1 in usual City.
INT: 3 Bridge Eng, 6 INF39, 3 6'' GUN, 3 AD 40mm. Bridge Engineers deployed in VH Cities. Each VH City have 5 units as in case of Unlimited PP but comparing with "Unlimited PP" deployment these units are worst deployed.
BAS: 6 39INF, 3 25Pdr GUN, 7 Matilda I

400 PP/turn:
ADV: 8 HW INF, 5 PARA, 1 39INF, 7 6'' GUN, 4 20mm AD, 1 40mm AD, 2 3'' AD.
AI deployed 21 unit in 3 VH Cities till 6th turn. Then started with usual Cities.
PARA units deployed as infantry in usual cities without any airfields. Overall units deployment in "Seventh" is a little strange.
INT: 3 Bridge Eng, 7 39INF, 1 HW, 2 PARA, 3 6'' GUN, 1 20mm AD, 1 40mm AD, 1 3'' AD. 1 Cruiser I, 6 Matilda I.*(See NOTE below)
AI deployed 15 units in 3 VH Cities till 5th turn. Then started with usual Cities.
BAS: 6 39INF, 3 25Pdr GUN, 8 Matilda I.

* NOTE
I was interested to check more times AI Advanced/Intermediate in 400PP/turn scenarios. If AI purchases same set of units or not? Additional 3 test's results are below. Previous testing result added as "0" number:

0 ADV: 1 39INF, 8 HW, 5 PARA, 0 BE, 0 25Pdr, 7 6'', 4 20mm, 1 40mm, 2 3'', 0 MatI, 0 CruiserI
1 ADV: 0 39INF, 8 HW, 3 PARA, 0 BE, 0 25Pdr, 6 6'', 4 20mm, 1 40mm, 2 3'', 0 MatI, 0 CruiserI
2 ADV: 11 39INF, 0 HW, 0 PARA, 3 BE, 1 25Pdr, 5 6'', 4 20mm, 0 40mm, 3 3'', 8 MatI, 0 CruiserI
3 ADV: 11 39INF, 0 HW, 0 PARA, 3 BE, 1 25Pdr, 5 6'', 4 20mm, 0 40mm, 3 3'', 8 MatI, 0 CruiserI

0 INT: 7 39INF, 1 HW, 2 PARA, 3 BE, 0 25Pdr, 3 6'', 1 20mm, 1 40mm, 1 3'', 6 MatI, 1 CruiserI
1 INT: 7 39INF, 1 HW, 2 PARA, 4 BE, 0 25Pdr, 3 6'', 1 20mm, 1 40mm, 1 3'', 6 MatI, 0 CruiserI
2 INT: 2 39INF, 9 HW, 3 PARA, 0 BE, 0 25Pdr, 3 6'', 3 20mm, 0 40mm, 1 3'', 8 MatI, 0 CruiserI
3 INT: 2 39INF, 8 HW, 4 PARA, 0 BE, 0 25Pdr, 3 6'', 3 20mm, 0 40mm, 1 3'', 8 MatI, 0 CruiserI

Average ADVanced: 14 INF, 6-7 GUN, 7 AD, 0-8 tanks(cheapest)
Average INTermediate: 13-14 INF, 3 GUN, 4-5 AD, 6-8 tanks (cheapest, one time one average)
Average BASic: 6 INF, 3 GUN (cheapest), 8 tanks (cheapest)

My short comparison of ADV, INT and BAS modes

BASic AI it's very specific and provide very strange reinforcement set. I'll ignore this mode.
In INT mode AI creates more balanced forces and using with INF and GUNs will try to defend not only VH Cities. This mode is predictable in reinforcement's sets.
In ADV mode AI will much better defend VH Cities, but is providing very poor defense for usual, non VH Cities. It is unbalanced mode in reinforcement sets. But the best AD from three AI modes and the best artillery set.
Scenario Developer have to choose which one he will use

[Lettos]

6) Always prioritizes Submarine Class units over Capital Ship Class units. Priorities involving Destroyer, Air Carrier and Naval Transport Class units aren't clear to me yet.

It's good that there was a big puddle with toy ships next to the sandbox.
Common rule for all except transports, in puddle: Unit with less MVT will be moved first.
Movement priorities:
1. Destroyer class.
2. U-boat class.
3. Capital ships.
4. Air Carriers.
5. Transports. Priority by Ground Class.
Actually, I did not even hope that in this puddle there would be the "Flying Dutchman" from AI.
And it is logical that in the puddle there is still the same tactical principle - "go ahead, hit the first one you find".
Great generals silently sympathize with great admirals.

[Lettos]

My short comparison of ADV, INT and BAS modes

BASic AI it's very specific and provide very strange reinforcement set. I'll ignore this mode.
In INT mode AI creates more balanced forces and using with INF and GUNs will try to defend not only VH Cities. This mode is predictable in reinforcement's sets.
In ADV mode AI will much better defend VH Cities, but is providing very poor defense for usual, non VH Cities. It is unbalanced mode in reinforcement sets. But the best AD from three AI modes and the best artillery set.
Scenario Developer have to choose which one he will use

In all performed tests I've deployed AXIS army 20 INF units 5 hexes far from Allied side's closest city. I've not moved AXIS units at all.
But what changes in AI decisions if AXIS units will attack?
I set PP 500 per turn in 1-5 turns.
Tested. AI can see AXIS INF units close to cities. Nothing. Still same AI algorithm in VH Cities: INF, INF, GUN etc. Only tanks purchased not in usual cities in closest to AXIS but in VH Cities.
Next test. 10 tanks added to AXIS units and attacked Allied usual cities. AI changed nothing in his purchasing algorithm. Only 4 classes purchased: INF, GUN, AD, Tank.

Question is about ATG

On last turns in my OoB France scenario AI purchased few 25mm ATG and deployed in VH Cities. In sandbox 2Pdr Gun = 60PP unit was available for purchasing but AI didn't purchased ATG.
Question to ALL: Could You please share Your experience about AI ATG buys?

[Radoye]

I haven't done any extensive empirical testing like yours, but in my experience yes, AI buys a lot of towed AT guns - under normal campaign play each victory hex gets surrounded by cheap AT, AD and heaviest ATY that AI could afford.

[Lettos]

I haven't done any extensive empirical testing like yours, but in my experience yes, AI buys a lot of towed AT guns - under normal campaign play each victory hex gets surrounded by cheap AT, AD and heaviest ATY that AI could afford.

I'll think about it.
ATY - it is Artillery?

[Radoye]

ATY = short for Artillery, yes

[Lettos]

I haven't done any extensive empirical testing like yours, but in my experience yes, AI buys a lot of towed AT guns - under normal campaign play each victory hex gets surrounded by cheap AT, AD and heaviest ATY that AI could afford.

