My Feedback

[murdock762]

Just finished my first play through (as Allies, default difficulty settings, ver 1/00/03/01)

Overall really like the game, functional graphics keep me informed and like the combat system overall. Good overall job of keeping the focus on the big picture (fun) and not micromanage everything

I read on some other posts that technology research is getting patched, was horribly out of balance in my play through.

My main observation is some serious tweaking of the AI would make game much more enjoyable. In my play through I (accidentally) brought Italy into the war on the first turn and it just seemed to totally throw the Axis timetable off. Germany was divert planes to africa and stacks of units stuck behind the line in northern italy and they really struggled to knock France out. France didnt fall until mid/late 1941. AI should be laser focused on knock out France as the early game is so predictable based on the Sept 1, 1939 start. After France surrendered they immeadiatelly declared on Vichy France. Finally they attacked and overran Spain. I understand the benefit of getting gibralter, but eastern Front was totally undefended. As soon as Russia was able to declare war I steamrolled in and game was over in mid 1943 (most of the land units still in Spain when I declared war).

To summarize I feel Germany needs a special script to focus on Northern France right away, and shouldn't even consider open up a new front in Spain until some minimum amount of forces have been moved to build up reasonable defense on the Eastern Front. (ie if war is going extremely well or ahead of schedule Spain is in play, but not make it a common opening)

I know is brand new game so hopefully some of these issues will be addressed in patches. Is a really good game and am already thinking about my upcoming Axis playthough

[AlvaroSousa]

A.I. is always being worked on. But to compare.... SC has been around for a long time. I barely read the rules, I barely play it. On the default level I stuff the Axis in France every time.

A.I. for games like SC and WP are incredibly difficult to do. For a single front game like WitE or UoC or DC it is very easy to do. I designed Assault on Communism and the A.I. was very easy to do. Players were complaining it was too hard. Completely different game in terms of dynamics. Single front games have a very specific static objective. I'd say exponentially easier to program. UoC did an excellent job with creating some 10 tactics the A.I. can use. For a limited front it was a lovely piece of coding. In a game this large you just can't do it because you all will be sitting there for 10m waiting for the A.I. to think. So there are compromises.

For example I played an SC Europe 1944 game as the Axis. The Russians abandoned the Romanian Front and ran for Warsaw. ALL OF THEM.

Try on a higher level. I am temped to increase the bonuses for the A.I.

A.I. is really that hard to code for these kind of games. I would love nothing more than giving this to Alpha Go and saying beat me.

But game A.I. is incredibly interesting to me so it is one of my main focuses when I have not much to do to improve WarPlan.

[murdock762]

Thanks for the response.

Was only one playthrough and on default level so maybe I have different experience next game.

But it seems like it would be simple (in concept, maybe not in coding) to not have Germany declare war on Spain until some minimum condition is met. Again I dont know if this is an outlier or this happens in most games

[AlvaroSousa]

The Axis have 2 major strategies and a 2-3 small ones, and some exploits.

It should DOW on Spain if France is conquered By X date. Sometimes it's real borderline.

[Meteor2]

You just said, that the Axis AI should DOW Spain if France is conquered?
What is the rational behind this? Franco was an semi-ally to Hitler and there is no real advantage for Germany in doing so.

[tyronec]

A.I. for games like SC and WP are incredibly difficult to do. For a single front game like WitE or UoC or DC it is very easy to do.

Have to agree and disagree with you on this. Am a play tester on WITE and the amount of effort in terms of programming and play testing is huge. Even getting the AI to form a reasonable front line and then to retreat or attack as appropriate has taken many iterations or adapting and testing, adapting and testing, adapting and testing.... over several years now.
It is really difficult to get the AI to work well on a single front game. Here you have the strategic decisions
to work on too. Given how much time and effort to get all that to function to a high standard makes me wonder if it is viable. Hopefully Alvaro can continue to make improvements but I don't think we should be expecting too much.

[AlvaroSousa]

I only said that from my own experience. I had an easy time designing AoC A.I. Didn't mean to offend anyone's talent. Granted I didn't have divisions to worry about. There is also a difference between busy code in which you have to bug hunt things you missed and complex A.I. code. Busy code is easy.

For example when you move a fleet out of port.
# of units picked > 0
Are these your ships
on your turn
are you docked with subs
are there any subs selected
do you have opPts

I am sure I missed 2-3 other checks
So the above is an example of busy code. It is tedious but doesn't take a lot of complex thinking.



For complex A.I. code I have something like 5-6 influence maps checking strengths, air power, threats, different conditions of attacks and defense. Then I have to formula the values and create a logic decision neural tree to have the A.I. decide what to do. That is not only time consuming but you are basically feeling in the dark because you don't know exactly how the code will be interpreted. The A.I. will follow the math to the letter doing things you don't expect because some exception came into play. If you youtube A.I. and problem solving it is one of the most fascinating things you could watch. They come up with ideas on how to solve problems based on the rules given in creative ways humans can't think about.

So when I mean more difficult I mean the complex due to exponentially more things that can be done. AoC was still tedious but all I had to do was write scripts for defensive lines and write unit gatherings for surprise offensives during the winter of 1941 to keep the Germans on their toes. Then go back to defense until they build up. Defending France is much easier to code then defending Russia would be a good example.

I wish I could rework it but that would take about 2000 hours of work. I wrote the entire A.I. system from scratch.

[tyronec]

I only said that from my own experience. I had an easy time designing AoC A.I. Didn't mean to offend anyone's talent

Alvaro I wasn't taking offence if that is how I wrote it my mistake. Just making the point that WITE are now on their 4th generation of that game system, building on their past code and experience and with more guys on hand to develop the game than you have and STILL it is a long and difficult job to get the AI to work well.