EIA v1.25.03 Beta Hotfix #6 Release Available

[pzgndr]

I'm so sorry. But I have another error. But this time I have a save game with which I can produce the error.

Don't be sorry, this is fine. Just another issue for me to investigate and resolve. Thanks!

This one was interesting. First of all, I cannot understand why the code is trying to add a fleet's ships to the area builds if it must evacuate but cannot? In this case, the French-controlled Swedish I Hvy Flt was trying to evacuate from Abo to Lille. So I have commented out that section of code. Secondly, the only reason the code got to this check was because it cycles through nations to eliminate and convert control, and Swedish-contolled Finland (#34) comes before Sweden (#35). So for Finland it got to the point of trying to evacuate Abo but couldn't. When it got to Sweden it deleted all remaining Swedish units, which was just the I Hvy Flt. So you got this obsure error message, clicked through it, and all was well. Russia controlled Sweden and Finland and all Swedish units are removed.

I am not going to update the hotfix but this little change will be in the next build.

[hapshott]

Thanks for checking this.

I must say, that I'm impressed with the support you give. I really like this game.

I will continue with this savegame. As you said. After clicking the ok button I could continue.

Best regards,

Hapshott

[pzgndr]

A quick update. I've been working on Hotfix #5. I received a bug report regarding the release of prisoners and a couple of reports regarding some AI weaknesses. I've fixed the prisoner release and added some code to look at local naval superiority between Turkey and Russia, considering that the Russian fleet is split between Baltic Sea and Black Sea; this should restrict some reckless blockades. We're still seeing weak garrisons attacking stronger besiegers, and I think I found and fixed the issue but not sure.

Something else I found and worked on is what the AI does when it can't assign enough forces to a target and then looks at an adjacent 'prep_area' to advance to. There was code to unassign units from the target area, but nothing to unassign units from the prep_area if it couldn't assign enough forces. This is probably what has caused piecemeal advancements and puts too many weak forces at risk being over-extended. I still need to playtest this some more to see how it goes.

Another area I relooked (again) was how the AI places depots. This has been ridiculously challenging over the past couple of years. I had a "cheat" coded for the AI to allow moved units to place depots to help it keep armies supplied and minimize forage losses. Well that wasn't working and I figured out why so now it is working. AI France in 1812 is now advancing quite well into Russia and playing Russia I've had to run away fast. In one playtest I got butchered pretty good, and not so bad in another playtest.

1812 has been a good scenario to watch and fine-tune AI performance. Spain usually surrenders which frees up French corps and depots for Russia. And French corps were leaving Spain but stuck in southern France. British forces would advance into Spain some but then get stuck there, too weak to attack France but not returning to Portugal for other operations. So I've been working on these issues some.

I mentioned "cheats" above. Other than several AI bonuses that are documented, there are only a few minor cheats where the AI gets a pass on some rules. One is for placing depots because the AI needs help here. There's another cheat for allowing the AI to move through neutral territory with no access if moving back towards its capital. One other cheat is some help for AI Turkey versus Egypt, just a bit to get it going in 1805. Other than those, the AI is bound by the same rules as human players. Just so folks know.

I hope to wrap up my playtests and get Hotfix #5 out later this week. Assuming no more annoying bugs or AI weaknesses, I should be sending files to Matrix by end of the month for the next official patch. That's the plan.

[pzgndr]

Hotfix #5 posted. As discussed above, a few bug fixes and some AI improvements. AI placing depots is working well; the computer opponent is supporting its line of advance much better now. And that prep_area issue is now working well. I think this was a major bugaboo in why the AI offensive actions were not as aggressive or efficient as they were intended to be. Watch AI France in 1805 and 1812 now and notice some differences!

I still want to playtest some more over the next week to verify a few things and wrap up documentation. If anybody spots something that I need to review, let me know and provide a save game. We should be very close now. I should be sending files to Matrix by end of the month for the next official patch. That's the plan.

[pzgndr]

I've posted a couple of AARs above showing my playtesting of the 1805 and 1812 scenarios with Hotfix #5. I'm still spotting and fixing numerous little issues, more than I care for at this point but with every little fix we get closer and closer to having a game without any little issues. I intend to do a couple more rounds of playtesting to ensure this upcoming release is about as rock-solid as I can get it. I may release another hotfix when I wrap up everything and send files to Matrix, so that Hotfix #6 will be the same as the official v1.25.03 release.

Several players have downloaded Hotfix #5. I am still interested in anything anyone has to report from playing it. You may have spotted something that I haven't seen yet. So, please speak up with any feedback you may have. Thanks!

[pzgndr]

Another quick update. I've been banging on this game hard for over a month. I think I've made noticeable improvements in many AI areas: placing/removing depots, DOWs, surrenders, reinforcing battles, redeploying forces, naval actions, retreats, advances, and strategic redeployments... I've lost count. My routine has been to run a playtest game, look for issues, stop to investigate/resolve those issues, continue or repeat. And then playtest again.

Today I completed two rounds of 1805 and 1812 and only had minor AI adjustments to make. I'll run a couple more confirmatory playtests next week and most likely wrap it up. There are still some deeper AI enhancements I could make but for now the "generic" AI computer opponent is doing ok. Good enough for the next official patch.

[roos]

This is an odd bug:

Hotfix 5
France managed to conquer Munster from CotR (controlled by Austria) without being at war with neither Austria nor CotR.