I'll think about it.

I've created a sandbox scenario, AXIS=GER, AI Allied=UK. AI Advanced. ATG class allowed to purchase. I can do it in H2H mode. But..
I've decreased ATG cost even to =1. AI not purchased ATG. So here is not any AI logic or "thinking", but there is a simple trigger switching on ATG purchasing.
I've tryed following:
1. To occupy 15VH AI Cities and 15 usual cities.
2. To lost 2 my units.
3. To destroy a lot(15-20) of AI units.
4. At start of Scenario to deploy a lot of ATG units and destroy them later.
5. To destroy 15 AI aircrafts.
6. To give to AI limited or unlimited PP.
7. To wait till 30th turn.
8. To leave not occupied only one VH City.
9. Attack with Inf and Tanks.
AI purchased INF, ATY, AD and tanks. No ATG buys.

Conclusion:
1) If You will see a fact of purchasing of ATG by AI please try to fix all (as possible) conditions when it occured. I can't emulate now ATG purchasing in sandbox. May be I need to change AI country to France or Poland ?..
2) If it is so hard to force AI to purchase ATG even in sandbox there is another question: what we are waiting from AI in battles when ATG purchasing is highly necessary? Is AI will do it in Scenario in moment when Scenario Developer "predicted" and planned this AI action by battlefield logic?

[Radoye]

There's no special conditions, AI buys towed AT units regularly. I've seen it more often than not, campaign play, single scenario, standard SSI conversion files or mods...

Here's a screenshot from before Axis Turn 1, PG1 SSI campaign with all files unmodified, Poland scenario:



Here's before Axis Turn 4 (or after Allied Turn 3), same play - and there's a newly purchased Allied ATG next to the top victory hex:



I'm actually surprised that you find this to be a rare thing

[Lettos]

There's no special conditions, AI buys towed AT units regularly. I've seen it more often than not, campaign play, single scenario, standard SSI conversion files or mods...

I'm actually surprised that you find this to be a rare thing

You have no idea how surprised I was that AI does not buy ATG in the sandbox! He should have bought ATG!

I replaced in sandbox AI Allied side GB with France now. AI purchased ATG instead of all INF in my experiment described above. I replaced France with USA. AI purchased ATG and not INF.
I replaced USA with GB. AI purchased ... not ATG but INF.
It became quite interesting.
In the experiment, I've used the equipment file where only one 2Pdr ATG was available for GB in 1939. Number 217 in file.
I've copied 218 6Pdr row to 217. Changed price from 108 to 70 and Production year to "from 38". And no bug more! AI purchased a lot of 2Pdr ATG instead of INF!
Purchasing algorithm is exactly the same as in sandbox experiment described above.
By the way AI also purchased one 6Pdr ATG as more expensive and "better". But in my equipment file parameters for units are same and only price is different... AI is looking only to price.
I really don't know the nature of occurred bug. Looks as my equipment file was corrupted on row 220, unit Nr.217? How many such errors can be hidden in the equipment file?
Ok, now I see ATG!
And thanking to this bug now is clear which units will purchase AI if units from ATG class not available. Next class to "up", INF.

[Radoye]

Yes, the PGF AI has no concept of which unit is "better" according to stats when purchasing, it just assumes that the more expensive is the better one.

[Lettos]

Yes, the PGF AI has no concept of which unit is "better" according to stats when purchasing, it just assumes that the more expensive is the better one.

Thanks for help!

[Lettos]

One "auxiliary" topic.
How PGF Engine will give You PP for destroyed enemy units.
In test participated enemy unit with Cost=2000. 6 EXP levels from 0 to 5.
If enemy unit was "killed" by one hit the gained PP are:
EXP=0 - 83
1 - 166 = 83 * 2
2 - 249 = 83 * 3
3 - 332 = 83 * 4
4 - 415 = 83 * 5
5 - 498 = 83 * 6
So each (experience star +1) is a multiplier for basic "83".
Coefficient or basic multiplier is = 83/2000 = 0.0415

Additionally here is one hint how to gain more PP:
I've tested with same Enemy unit Cost=2000 and EXP=3.
Each hit causing Enemy unit's losses till 50% will give You half of Basic PP. I.e. if I've hit this 2000 worth unit and "killed" 1-5 this unit's "strengths" I'll got 42 PP. Doesn't matter if I've hit Enemy unit with STR=10 or STR=5. Here is only percentage from real Enemy unit's strength at the hit moment.
For hit caused losses more than 50% engine will give to Player the Basic "award" 83 PP.
So if Player is interested to gain a maximum PP he need to attack an Enemy unit 9 times "killing" only one "strength" per attack. In this example:
9 * 42 = 378
And last attack of this enemy unit with EXP=3 and STR=1 will give 83*4 = 332 PP.
Sure it is not possible to attack one unit nine times. But...
"Don't kill them all immediately, try a limited hunt so long as possible!"

[Lettos]

Also waiting to hear from # Lettos # regarding some promising AI experimentation.

I've tested. In two words: "IT WORK".
I've created sandbox AI Allied Defensive scenario. 10 turns.
Two events: 2000 PP on 3rd turn and 2000 PP on 9th turn.
Allied(AI) units deployed (Quantity x STR):
TB 3x10, LB 4x10, Fighters 14x1 3x10 all with EX=0, Recons 5x1, Tanks 17x1, AA 12x1, ATY 13x1 2x10, AD 19x1, INF 5x10 20x1, ATG 10x1
And deployed on airfield "Batman Gen HQ" unit: Fighter class STR=1 EXP=5 Cost=2000 MVT=0 Fuel=300 all attack and defense = 0.
Tested in AI Advanced and Intermediate modes.
1-2nd turns - AI send some PP gained in clashes to cheapest(from STR=1 to 3) tank and to one inf (from STR=1 to 2)
3rd turn - AI send replacements to all possible Air units which were deployed on airfields and 5 strengths to Batman Gen HQ. Some fighters were not reinforced because here is complicated AI algorithm to which unit in Class to send replacements first. I'll investigate this algorithm ASAP.
4-8th turns - AI actions on battlefield.
9th turn - "Batman Gen HQ" STR=10, all possible air units reinforced too, and the rest PP AI spend to replacements for Recons, Tanks, AA and some ATY.

It works even without empty Air transport. But Air transport in role of "Black hole" (if AI wil send replacements to it) will be a little bit predictable due to strange and still not investigated AI replacement's sequence inside one Class.

[Lettos]

Verification time ! Remember, the "theory" pertains to PG1-DOS... Even then, if memory serves me correctly, experimentation revealed that fraction one-fourth should have been one-eighth or vice versa.

Caveat: Due to the different way that PG1-DOS "defines" a unit's listed cost, all PGF "experimental" results may have to be divided by 12 !

