Pac 3.2.15 test

[bradk]

ORIGINAL: Istfemer

It is apparent to me that 3.2.15 is unsuitable for our PBEM game. Some important (and questionable) changes it brought appear to have been underdocumented.
My concern here is that 3.2.16 may turn out to be not much of an improvement over 3.2.15 in this regard. And I want it to be an improvement.
---
Keep on reporting, Zeke. I appreciate your feedback. We all do.

The game has a simulation of historical oil levels. Well, the game doesn't run at low simulated historical oil levels. New factories don't open, existing factories don't expand, and its possible for oil to get so low even open factories stop producing.

AI makes only a cursory effort to defend convoys from subs. But if a human player loads up Nagoya with DDs and DEs, and plasters the convoy routes with air - including pulling G3Ms and G4Ms from offensive activity for ASW work - there is negligible effect on the amount of oil. IJ just plain shuts down and the game isn't playable. Yes, it is unsuitable for your PBEM with Zeke.

[Rich Dionne]

ORIGINAL: bradk

There are undocumented changes in scoring beyond the multiplier feature.

Why do you say that? If you think there are some other effects on scoring beyond the multiplier feature, why not just ask me rather than making statements unsupported by facts. I don't think I've made any other changes to scoring other than those that I have documented, but if there's a bug, please explain what you've seen so I can try to fix it...

[Rich Dionne]

ORIGINAL: bradk
The game has a simulation of historical oil levels. Well, the game doesn't run at low simulated historical oil levels. New factories don't open, existing factories don't expand, and its possible for oil to get so low even open factories stop producing.

You got it partially right, Brad. Existing factories don't expand. New factories do open, but if they start with no production and they can't expand, then it's about the same thing. You're wrong about stopping factories producing. There's nothing in the code that stops existing factories from producing due to low oil. Please don't make statements unsupported by facts...

We must agree that oil and resources are important in this game, right? It's the reason Japan started the war, after all. So why isn't it reasonable to try to model this well? This is how the original version 1.22 runs oil and resources:

Date / Oil / Resources
7-Dec-41 / 53,000 / 50,000
21-Dec-41 / 31,805 / 32,799
4-Jan-42 / 13,855 / 18,104
18-Jan-42 / 8,614 / 2,835
1-Feb-42 / 4,229 / 3,035
26-Apr-42 / 6,146 / 18,290
6-Sep-42 / 4,551 / 25,938
10-Jan-43 / 3,621 / 56,606
6-Jun-43 / 1,854 / 211,009
9-Jan-44 / 846 / 655,059
29-Jun-44 / 3,029 / 880,399
9-Oct-44 / 3,254 / 999,969
1-Jan-45 / 4,400 / 1,031,162
25-Jun-45 / 4,115 / 1,049,592
6-Aug-45 / 3,765 / 1,043,143
7-Jan-46 / 3,846 / 1,014,019

Does this look reasonable to you? Is this what you'd prefer to stay with? Not me, I know I can improve this. Version 3.2 actually rolled over the resource reserves back to zero!

This keeps oil flat throughout the game. It really reduces the importance of the submarine war if you can't affect the oil reserves. These values stay constant even after sinking 267 TKs (which is very close to my latest loss numbers, by the way).

Look at resources. Did Japan really expand its resource reserves on a straight line towards infinity? No! They ran out of resources and supplies also. Soldiers and Civilians were starving, factory production decreased. Here, resources keep increasing even though Japan has lost 1671 MCS (actually higher than my latest loss numbers)

So what's wrong in trying to improve this?

Yes, 3.2.15 is not ready for PBEM tournament play, but folks can still test it and help me improve it, rather than flatly rejecting it. I'm going to get this right, and I'm getting close...

AI makes only a cursory effort to defend convoys from subs. But if a human player loads up Nagoya with DDs and DEs, and plasters the convoy routes with air - including pulling G3Ms and G4Ms from offensive activity for ASW work - there is negligible effect on the amount of oil.

Yes, I agree with this. Maybe it's worth my looking at the code...

