Matrix Games Forums

Very interesting to see some of the the considerations on how the AI is handled.
And great to see the steady progress on the AI.
And not least - and still! - what an incredible commitment and amount of work fro you, Steve - it is truly extraordinary!
Thanks for this gigantic effort!

Regards
Nikolaj

Will version 5.x be available to all users, only Member's Area or only to Beta Testers?

Falken wrote: Wed Nov 09, 2022 11:45 pm Will version 5.x be available to all users, only Member's Area or only to Beta Testers?

Am also interested in this.

December 16, 2022 Status Report for Matrix Games’ MWIF Forum

General
My chorus has had 5 performances in the first half of December, which explains why this status report is so late.

Product Releases
In November, version 05.00.00.10 was made available as a hot patch for all customers. It contained the code for the AI Opponent to play the Barbarossa scenario as the USSR up to the beginning of the second impulse of the game. It also had the code for the new form on AIO statistics that I described in October’s monthly status report (in a post above).

Bugs
I did not make any substantive changes to fix bugs in November - only a couple of modifications for cosmetic improvements and another to avoid a non-fatal error.

Missing Optional Rules & Half Map Scenarios
Nothing new in November.

AI Opponent (AIO)
Just as last month, work on the AI Opponent for World in Flames continued with steady progress, although nothing spectacular. Using some saved games provided by customers, I was able to reproduce and fix several bugs in how the AIO, as the non-phasing side, processes land combat resolution. That phase needs more testing (by me) because there are numerous outcomes, depending on the attacking and defending units, whether the attack is Assault or Blitz, and (of course) the die rolls for the combat. This past month I have tested many of those outcomes but I want to rigorously check each of the possibilities, in order to convince myself that the code is completely clean.

I have been able to get the program to proceed to the point in the sequence of play that the AIO is the phasing side (second impulse) and deciding about its land moves. Presently, that decision making code by the AIO is incomplete.

I am still looking for saved games from the early part of the Barbarossa attack. They can be from any mode of play (solitaire, head-to-head, NetPlay, AI Opponent). If you have some, please post them to the thread Planning Land Movement in the sub-forum AI Opponent Discussion. I will use them to test how well the AIO makes decisions about its land moves under different circumstances.

August 8, 2023 Status Report for Matrix Games’ MWIF Forum

General
I have been working exclusively on the AI Opponent for the past 6 months. Most recently I took a 3 week cruise to Alaska that I got back from at the beginning of July.

Product Releases
The private beta testers for the AI Opponent code received version 6.0.0.0 for testing June 12th. While the code I had in place for the AIO to move units was functioning for my saved games in late May, it wasn’t able to perform correctly for the private beta testers we enlisted. However, I was eventually able to get it to work using their saved games - moving the USSR land units in the second impulse of the Barbarossa scenario. I just uploaded a new version (6.0.0.1) for the beta testers today.

Bugs
Other than to get the AIO to work, I did not make any substantiative changes to fix bugs.

Missing Optional Rules & Half Map Scenarios
Nothing new since last year.

AI Opponent (AIO)
While the code I had in place for the AIO to move units was functioning during my personal Alpha testing in late May, it wasn’t able to perform correctly for the private beta testers we enlisted to test the AIO performance. In particular, when Germany didn’t align Finland during the first impulse of Barbarossa, the AIO could not figure out a front line for the USSR stretching from the Arctic Ocean to the Black Sea.

After adding code to treat Finland as a potential Axis controlled minor country, with land units that could move and attack, the AIO found the existing front line and could work out where to move its land units.

Moving the USSR land units during the Land Movement phase is accomplished by the AIO in 3 major steps. First it evaluates each land region in the game (Barbarossa scenario) and assigns each a Status. Land regions completely controlled by Germany are either No Threat, Threat, or Neutral. The No Threat land regions are not adjacent to land regions where the USSR controls some hexes. Land regions completely controlled by the USSR are either Safe or Vulnerable. The latter are adjacent to land regions where Germany has land units (or potentially could have land units in the case of Finland and Hungary). Land regions where both sides have partial control have a Contested Status, which can range from Co_Awful to Co_Excellent, depending on how well the USSR is doing in controlling the land region.