Things need to be tested:
Replacement's sequence inside one class.
Amount of gained PP during attack and destroying enemy unit with EXP=0,1,2,4,5.
EXP gained by own unit in case of attack without/with own losses.
The sandbox will be a little dusty
I'll test!

[Lettos]

"PG1 Theory"

Here's what Prima's PG1 Strategy Guide (not to be confused with SSI's manual) has to say on the subject:

The base reward for fighting in combat is the prestige number of all the enemy unit's equipment x (the enemy unit's experience level + 1) and then halved

D-Day scenario, settings by default.
Destroyed bomber's (EXP=3) cost is 396. PP gained = 66. 66/396 = 1/6.
Another bomber's (EXP=3) cost is 420. PP gained = 70. 70/420 = 1/6.
Something wrong in Strategy guide...

[Lettos]

The base reward for fighting in combat is the prestige number of all the enemy unit's equipment x (the enemy unit's experience level + 1) and then halved.

Caveat: Due to the different way that PG1-DOS "defines" a unit's listed cost, all PGF calculation results should be divided by 12 !

Let's take your first example:

Destroyed bomber's (EXP=3) cost is 396. PP gained = 66. 66/396 = 1/6.

PP GAIN = ((396*(3+1))/2)/12) = ((396*4)/2)/12) = ((1584/2)/12) = (792/12) = 66

Bang on !

Let's take your second example:

Another bomber's (EXP=3) cost is 420. PP gained = 70. 70/420 = 1/6.

PP GAIN = ((420*(3+1))/2)/12) = ((420*4)/2)/12) = ((1680/2)/12) = (840/12) = 70

Bang on again !

So far, PG1 "theory" works like a charm !

Hexcode,
I apologize for one small inaccuracy that caused big consequences. Please understand that I did not do this on purpose.
I wrote:

D-Day scenario, settings by default.
Destroyed bomber's (EXP=3) cost is 396. PP gained = 66. 66/396 = 1/6.
Another bomber's (EXP=3) cost is 420. PP gained = 70. 70/420 = 1/6.

Should be:
D-Day scenario, PG1, settings by default.
Destroyed bomber's (EXP=3) cost is 396. PP gained = 66. 66/396 = 1/6.
Another bomber's (EXP=3) cost is 420. PP gained = 70. 70/420 = 1/6.
So taking into account your calculations, the Caveat

Caveat: Due to the different way that PG1-DOS "defines" a unit's listed cost, all PGF calculation results should be divided by 12 !

should look as:

Caveat: PG1-DOS calculation results should be divided by 12 !

[Lettos]

My Findings

1) The "Enemy Unit Elimination Prestige Award" (EUEPA) equals:

([Enemy Unit Listed Cost] / 24) * (1 + [Enemy Unit Experience Level])

The unit cost includes the cost of its companion Transport Class unit, if any.

Yes! It is correct!

2) The "Enemy Unit Relative Strength Reduction Prestige Award" (EURSRPA) equals:

one quarter (1/4) of EUEPA.

Partially correct. There is a nuance here:
If Enemy unit's Relative Strength Reduction (EURSR) is 70% or more -> EURSRPA = 1/4
If EURSR is less than 50% -> EURSRPA = 1/8
If EURSR is equal or more than 50% and less than 70% -> EURSRPA is distributed unevenly randomly and can be 1/8 or 1/4.

Examples from sandbox battlefield and illustrating it:

Strength reduced From-To -> EURSPA:
10-5 -> 1/8 always (or very often only?)
10-6 -> 1/8 (more often), 1/4 (less often). Probably, distribution of random is something like 1/8 = 66% and 1/4 = 33%.
8-4 -> 1/8
4-2 -> 1/4 (occurred one time only. Not 1/8 but 1/4 !)
7-3 -> 1/4 (!)

3) The "Enemy Unit Retreat Prestige Award" (EURPA) also equals:

one quarter (1/4) of EUEPA.

In the case of a single combat outcome, prestige awards are never combined ! PGF's engine just chooses the most beneficial prestige award type applicable to the situation.

I've not investigated EURPA. Probably will do it today.

[Lettos]

One error corrected:

Enemy unit's Strength reduced From-To -> EURSPA:
10-5 -> 1/8 always (or very often only?)
10- 4(four - it's my typo!)6! -> 1/8 (more often), 1/4 (less often). Probably, distribution of random is something like 1/8 = 66% and 1/4 = 33%.
8-4 -> 1/8
4-2 -> 1/4 (occurred one time only. Not 1/8 but 1/4 !)
7-3 -> 1/4 (!)

[Lettos]

Genesis

The subject recently came up under this topic.

"PG1 Theory"

Here's what Prima's PG1 Strategy Guide (not to be confused with SSI's manual) has to say on the subject:

A unit gains experience points whenever it takes part in (and survives) a battle with another enemy unit. Even bombarding artillery and capital ship units improve their experience with every long-range shot they lob, but they do so more slowly than when units engage in combat directly. The formula for the number of experience points a unit gains in battle is given below. The better a unit performs in battle, the more experience points it gains.

To determine the amount of experience a unit gains from a given engagement, the program goes through a number of steps. We present the final equation first and then show you how each variable is determined:

Base experience points earned x present experience fractional multiplier = final experience points earned.

Units accumulate base experience points as follows:

15 per attack for level bombers

2 per friendly strength factor lost (the survivors wise up quickly!) plus any kills experience (for killing enemy strength factors)

The key to the formula abbreviations used in the table is shown after the formula.

Step 1: Calculate (6 + enemy Astrength - friendly Dstrength x (enemy on-line / 10)). Note: This value is 0 when you're conducting a ranged attack).

Step 2: Add (6 + enemy Dstrength - friendly Astrength) to the value of Step 1.

Step 3: Adjust the total obtained in Step 2 so that it is at least 1 (if it was less than 1) or no greater than 15 (if it was more than 15).

Step 4: Multiply the adjusted number from Step 3 by the number of enemy strength factors killed (not suppressed!). The result is the base experience points earned by the unit in that battle.

Astrength is the unit's attack value vs. the appropriate enemy target type.

Dstrength is the unit's appropriate defense value vs. the enemy unit type (if this value is 0 treat it as 1 for this formula's purposes).

On-line are strength factors actually firing (that is excluding strength factors that were suppressed, unsupplied, reduced for mud, etc.)

Ranged attack is an attack conducted by any unit with a firing range greater than 0, such as artillery, air-defense and capital ship class units.

The base experience points earned are then multiplied by the unit's "Ability to Learn" fraction specified in the table below:

Code: Select all

Unit's              Present  Ability to Learn
Experience Level    Fractional Multiplier

0 or 1              2/3
2                   1/2
3                   2/5
4                   1/3
5                   1/4

The more experienced a unit becomes the greater difficulty that unit has in gaining additional experience. In other words, it is very tough to confront veteran troops with situations they have not encountered before. This change is reflected in the game by decreasing the fraction of the experience point multiplier as experience levels rise. As such, it is also much less likely that they will learn something new if they reach the elite, 5-star level.