There was a Russian garrison in Munster and France is at war with Russia.

Ill try to recreate a save fil, but i didnt notice until too late, so i will have to retrace from an economic phase. May take a few days until i have the time.

PS
About GBR AI. I have never seen AI move any units out of GBR, not even when AI-controlled Prussia and AI-controlled Russia almost crushed France.

[pzgndr]

France managed to conquer Munster from CotR (controlled by Austria) without being at war with neither Austria nor CotR.
There was a Russian garrison in Munster and France is at war with Russia.

About GBR AI. I have never seen AI move any units out of GBR, not even when AI-controlled Prussia and AI-controlled Russia almost crushed France.

Good feedback, thanks. I'll take a look and make some adjustments. There's probably an issue with converting control of a kingdom minor. And GBR should be looking at amphib ops against France in the later game after 1812, but I wouldn't be surprised if my logics need to get fixed. (It wouldn't be the first time... LOL)

Update. If you still have that game and could at least provide me some excerpts from the game log for when France 'conquered' Munster, I may be able to track down what may have been calling the nation_convert_control function. I don't want to mess with the function itself, but I agree a nation not at war with a minor or its controlling major power should not be able to gain control. So how and when the code calls for it needs to be reviewed. If you don't have a save, this will go on the back burner for a while.

[roos]

Screenshot (28).png (6.94 MiB) Viewed 887 times

[pzgndr]

France managed to conquer Munster from CotR (controlled by Austria) without being at war with neither Austria nor CotR.

Can you provide me a save of your game? Munster fell in Apr 1810 and your screenshot is May 1810. I can further review your game log and perhaps recreate the situation. That would be helpful. For GBR amphib, I relaxed some of the restrictions slightly, so maybe they will be more active in the later game.

BTW, I am just about ready to wrap up what I have and send files to Matrix for the next official release. I'd like to investigate this issue (one more last issue!) before I go final.

UPDATE. Nevermind. By process of elimination I determined that something in the endphaseturn nation_scan_all_areas_for_control check was the likely culprit and further determined that the area_determine_owner function needed a redundant check to verify that a majority invader was actually at war with the minor or controlling major power. I had recently added something for determination of majority control of a contested area but did not consider a case like this. I'll playtest some to verify I haven't messed anything up, and then wrap this up.

[pzgndr]

FYI, I have wrapped up Hotfix #6 and sent update files to Matrix for the next official update. This morning I posted a short AAR at https://www.matrixgames.com/forums/view ... 8#p5116898 showing some of the recent improvements.

The area_determine_owner function changes that I discussed above did not make the cut because it messed things up. It's still on my backburner list along with a few other minor issues to resolve later. We'll get there.

[pzgndr]

Hotfix #6 release version posted. This update has all of the files that I sent to Matrix last week except for the large documentation files (Game Manual and Help files). As discussed recently, I've been playtesting and adjusting the AI, over and over again, until I achieved consistent results with AI behaviors that I expected to see. Most everything is now working well to my satisfaction. Hopefully we'll get the official update out later this week or next week. Except for the documentation, folks can start playing the release version now and see what they think.

Right now I'm taking a break from code development and I'll be on travel in early September. After that, I'll work on another hotfix EXE for the official update to resolve a couple of minor issues that popped up recently (e.g., the area_determine_owner function problem) plus any reported bugs or obvious AI deficiencies with the official update (I'm sure I didn't catch everything and some players will surely spot something). My intent is to have THIS official update as playable and as challenging and as rock-solid as possible before moving forward.

After that, I'll post another State of the Game update and my thoughts about future plans. Players are always welcomed to provide feedback and suggestions for improvements.

UPDATE 8/25/2023. I have verified the Matrix installer so the official update should be posted soon, probably next Tuesday 8/29.

[pzgndr]

Bug!
Supply route broken.

I have circled back to this bug and I've started working on finding an alternate path if the primary path fails. So far it looks like I'm on the right track. Austrian units in Adrianople are now tracing supply back to the depot in Nicropolis 3 areas away via the counterclockwise route around the Turkish feudal corps. Basically I'm stepping through a path that avoids mountains and marsh except for the destination area. I need to continue to test this as the find_path function is used in many ways and I want to ensure my new find_alt_path does not create new issues. This should be in the hotfix EXE in another month or so for the v1.25.03 official update just released.

bug supply.jpg (983.2 KiB) Viewed 616 times

[pzgndr]

Part II of this bug fix. Simple test case, moving from Venice to Milan. The code would always try to find the path through the Tyrol mountains to reach Milan, which was highlighted as a valid movement but then couldn't make it. Now the alternate path avoids the mountains and moves through Mantua and Lombardy to reach Milan. I'm trying to restrict this new find_alt_path function to cases of supply path tracing up to three areas and land unit moves up to three areas for non-French and up to four areas for French. I'm excluding cavalry moves because I want the code to optimize for three areas (four for French) and not go beyond that for a failure. I've had to go back and forth on the logics several times but I think this is working now and should be a nice improvement to the game. A human player could always undo a "bad" pathfinding move and manually make a good move, but where this will help a lot is for planned AI moves that sometimes fail and result in a lapse of war or units not reaching a battle.

Frankly, I've thought about this over the years and moreso over the past couple of months since the supply bug was reported, but I've need some quiet time to really dig deep into the pathfinding. This is good.

bug supply 2.jpg (1.04 MiB) Viewed 594 times