For the second step in determining land moves for the USSR, the AIO identifies the current front line. The front line consists of all friendly hexes which are adjacent to enemy controlled hexes. That starts at the Black Sea and continues northward to the Arctic Ocean. If frequently is broken into two pieces by the Baltic Sea.

The third and final step is for the AIO to decide in which land regions it wants to Attack, Hold, and Retreat. Based on those characterizations and given how far each land unit can move, the AIO determines each unit’s best destination hex. There are a lot of factors that feed into those decisions. Keeping the units in supply, not making them disorganized, and finding hexes with good defensive value are key considerations. A bunch of other stuff also comes into play, such as stacking in the destination hexes.

When the AIO moves units, the program centers on the destinations, one by one, and displays a small message as to which unit just moved, its starting hex, and its destination hex. The unit always ends up on the top of the destination hex, so it is easy for the player to identify the unit that moved. See figure below.

Hello Steve.

Thanks for a good game. However, I am wondering why to prioritize AI when basic mechanisms either seems not to work or be extremely difficult to use and comprehend. Especially the resource/convot transportation leads to a lot of frustration among players. For me the present system or rather how the system handles the convoys has lead me to abandon the game for months now. I know other players have quitted the game entirely mainly due to this issue.

I understand the challenges connected to develong a functioning AI but somehow AI doesnt seem very important to the average player either playing solitaire training sessions or online with other players.

Is it possible at all to prioritize the convoy system?

Best regards

Anything since August?

Please see the End of Year AI Progress Report here:

https://www.matrixgames.com/forums/view ... 0&t=400470

Seems that the multiple and annoying bugs that the game still has, will be appointed (maybe) some day in the distant future, because it looks the AI matter is going to take one lustrum at least.

dyrn wrote: Thu Aug 10, 2023 7:07 pm Hello Steve.

Thanks for a good game. However, I am wondering why to prioritize AI when basic mechanisms either seems not to work or be extremely difficult to use and comprehend. Especially the resource/convot transportation leads to a lot of frustration among players. For me the present system or rather how the system handles the convoys has lead me to abandon the game for months now. I know other players have quitted the game entirely mainly due to this issue.

I understand the challenges connected to develong a functioning AI but somehow AI doesnt seem very important to the average player either playing solitaire training sessions or online with other players.

Is it possible at all to prioritize the convoy system?

Best regards

+1

lucliu wrote: Mon Jan 01, 2024 4:03 am

dyrn wrote: Thu Aug 10, 2023 7:07 pm Hello Steve.

Thanks for a good game. However, I am wondering why to prioritize AI when basic mechanisms either seems not to work or be extremely difficult to use and comprehend. Especially the resource/convot transportation leads to a lot of frustration among players. For me the present system or rather how the system handles the convoys has lead me to abandon the game for months now. I know other players have quitted the game entirely mainly due to this issue.

I understand the challenges connected to develong a functioning AI but somehow AI doesnt seem very important to the average player either playing solitaire training sessions or online with other players.

Is it possible at all to prioritize the convoy system?

Best regards

+1

I have to agree with this post as I too have abandoned this fine game for years waiting for improvements to the convoy system. I feel it's a massive mistake to not include oilers, that has been the biggest problem with the convoy system ever since the original WiF computer game put out decades ago by ADG. Put oilers into the game and fix the convoy system. Wasting time on the Artificial Ignorance is never going to produce an AI capable of playing this complex complicated game well.

Hate tankers. They do not accurately represent the period. They unbalance the game absolutely to favor the Axis.

In fact, the way the game represents oil expenditure is absurd and exaggerate. The new version, WIFCE , which is not MWIF's, is way more equilibrate.