The final result is the actual number of experience points, rounded up, that the unit has earned in that battle.

Next Steps

Well ?

In first sandbox experiment participated Wermacht39 INF SA=6 GD=7 and AI unit Poland INF SA=5 GD=4. Both units with STR=10, EXP=0, ENT=0. Wermacht attacked.

After attack:
Wer39 INF STR=8, Poland INF STR=4.
Step 1 + Step 2 = (6+5-7*(10/10)) + (6+4-6)) = 8 for all examples below. Calculating:
(2*2 + 8*6)*2/3=35. EXP gained in sandbox is 34.

Wer39 INF STR=10, Poland INF STR=2.
(2*0 + 8*8)*2/3 = 44. Sandbox = 42.

Wer39 INF STR=4, Poland INF STR=5.
(2*6 + 8*5)*2/3 = 34.8 Sandbox = 35.

Wer39 INF STR=10, Poland INF STR=7.
(2*0 + 8*3)*2/3 = 16. Sandbox = 16.

Wer39 INF STR=4, Poland INF STR=10.
(2*6+ 8*0)*2/3 = 8. Sandbox = 8.

Wer39 INF STR=6, Poland INF STR=9.
(2*4 + 8*1)*2/3 = 10.7 . Sandbox = 10.

Wer39 INF STR=7, Poland INF STR=7.
(2*3+8*3)*2/3 = 20. Sandbox = 20.

So is it time to close a book and sandbox too? Probably. But... next experiment with same units, terrain, STR. Only Poland INF ENT increased instead of ENT=0:
Same battle as in last example from previous experiment, Wer39 INF STR=7, Poland INF STR=7. EXP calculated = 20.
ENT=2. Sandbox result is EXP=22.
ENT=4. Sandbox EXP=24.
ENT=6. Sandbox EXP=26.
ENT=9. Sandbox result is EXP=28.

Looks as ENT is simply added to EXP calculated by PG1 formula.

And what about EXP "Levels" and "Ability to Learn"?
Experiment with same units, terrain, STR. Only Wer39 INF EXP increased instead of EXP=0.
a) EXP=1.
Wer39 INF STR=10, Poland INF STR=4. Calculated EXP should be 32. But... only 24 in sandbox.
Wer39 INF STR=10, Poland INF STR=5. Calculated EXP should be 26. Sandbox EXP = 20.
Wer39 INF STR=7, Poland INF STR=7. 16 instead of 20.
Multiplier is REDUCED from 2/3 to 1/2.
b) EXP=2.
Wer39 INF STR=10, Poland INF STR=4. Calculated EXP should be 32. But... only 12 in sandbox.
Wer39 INF STR=10, Poland INF STR=5. Calculated EXP should be 26. Sandbox EXP=10.
Multiplier is REDUCED from 1/2 to 1/4.
c) EXP=3.
Wer39 INF STR=10, Poland INF STR=4. Calculated EXP should be 32. But... only 4 in sandbox.
Wer39 INF STR=10, Poland INF STR=5. Calculated EXP should be 26. Sandbox EXP=4.
Wer39 INF STR=6, Poland INF STR=3. Calculated EXP should be 26. But only 8 in sandbox.
Multiplier is REDUCED to 2/5 to 1/8.
d) EXP=4. Multiplier is 1/12. Tested and approved.
e) EXP=5. 1/16. Tested and approved.

Only integer part (i.e. digits after comma will be ignored) of calculated amount will be added as EXP points.

Hexcode, books + testing + thinking = a good result! Thanks!

[Lettos]

Strange Event: Prestige was awarded in an air combat situation where both the friendly and enemy air units suffered precisely the same number of casualties as a result of their combat.

Tested now.
If 8=8 in combat between Air units, or 7=7 (AirU), 6=6 (GroundU) and 5=5 (AirU) Prestige award in tests was 0.
But: 4=4 (three times!) and 2=4 (Sic! Enemy casualities are LESS than our!) Prestige award was 1/4.
I'm supposing:
1) here is randomization for 5=5 result.
2) And rule: If enemy loses are more than 50-60% calculate +1/4 from EUPA in all cases.

[Lettos]

I've investigated:
How AI distributes available amount of PP for replacements inside one Class for only one Unit type with EXP from 0 to 5.
Only empirical rules in conclusion:

1st step. AI will send about 75% of available PP amount to replacements for units with EXP=0. Doesn't matter in one turn or in few turns. The rest 25% of PP amount will be send to EXP=1. And only If 40-50% of EXP=0 units has "replenished" to STR=10, AI will go to 2nd step

2nd step. AI will send about 75% of PP amount to EXP=1 until replenished units quantity will achieve 40-50%. The rest 25% of PP amount will be randomly distributed between EXP=0 and EXP=2.

If on this Step AI haven't enough amount of PP to send replacements to ALL his units AI will start randomly increase EXP level of only some units with EXP>0. In final of experiment I even saw units with STR=14. This is a randomized process how AI decide which EXP Level will get more "overstrength" each turn.

3rd step. AI will increase EXP=3 units until get 40-50% of them replenished. About 75% to EXP=3 and only 25% will be randomly distributed between EXP=0,1,2.

4th step. The same as 2nd and 3rd but for EXP=4,5. After finished this first stage of process AI will have 60-70% of EXP=0,1,2,3 units with STR=10, and 50-60% of EXP=4,5 units with STR=10.

5th step. About 60-75% of available PP amount AI will spend to replacements for EXP=0,1 units. And the rest PP - to replacements for EXP=2,3,4,5.

6th step. AI will try to get all units EXP=1,2 with STR=10. And only after it AI will send necessary replacements to have all units EXP=0,3,4,5 with STR=10.

[Lettos]

I've investigated:
How AI distributes available amount of PP for replacements inside one Class for only five Unit types with different PP cost(conditionally Cost from 1 to 5), with EXP=0.
Only empirical rules in conclusion:

1st step. AI will send all available PP amount to cheapest type units (Cost 1). When AI finished it will go to Step 2.

2nd step. All rest unit types will be replenished randomly. It mean:
Experiment N1. Cost 2 and 4 ignored during two turns, but all replacements send to Cost 3 and 5. On 3rd turn Cost 2 and Cost 4 fully replenished too.
Experiment N2. Only Cost 4 ignored during two turns.