[bradk]

ORIGINAL: Rich Dionne

ORIGINAL: bradk
The game has a simulation of historical oil levels. Well, the game doesn't run at low simulated historical oil levels. New factories don't open, existing factories don't expand, and its possible for oil to get so low even open factories stop producing.

You got it partially right, Brad. Existing factories don't expand. New factories do open, but if they start with no production and they can't expand, then it's about the same thing. You're wrong about stopping factories producing. There's nothing in the code that stops existing factories from producing due to low oil. Please don't make statements unsupported by facts...

Sorry. Oil below 1000 factories shut down. Thats a fact. Check your test files.

But that's hardly the central issue.

[bradk]

Resources is a different argument. Its never threatened IJs ability in any version. We're talking about oil here that kills the game for IJ.

Applying the historical scale to oil decline doesn't work in the game. We've talked before about the end result being what matters and games often may not have the ability to give historical output with historic inputs.

Most interestingly, after a significant - but a lot less than Matrix or GG - initial drop in oil, the 1705 obcs with 3.2.9 give a gradual drop from 20,000 or so to circa 6000 by Jan 45 in AI/AI. A human can do better but has to give up offense to cover the convoy routes. A good decision to ask a player to make.

At 6000 factories run but PPs are reduced. The effect wanted. And the result didn't even require putting in a system humans can't influence or sinking 300 plus TKs.

Now, a human Allies player could try to take back the oil bases to kill the oil. Wouldn't that be interesting? But oil is not going to just fall off the table automatically.

[bradk]

ORIGINAL: Rich Dionne

Yes, 3.2.15 is not ready for PBEM tournament play, but folks can still test it and help me improve it, rather than flatly rejecting it. I'm going to get this right, and I'm getting close...

No you're not close. This isn't even ready for Beta testing. I was stunned you posted about it here.

Yet the error corrected pair of 1705 obcs, now 1706, ARE ready for PBEM play. Bill and I have done it. Oil even works. What those obcs need is an exe not crippled by Matrix during its tantrum about AI cheating. That's the game that's close to ready. We could put it out for PBEM play right now. But since most people play AI, what's the point?

[bradk]

ORIGINAL: Rich Dionne

ORIGINAL: bradk

There are undocumented changes in scoring beyond the multiplier feature.

Why do you say that? If you think there are some other effects on scoring beyond the multiplier feature, why not just ask me rather than making statements unsupported by facts. I don't think I've made any other changes to scoring other than those that I have documented, but if there's a bug, please explain what you've seen so I can try to fix it...

It seems 169k kills Jan 45 with standard multiplier indicates something else is going on with kill points. Especially since you indicated you wanted to make changes in basic calculations.... civilian casualties, aircraft destroyed on the ground (how do you determine how many reported destroyed were repaired?), y'know that sort of thing.

[Rich Dionne]

ORIGINAL: bradk

ORIGINAL: Rich Dionne

ORIGINAL: bradk
The game has a simulation of historical oil levels. Well, the game doesn't run at low simulated historical oil levels. New factories don't open, existing factories don't expand, and its possible for oil to get so low even open factories stop producing.

You got it partially right, Brad. Existing factories don't expand. New factories do open, but if they start with no production and they can't expand, then it's about the same thing. You're wrong about stopping factories producing. There's nothing in the code that stops existing factories from producing due to low oil. Please don't make statements unsupported by facts...

Sorry. Oil below 1000 factories shut down. Thats a fact. Check your test files.

But that's hardly the central issue.

Just tested it. It's just as I said. Loss of oil does not stop production; it prevents expansion of factories, period! Please conduct the test yourself if you don't believe me, and again, please be careful about giving factual information.

Loss of resources are what shuts down production. I tested this also. It has nothing to do with 1000, it simply subtracts resources to make stuff until it runs out. Test it yourself.

And these are actually both fairly central issues, at least as far as I'm concerned.

[Rich Dionne]

ORIGINAL: bradk

ORIGINAL: Rich Dionne

ORIGINAL: bradk

There are undocumented changes in scoring beyond the multiplier feature.

Why do you say that? If you think there are some other effects on scoring beyond the multiplier feature, why not just ask me rather than making statements unsupported by facts. I don't think I've made any other changes to scoring other than those that I have documented, but if there's a bug, please explain what you've seen so I can try to fix it...

