Dev Diary: AI and Feature Updates

Today's Development Diary by Podcat

Hi everyone! From the time we show off features in dev diaries and release a lot happens. Usually its just bugfixing and work behind the scenes, but sometimes it can be bigger changes too. So today we will pick out some stuff thats changed since you last read about it. But first, but not least, lets talk a bit about AI development!

AI Updates

When we started with 1.9 'Husky' and La Résistance there were 2 main things we wanted to achieve. First we wanted to nerf germany (without actually nerfing them) by giving resistance more teeth. This worked pretty well, as after we had made the new resistance system Germany would consistently collapse to large polish uprisings behind its lines and needed to be expanded how to manage all that as well as control its equipment production better. Resistance is now constantly hurting garrisons and you don't pre-build garrisons, so stockpiling and not rushing units is more important.

The second thing is also related to Germany (its ww2 after all). For HOI (no matter the version really) we have always strived for a balance where Germany defeats the Soviet Union if the allies do not take pressure off by invading. This ensures that all of the big sides need to be active and if played by a player you need to step up. Now before we jump to the 10 pages of comments discussing if this is a realistic target... It is, for gameplay I think. History wise I am personally convinced that the Soviets would have beaten Germany on their own. It just would have taken a lot longer and been much costlier, and there isnt much point in playing the UK or USA if Soviets will just take care of Germany on their own (now now, please keep reading the rest of the diary before we start arguing what-ifs :p).

Anyways, in 1.7 the allies seldom pulled off invasions well enough, and pretty much only if they got lucky in italy or punched up through greece or something like that. Firstly we have made a bunch of improvements and bugfixes to wierf parts of the invasion logic for the AI that makes them perform better. The biggest impacts came from limiting the AI from shipping troops back and forth across the globe and simply getting them to buffer up its forces in specific allied territories. USA specifically here will now station troops ahead of time in UK so that when its time to do invasions they are already in the neighborhood. We have also improved prioritization of different fronts to make sure AI gets the importance of Europe and when landings have been made successfully.

One of our most important tools when working on high level AI stuff and history tuning is the "History Logger".

This is a tool that we developed to observe AI gameplay. When enabled, it collects & dumps data on how game progressed during a particular run and its output looks something like this when run through a web interface:

Other than showing country borders, it can also show many other other details for a selected country such as diplomacy status, army/navy mission/air mission positions, orders, taken decisions/focuses/ideas, executed operations/operative missions and many other things. On top of that, the things are logged using log effect will also appear in these logs so you can have custom logs in your effects if you want. Here is an example from USA:

The lines are invasion orders and red