Paying attention to: viewtopic.php?f=95&p=7953#p7953
for now there is at least two empirical rules:
1. Cheaper and weaker unit have much more chances than more expensive and more experienced units to have replacements in the first turns when Prestige Points became available for AI. But at least 40% of cheapest and weakest units will get a replacements first of all.
2. Neither more experienced unit's experience level neither more expensive unit's cost not provide a warranty from PGF engine's side to certain unit to have replacements earlier than other units. There is a Probability only.

If someone will investigate the replacement process more deep please share the results to edit and correct these rules!

[Lettos]

Two key questions:

1) Does PGF's AI provide its units with

a) Elite Replacements only ?
b) Normal Replacements only ?
c) Either Elite or Normal Replacements, depending on the specific situation ?

2) Is PGF's AI programmed to spend all of its available prestige in every single turn ? If not, is PGF's AI allowed to carry over unspent prestige to subsequent turns ?

Answering:
1) c) only has been observed.
If Unit's EXP=0 AI will provide Normal Replacements. If EXP Level 1 star or more - only Elite Replacements. AI programmed to wait for PP and not thinking about "is it better to provide Normal Replacements now instead of waiting for future quality's profit or not?"

2) AI've spend ALL his PP in one turn in all tests. It was greatly observed when AI send all PP to Air units and remained "some cents change" PP which AI've spend for replacements for 1 Poland TK3 to increase STR from 1 to 3. It not looks as collecting something for future.

There are not AI thinking at all. Only formulas and simple algorithms such as: "Always keep movement straight on green light"

[Lettos]

Two key questions:<...>

There is xls file with Experiment N1 results.
How AI distributes available amount of PP for replacements inside one Class for only one Unit type with EXP from 0 to 5.
Link: https://drive.google.com/file/d/1aV-BQ3 ... sp=sharing

6 groups of Poland INF per 20 units in group. Defense scenario. Only 4 Wer39 INF from AXIS side. All Poland INF not moved by AI till 10th turn.
EXP in groups = 0,1,2,3,4,5,6.
Each turn AI PP settings = 1000 or something close, may be 600. I don't remember now.
Column A = EXP Level.
Row 1 = Turn
Digits in cells = Total Q-ty of units with STR=10. Digits in parentheses () : how much units on this turn are with EXP Stars, and **** mean the Star's Q-ty.

[Lettos]

If Unit's EXP=0 AI will provide Normal Replacements. If EXP Level 1 star or more - only Elite Replacements.

This is interesting. It reminds me of an often used, restrictive "house rule" in longish scenarios. The idea here is for the human player never to give Elite Replacements to zero-star units. But if, per chance, a unit crosses the 99 experience points threshold due to a particularly successful combat and achieves 1-star status, the human player can do as he pleases as long as the unit doesn't revert back to zero-star status.

I have not explored such subtleties. The sandbox was already very dusty

... is PGF's AI allowed to carry over unspent prestige to subsequent turns ?

To definitively answer this question, one needs to know where exactly PGF's engine saves the AI's prestige in the game-state file. I may be able to do something about that very soon.

All I can do is wish you luck friend Hexcode!

[Lettos]

Since both tests were aimed at figuring out the correct location for the "Black Hole"(or Batman/Pumpkin General's HQ) for Event creation, I've performed conclusion test.
20 Air units, Pumpkin as Recon MVT=0 and EXP=0.
Test results:
1) Air units will got a full replacements from STR=1 to STR=10.
2) Pumpkin will be replenished simultaniosly with another Recons with EXP=0.
3) After it Replacements will be send to another Recons and "lower" unit classes starting from tanks.
Conclusion again: IT WORK.

So empirical rule:
if Scenario Developer will need to create a "Black Hole" to separate replacements send to "upper" classes from send to lower classes, he should prefer to create the unit with EXP=0 (and MVT=0, Cost = unlimited) from next lower class.

[Lettos]

Referring to viewtopic.php?f=95&t=470#p7936

So is it time to close a book and sandbox too? Probably. But... next experiment with same units, terrain, STR. Only Poland INF ENT increased instead of ENT=0:
Same battle as in last example from previous experiment, Wer39 INF STR=7, Poland INF STR=7. EXP calculated = 20.
ENT=2. Sandbox result is EXP=22.
ENT=4. Sandbox EXP=24.
ENT=6. Sandbox EXP=26.
ENT=9. Sandbox result is EXP=28.

Looks as ENT is simply added to EXP calculated by PG1 formula.

NOTE: If an Enemy entrenched unit casualties after attack are 0(zero) Enemy ENT will not be added to Friendly unit's gained experience.
1st Example: Wer39 INF EXP=0 STR after attack is 5, Poland INF STR=9 ENT=6 (before attack)
EXP gained = (2*5+8*1)*2/3 + 6 = 16 (Sandbox result is the same)
2nd Example: Wer39 INF EXP=0 STR after attack is 5, Poland INF still have STR=10, and ENT=6 before attack.
EXP gained in sandbox is "6". It mean : (2*5+8*0)*2/3 + "0".

[Lettos]

Am I crazy here ? My own extensive experimentation indicates that, in the case of PGF's AI, its prestige selection slider is a DUD ! No effect whatsoever. Totally useless...

Is it possible that this 10-year old "hobby" never realized the above "finding" ? I sincerely hope I'm wrong here !! I mean, wow...

Yes, exactly! Prestige slider affects nothing! Tested now on all three AI settings including combination with negative or positive EXP level.

[Lettos]

The other day, Mr. AI himself came to the sandbox. I have long wanted to ask him a couple of questions about the choice of strategy in his offensive. He agreed to a short interview. AI is difficult to communicate with us, see details here: http://forum.open-general.com/viewtopi ... =95&t=148 . It has already been noted that in most cases the AI reaction is expressed by the phrase: "hey, I didn't understand a word here" !!
Therefore the experimental battles in the sandbox served as the language of communication, which I tried to translate from the AI language into a more understandable one.
- Mr. AI, you are offered a "Offensive" scenario in which you will be the attacking side. Axis will move first. 8 turns. There are no prestige points for both sides, i.e. "0" at the beginning and "0" every turn. Your army consists only of Poland INF EXP=0. All Axis Cities are Victory Hexes.
(In all experiments Axis have only Wer39 INF units EXP=0. FW189A with spotting=35 is placed only to observe battlefield. During Scenario Axis units not moved at all)

Experiment N1.
Turn 1.

AI actions on Turn 4.

- Mr. AI, you do not have time to capture all seven cities. The only chance to win this scenario is to split your 28 units into 7 groups of 4 units and will attack simultaneously in seven directions. Then the scenario is won on turn 7. Do you understand what the number of "Turn" is?
- Not.
- Will you change your tactics if you can see the whole map?

Experiment N2. (Spotting for Poland INF = 35)
Same AI tactic as in Experiment 1.
- Do you care if you see the map or not?
- Yes. The main thing is tactics!
- Can we check it out?