It seems 169k kills Jan 45 with standard multiplier indicates something else is going on with kill points. Especially since you indicated you wanted to make changes in basic calculations.... civilian casualties, aircraft destroyed on the ground (how do you determine how many reported destroyed were repaired?), y'know that sort of thing.

In Jan 45, the 3.2.15 kill multiplier will give about 50% higher total kills than version 1.22. So I'd say the values you are seeing are in line with my documented changes. The other changes are indeed changes I have made, but I documented these, right?

Mind you, I haven't been perfect in documenting everything, but I'm trying...

[Rich Dionne]

ORIGINAL: bradk

ORIGINAL: Rich Dionne

Yes, 3.2.15 is not ready for PBEM tournament play, but folks can still test it and help me improve it, rather than flatly rejecting it. I'm going to get this right, and I'm getting close...

No you're not close. This isn't even ready for Beta testing. I was stunned you posted about it here.

Brad, you seriously have no clue as to how close I am.

I am indeed getting close, and it will build on the coding work I put into the routine convoy system in version 3.2.9 and the all the versions since, which are all superior to 3.2.9, by the way. I'm guessing, but I imagine you haven't tried versions 3.2.13 or 3.2.14.

I've always posted updates here at the same time as on "Pacific War Update" in case not everyone here tunes in there.

[bradk]

ORIGINAL: Rich Dionne

ORIGINAL: bradk

ORIGINAL: Rich Dionne

Yes, 3.2.15 is not ready for PBEM tournament play, but folks can still test it and help me improve it, rather than flatly rejecting it. I'm going to get this right, and I'm getting close...

No you're not close. This isn't even ready for Beta testing. I was stunned you posted about it here.

Brad, you seriously have no clue as to how close I am.

I've tested four of your scenarios and as a comparison GG 1.1, Matrix 3.2, and your 3.2.9 with obca1705. I have posted the results of those tests to Pacific War Update. So I really do know how close you are...(n't).

The game not running on your simulation of historical oil is a fatal problem. Reason enough not to let this version out in public.

[Rich Dionne]

ORIGINAL: bradk
I've tested four of your scenarios and as a comparison GG 1.1, Matrix 3.2, and your 3.2.9 with obca1705. I have posted the results of those tests to Pacific War Update. So I really do know how close you are...(n't).

The game not running on your simulation of historical oil is a fatal problem. Reason enough not to let this version out in public.

Brad, thanks for running the scenario a few times; it's helpful to get constructive feedback. As I posted over on "Pacific War Update" on 26 June, this problem had been driving me crazy for some time. I finally found (and fixed on 26 June) a couple of big issues in GGs original routine convoy coding.

The first was in some faulty logic comparing cargo carrying capacity in a current ship unit versus required pickup or delivery requirement versus ship availability in the pool that caused additional ships to randomly appear in the pool, particularly in the tanker ship pool.

The second was a flaw in the number of ships selected in the routine convoy system. It is also random. If the code comes upon a ship unit with, for example, only one ship, and there are no ships in the pool, then only that single ship is used to meet the need. If this happens when the code is trying to pick up resources from, say Shanghai (95), then much resource, or oil, is not picked up. This issue really made for an unstable routine convoy system. The other thing this does is only use the number of ship units equal to the number of, say oil sites, that are available for collection. So many of the ship units never get used in the routine convoy system.

I'm now seeing a much more stable routine convoy system, and as a result, more stable oil and resource levels.

Brad, don't forget, this game is free, we only have maybe a dozen people currently involved with this game. Why wouldn't we put it out there for folks to try? There are many other updates besides routine convoy. Getting the few folks out there who have time to help and test the updates is a great benefit.

I think I've already fixed the routine convoy / oil & resource problem, and am currently testing it, plus working on other things...

So no, you have no idea what I'm working on, no clue what's been fixed, and no clue how close I am to getting things working right. Stop saying you do. Are you standing over my shoulder watching me work on this?? Thankfully, No!