Experiment N3. (All AXIS INF unit were removed from cities. So AI need to attack an "empty" cities).
Turn 4.

- Do you care if there is an enemy on the map or not?
- Yes. The enemy can appear at any moment. And I have a tactic for this case!
- Can your tactics change depending on the starting position of the units?
- Да, безусловно!

Experiment N4.
Turn 1.

Turn 5.

- I can see that the deploying of your units as "hexagon" has helped you.
- Yes of course! This is a great units deployment!
- But why are you still spreading power so unevenly? Is this the main secret of your tactics?
- Yes. This great tactic was used by Epaminodas at the Battle of Leucta.
- Do You mean Epaminondas in Battle of Leuctra?
- Not. Epaminodas and Leucta. That's what my creators told me.
- Do you know any other tactics?
- Not. Tactics of Epaminodas in Battle of Leucta is the best.
- How do you plan on dividing your forces into major and minor areas?

Experiment N5.
Turn 1.

Turn 4.

- You are not planning anything at all?
- I don't understand the concept of "planning" anything...
- Thanks for the informative conversation, Mr.AI!

[Lettos]

# Lettos #, I'm really pleased that you came across the "sandbox" play-testing mode where neither side is being given any planned prestige whatsoever. Basically, players will have to get whatever prestige they can by taking cities / airfields / ports away from their opponent and by successful combat engagements.

In the creation of the historic Campaign, three things hindered me a lot:
1) incomprehensible algorithm for calculating PP for each move for AI in PG1,
2) bonuses for captured cities and airfields
3) accrual of PP for partial or complete destruction of an enemy unit.

I was only satisfied with:
4) the option of calculating PP between scenarios.
In fact, this is the only option and is the ability to model the economy of a belligerent state in PG1 / PGF.
That is, 1-3) is "salty-sweet", and 4) is "soft-hard"
The PP PG1 legacy is a mixture of these two completely different categories.

Now I am very glad that I can at PGF:
Completely control 1)
Remove completely 2)
Partially mitigate the consequences of 3) with the help of the "Black Hole" (for AI) and compensate for the imbalance from getting PP for the destruction of this expensive unit, as well as ordinary AI units, with 4).

This is a real step forward on my path to OoB Campaign!

As far as H2H scenarios are concerned, your approach is perfectly correct from my point of view. First, create a scenario that is as deterministic as possible in obtaining PP, and gradually complicate it by adding "sweet-salty" bonuses.

[Lettos]

A) Tested in sandbox: AI will first move tank units with less MVT. Very simple. Very intellectual. Very "unpredictable".
If there is not possible to move these slow units at the beginning of Tank units movement stage AI will not move them during current turn. Even if after faster tanks were moved and necessary hexes has been freed.
IMO: probably Tank Bridge units should have MVT 3 or 4.

B) Nerdyoldman already created Bridge tank in his mode viewtopic.php?f=95&t=174#p4746
"The bridgelayer is good in some scenarios, but SO slow."
It is now clear that the fast tank will not always be the first on the river bank. For example, if the starting position is 7 hexes from the river and the Tank Bridge has MVT=7, then yes, it will come running first. If the group of tank units is located three hexes far from the river, then Valentine with MVT=3 will be the first in right place.}

Some thoughts on Bridging modules in PGF.
The engineering battalions in WW2 consisted of two parts-units completely different in their functions.
- Some Companies (approximately 2-3 per battalion) are specialized troops (infantry), which could build friendly structures and, if necessary, effectively destroy enemy ones. And these companies, among other things, helped to build serious long bridges with a large carrying capacity. These companies are "Bridge Engineers" infantry, with Ignore Entrenchment option, with Bridging Rivers option, with good Hard / Soft Attacks. Should be called as Combat Engineers.
- Another Company (1 per battalion) is actually a pontoon transport column. Which brings pontoons to the river for the construction of the bridge and participates in this construction. This company has zero attack indicators, it does not participate in battles at all as a combat unit. Such a unit has much more rights to be called "Bridge Engineers" but this name already is in use.

Historically, the legacy of "mix soft and salty" has come from the old PG1. One well-known "generalizing" unit "Bridge Engineers" was made. By the way, in the Wehrmacht the battalions of engineers were separated from the divisions into separate units only in 1943. But for the game, such approximations are normal. Apparently, the creators of the game understood this, and also made it possible for each "normal" unit to cross the rivers. If the Scale is unit=division, then it's completely normal that it has a couple of companies of engineering specialization (with rubber boats, pontoons, etc.).

Taking into account the fact that the AI ​​is not able, in principle, to plan the actions of bridge infantry engineers, but by chance it can put forward a tank with MVT<4 first on the river, a small idea arose:

Make a unit named, for example, "Bridge Battalion".
- Unit class = 1 (Tank) or 2 (Recon). Class "2" guarantees that this AI will push this unit forward to some river before it moves any of its tanks. The MVT should be smaller than the other Recons, giving a small guarantee that this unit will be pushed forward before the other Recons.
- So that the unit is not destroyed immediately, but has the opportunity to complete its task, assign it a sufficiently large Defense values. (In real combat life, this can mean something like the fact that a unit cleverly disguises itself, goes out onto the river on a dark night and begins its work under the protection of artillery fire).
- Unit cost - high enough so that the AI ​​does not buy these units.
- Unit icon. It requires that such unit be easily recognized by the player who saw it for the first time. Those icons of tanks with a segment of a bridge instead of a tank tower, all kinds of truck cranes, bulldozers and other engineering equipment, although they graphically display something "engineering", but create logical chaos. Since it is not clear what kind of a bulldozer or crane it is and what it is for. I suggest another solution - cut the "pontoon" image from the Bridge Engineers icon and place it in the any icon at the bottom of the unit (just partly closing wheels or trucks) that we want to make the "Bridging". I did this with two icons: Wehrmacht Sdkfz.7 and Soviet ZIS-6. It looks very good and it is perfectly clear what kind of military unit it is.

It remains to test these wunderwaffes in Scenarios

[Cat Leon]

Experiment N3. (All AXIS INF unit were removed from cities. So AI need to attack an "empty" cities).
Turn 4.

- Do you care if there is an enemy on the map or not?
- Yes. The enemy can appear at any moment. And I have a tactic for this case!
- Can your tactics change depending on the starting position of the units?
- Да, безусловно!

Да, безусловно!

Что это было?

[Lettos]

Да, безусловно!

Что это было?

This was a description of another experiment
Instead of a phrase in Russian, it should have been "Yes, definitely!"
Apparently, no one has read this description of the experiment, except you. Or something else ...
My English is not average but generally bad. I use google translate to write my posts. I edit as much as I can. But there are often gross mistakes. Even copy-paste errors ... I'm sorry for such cases. And I'm sincerely grateful to my friends here for their high ethics and patience with my linguistically illiterate posts.

[Radoye]

Apparently, no one has read this description of the experiment, except you. Or something else ...

Maybe some people understand basic Russian.

[Lettos]

Apparently, no one has read this description of the experiment, except you. Or something else ...

Maybe some people understand basic Russian.

Thank you for understanding both me and the Russian language, Radoye!

[Lettos]

Yeah, the following is a technical matter.

I use google translate to write my posts. I edit as much as I can. But there are often gross mistakes.

You might want to try

DeepL.com

Granted, I've never tested this translator with Russian, but when it comes to French, German and Italian, the quality of its output is head and shoulders above Google Translate.

Now I tried it. According to the first test this translator looks much more clever
Somehow, without concentrating and without making much effort, it solved one small but important technical problem
Thank you very much for your help!

[Parabellum]

...
You might want to try

DeepL.com
...

Even though none of you probably expected it, but I also read the PGF forum regularly. I'm just a generally interested player and don't understand most of the technical descriptions of HexCode, but I find the tip about this translator great. Thank you!

[Radoye]

If it was possible to have minefields as AD class with AA = 0, then they would only passively "attack" ground units.

But they would still retreat.

[Lettos]

By the way, shouldn't LM units not be allowed to receive replacements ?

Not always.
Most often the minefields were set in advance and then no one touched them, but there were also improvised overnight, and there were prepared but not activated.

An excellent book on the work of the miners (in Russian)
"Beware of mines!" by Nemchinsky Alexander Borisovich.
http://militera.lib.ru/memo/russian/nem ... index.html

I have selected from Сhapter 4 dedicated to Kursk Bulge some facts and quotations with exact numerical data.
The text volume is large, please use automatic translation RU-EN.

Глава четвертая. Минный пояс Курской дуги

***
Вновь в неведомой для нас обстановке пришлось двинуться на восток. У железной дороги Волчанск — Белгород рота наткнулась на опередившие нас фашистские танки. Темнело. Поставили в грязь оставшиеся на повозках мины: наверняка клюнет какой-нибудь немецкий танк...

***
Весь апрель и май каждую ночь ставили мины. В полосе 67-й и 52-й гвардейских дивизий наша рота установила около пяти тысяч мин.

***
Минные поля устанавливались из противотанковых деревянных мин ТМД-Б, ЯМ-5 и трофейных металлических ТМи. Позже ставились противопехотные мины, которые прикрывали установленные противотанковые минные поля.

Количество устанавливаемых мин зависело больше от поведения противника, чем от нас. Иногда рота за ночь устанавливала до 250 мин, но чаще только 80–150. Было немало ночей, когда противник вообще не давал возможности минировать.

***
Минные поля, фугасы и подготовленные к взрыву мосты включались в общую глубоко эшелонированную систему заграждений, прикрывавшую оборонительные рубежи и отсечные позиции. Они опоясывали и батальонные районы обороны, и противотанковые районы. Сочетание артиллерийского огня и минновзрывных заграждений являлось основой устойчивости обороны.

И так по всей дуге. Но плотности минирования создавались разные и зависели от важности направления. Если в среднем по всей дуге плотность минирования оценивалась в 1000 противотанковых и столько же противопехотных мин на один километр, то в полосе 6-й гвардейской армии она составляла в среднем 1500–1600 противотанковых и столько же противопехотных мин.

Если учесть еще то обстоятельство, что на южном фасе Курской дуги имелось много глубоких оврагов и обрывистых склонов, которые эскарпировались и контрэскарпировались, то фронт танкопроходимых участков резко сокращался. Таким образом, на основных танкоопасных направлениях плотность достигала трех-четырех тысяч противотанковых мин на один километр фронта.

В глубине обороны готовилась сложная система минновзрывных заграждений. В целях безопасности для своих войск минные поля обносились проволокой и обозначались табличками с лаконичной надписью: «Мины». Большая же часть подготовленных минных заграждений в глубине должна была приводиться в боевую готовность во вторую и третью очередь. Для них отрывались и временно маскировались лунки, составлялись формуляры, а мины и взрыватели складировались рядом. Чтобы привести в полную боевую готовность такое минное поле, имеющее по фронту 60–100 метров, одному отделению минеров требовалось около часа.

***
Условия для установки мин были особенно сложны из-за активизации взаимных действий разведывательных подразделений и большой насыщенности минновзрывными заграждениями полосы между передними краями обороны противостоящих армий.

И все же задание командования мы выполнили в срок. 1500 мин были установлены ротой за девять-десять ночей.

***
В первых числах июля роты батальона все еще продолжали минировать, но теперь уже в глубине обороны. Устанавливались осколочно-заградительные мины ОЗМ-152 с использованием артснарядов калибра 152 мм, взрываемые электрическим способом или по радио, а также немецкие трофейные снаряды, приспосабливаемые как мины натяжного действия.

***
В узких местах шоссе, где затруднен объезд, мы заранее подготовили тупиковые штольни, а в них заложили взрывчатку. В местах возможных объездов снарядили взрывателями группы противотанковых мин.

Всего батальон установил на шоссе 37 мин замедленного действия (МЗД). В качестве замыкателей использовались тогда электрохимические и часовые 16-суточные минные взрыватели. 16 МЗД были установлены нашей 2-й ротой. На каждую такую мину приходилось в среднем 5 зарядов, то есть по 25 килограммов взрывчатых веществ. Общий же вес тринитротолуола каждой группы мин составлял 100–125 килограммов. Сила взрыва такого заряда весьма значительна...

Наиболее сложной и ответственной задачей являлась установка времени замедления взрыва. Ведь если боевая обстановка изменится и наши войска будут на шоссе в период истечения сроков срабатывания взрывателей, то они понесут потери от своих же мин.

Наконец мы получили приказ, определявший сроки замедления в 6–8 суток. В электрохимические замыкатели мы тут же вставили сопротивления, а будильники установили на соответствующие деления. Пустили часы. Химическая реакция началась. Мины замедленного действия были приведены в боевую готовность...

Недалеко от Яковлево расположился наблюдательный пункт роты 211-го батальона спецминирования, входившего в состав нашей бригады. Командовал батальоном капитан Николай Хоменко. Отсюда отлично просматривалось широкое поле, заросшее сорняком. Это поле было минным. И стояли здесь не обычные мины, которыми минировал наш батальон, а специальные, управляемые на расстоянии по радио. Под толстым слоем земли были заложены фугасы по 30–50 килограммов взрывчатых веществ. Замаскированные у фугасов приборы принимали только один зашифрованный радиосигнал, служивший командой к срабатыванию электродетонаторов и взрыву группы фугасов.