[bradk]

Your 3.2.9 (2016) worked with an appropriate obc. Even oil. Best PBEM game ever. Your 3.2.15 (2019) doesn't work. No looking over your shoulder needed. Now, if you'd just look at 3.2.9 and put back the stuff Matrix took out during its "AI Cheats" tantrum we'd have a game everyone could play and enjoy. (During my testing it was a joy to see GG 1.1 take DEI and approaches BY APRIL without extra PPs, without free LCU activation, without weakened Allies LCUs, without a Max Help setting, and without any other nuclear bandaid "solutions".)

[bradk]

ORIGINAL: Rich Dionne

ORIGINAL: bradk

ORIGINAL: Rich Dionne


You got it partially right, Brad. Existing factories don't expand. New factories do open, but if they start with no production and they can't expand, then it's about the same thing. You're wrong about stopping factories producing. There's nothing in the code that stops existing factories from producing due to low oil. Please don't make statements unsupported by facts...

Sorry. Oil below 1000 factories shut down. Thats a fact. Check your test files.

But that's hardly the central issue.

Just tested it. It's just as I said. Loss of oil does not stop production; it prevents expansion of factories, period! Please conduct the test yourself if you don't believe me, and again, please be careful about giving factual information.

Loss of resources are what shuts down production. I tested this also. It has nothing to do with 1000, it simply subtracts resources to make stuff until it runs out. Test it yourself.

And these are actually both fairly central issues, at least as far as I'm concerned.

Its entirely possible I was concentrating on oil going above and below 1000 and not paying attention to resources. If so, then an AI opponent ran IJ out of BOTH oil AND resources. How embarrassing. How's a human to improve on THAT performance?

[Rich Dionne]

ORIGINAL: bradk
Your 3.2.9 (2016) worked with an appropriate obc. Even oil.

There's too much important stuff in 3.2.10 through 15 (I looked back through the list). You're hyper-focused on the oil issue in 3.2.15; yes, that was a major issue, but it's being fixed.

[bradk]

No, Rich, YOU'RE hyper-focused on the oil issue. Historical oil, historical oil, historical oil, that's all you've said for years. Even have a oil study in the game material.

Adjusting base quantities was't enough. Or changing the number of TKs by a little bit. No, no, you had to change the whole convoy and consumption system, adding hundreds of extra sinking of hundreds of extra TKs that for some reason only get loaded 15% or 20% of capacity so you could get historical oil only to prove the game still doesn't produce historical oil a lot of the time and when it does the game won't run.

[bradk]

"There's too much important stuff in 3.2.10 through 15 (I looked back through the list)"

Every time I read that list, after almost every line I think "Hey Rich, over there, the crippled AI, what about that?"

Its ludicrous that I, someone who doesn't play AI, is an advocate for fixing it while you, who I've never seen ask for or accept a PBEM game, ignore it.

[bradk]

In case anyone other than Rich and I are still reading this, here's the oil performance that's the point of contention. These are all Ai vs AI

I'm wondering if sinkings are even related to oil reserves. Don't see correlation.

And its a puzzle why the same input results in drastically different output.

[bradk]

"Brad, thanks for running the scenario a few times; it's helpful to get constructive feedback."

Lots of hours in those few times. Fortunately I could let two run at a time and do other stuff and come back and check on them every so often. (Which is why some of the dates don't exactly match.)

I figure the voluminous data provided is plenty constructive. I'd drop the commentary, which is apparently your objection, if I though the point would still be made.

[Rich Dionne]

ORIGINAL: bradk

In case anyone other than Rich and I are still reading this, here's the oil performance that's the point of contention. These are all Ai vs AI

I'm wondering if sinkings are even related to oil reserves. Don't see correlation.

And its a puzzle why the same input results in drastically different output.

You have no argument with me on this Brad.

I agree it is a major issue. As I said, I finally found the 2 bugs causing this (for once, I didn't cause these), one would randomly add significant numbers of ships to the pool because of an original code byte vs. word comparison error, and the other would randomly prevent oil and resources from being properly added to the stockpiles. They were causing the haphazard results you show above. Thankfully, I've fixed these and am currently testing, things are making sense now.