Пока мы с Жигаловым собирали своих людей с объектов МЗД, на это мирное с виду поле уже выползали из оврагов крупные танки. Это шли 60-тонные «тигры» с тяжелыми надульниками на длинных стволах 88-миллиметровых пушек. Вот танки подошли к невидимой границе минного поля, вздрагивая от ударов снарядов нашей артиллерии об их 100-миллиметровую лобовую броню.

Зловещий гул моторов усилился...

По ротной рации была доложена обстановка: три пятерки. Это означало, что к замаскированным фугасам подошли немецкие танки. В тот же миг мощная радиостанция, находившаяся в глубине обороны, послала в эфир условный сигнал...

Несколько сильных взрывов подняли на воздух непаханую землю. Один «тигр» запылал, у другого, словно порывом ветра, сбросило башню, а третий, несколько раз дернувшись, застыл на месте. Всего было уничтожено семь танков. Остальные остановились, а затем попятились в овраг.

Левее тоже были видны танки с белыми крестами. В грохот сражения вплелись звуки неистового лая собак. Да, собак... Они неслись навстречу танкам, будто соревнуясь в скорости с бронебойными снарядами, летевшими над их головами. Это были четвероногие помощники минеров 27-го батальона, которым командовал капитан Шишов. Специально натренированные псы с подрывными вьюками, закрепленными на спине, стремглав бросались под днище танков. Штыри, прикрепленные к заряду, отгибались, и срабатывал освобожденный от предохранительной чеки взрыватель. Заряд тринитротолуола пробивал днище, имевшее наиболее тонкую броню. Экипаж вражеского танка погибал вместе с взорвавшей его собакой.

Эффект применения специально подготовленных собак превзошел все ожидания. 27-й батальон выпустил 17 натренированных собак. 2 были убиты по пути, а 15 достигли цели. 11 вражеских танков подорвали четвероногие друзья минеров...

***
... на следующий день под носом у гитлеровских танкистов поставили более 100 мин.

[Lettos]

D) LMUTs should (or not) be allowed to initiate attacks against enemy units occupying adjacent hexes.

It's depending on the selected scale.

Minefields were usually part of a comprehensive defense.
In the case of an anti-tank minefield there was hidden anti-tank artillery somewhere not far behind it (about 1 kilometer away). The weight of explosives in an anti-tank mine was only enough to destroy light armored vehicles, while medium and heavy tanks damaged only the tracks and undercarriage and became stationary and very easy targets, which the artillery successfully shot in a couple of minutes.
Or the tanks had to move through the narrow, mined passageway in columns, which also made them very vulnerable.
The anti-personnel minefield was covered by machine gun positions. The infantry in the narrow, mined tunnels were also vulnerable.

None of this can be displayed in the game on a scale of 1 hex = 10 km unless the minefield is capable of attack. That is, it is simply as some sort of symbolic problem designation for attackers: "Clear this important minefield first, then attack the enemy." But you don't have to make all minefields on the game map like this.

If the scale is 2-5 km, then it's the scenario designer's free choice whether to attack the minefield or not. I'm more inclined to say that minefields should not attack at this scale.

But if the scale of 1 hex = 100 meters - 1 km, then you can display machine gun positions and anti-tank batteries. Here the minefield should lie quietly in the ground/snow, and the units should attack targets with fire through the minefield.

[Lettos]

# Lettos # has already proposed:

Unit Class = 1 (Tank) + Organic Transport: Destroyer or Capital ship with Range = 0, SA=0, HA=0. Ship is much better than Submarine for H2H and still great for H-vs-AI mode, because the minefield will never be located in the port. Target Type = 3 (Naval) MVT = 0

I'm a bit mystified by the Destroyer / Capital Ship Unit Class "Anchor". Why is it needed ?

Most likely he was thinking about the problem of submarines in ports.
Or maybe he wasn't thinking at all.
Or maybe he was thinking too much.
Both processes lead to the same result - he wrote some illogical crap.

[Lettos]

It may be a good idea to limit the envisaged, serious NA values to a few, highly specialized, demining units. This may be particularly important in scenarios featuring quite a bit of naval activity.

"in scenarios featuring quite a bit of naval activity" - Yes. The fewer ships on the map, the easier it is to work with minefields. I remember back in my childhood we had a joke about something unreal: "A submarine in the steppes of Ukraine". The time has come for this submarine

Do potential Level Bomber unit attacks against enemy minefield units require some "convenient" interpretation ?

Generally I want to make all Level Bombers NA=1. The only exception is the Do-217 with the FX-1400 radio-controlled bomb.

The idea of mine clearance by tactical bombers and artillery is also highly controversial.
Of course, individual groups of infantry can move through bomb craters. But not tanks, transports, or artillery.

Either the technical equipment of the army with bulldozers should be taken into account here. If bulldozers (or specially equipped engineer tanks) are available, they will quickly level the craters. Bulldozers were in the British and American armies.

Wehrmacht and Soviet armies did not seem to have them, or very few compared to the length of the front line. Ammunition was also spared, so minefields were cleared only by men or special tanks.

By this logic, American and British TacB and ATY units from 1944 should be able to clear minefields, while German and Soviet units should not.

[Lettos]

Decided to play the PGF standard Kursk scenario for Allied.
Advanced AI set an absolute record for stupidity, losing on the first turn. See screenshot.



Same mistake from AI side on Intermediate and Basic levels...
I'll have to not kill that infantry attacking across the river to look at the sequence of movement of the AI in the future and the sequence of movement of my units too.


But again the same observations in changing Random as before.

At the Basic level there are no such terrible losses from the attack, and in the case of Rugged defense the attacker also does not suffer such great losses as at the Advanced and Intermediate level. Isn't that a lot of overlap?

Otherwise it is the same: first the armored vehicles attack the Soviet army's anti-tank guns. They take heavy losses. Then the Wehrmacht tanks come in and also hit the ATG head-on. Then the infantry behind the tanks has absolutely nothing to do, and the AI sends them forward in trucks and armored personnel carriers to the ATG hexes vacated after the attacks, exposing the ATG survivors to attack.

Walter Model rolls over in its coffin, jingling its iron crosses...

Incidentally, again on Basic level, the infantry is not as active in moving forward on their vehicles to the anti-tank guns.

In general, it turns out to be very useful to play the same scenario at different AI levels playing for both sides and, if possible, in all three games: PG1, PGF and OG.
I guess I should have some serious fun with the Stalingrad scenario for this purpose.

[Radoye]

C) Purchased land units cannot be placed on hexes exhibiting terrain which would be inaccessible to them (i.e., "unenterable") during such units' normally envisaged movement phase (e.g., Sea, Rough Desert, Muddy Ground etc).

There's one very annoying exception to this rule, namely units with MOV = 0. Somehow they can be purchased and placed on hexes which normally they would not been able to access. They can also retreat onto the same if forced to do so by an enemy attack!