We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
You are using an out of date browser. It may not display this or other websites correctly. You should upgrade or use an alternative browser.
Go to the 1.10.2 page at the top of the forum and on the dev post there are changes added today.
One of the changes is
"the AI will now be more likely to bring back the Kaiser in non-historical mode (WTT only)"
Yeeeeeeeeees.
Welcomed (since the default chance was in a range of 0,5% or something)
I just hope it's not something crazy like 50% Fascist /50% Kaiser.
I prefer the games without historical focus to be like a more dynamic, randomized gameplay with somewhat plausible outcomes.
Not an explosion of releasables, ideology flips, weird factions, and civil wars that make no sense every game.
Britain is already decolonizing almost 90% of time in non-historical and it's really annoying. It causes a lot of lags, and doesn't bring anything at all to gameplay (100 AI releasables with 0 factories and 0 army are not useful in my opinion).
Germany's ambitions are also a bit too automatically opposed by Austria and Czechoslovakia in non-historic, making it really weak.
Welcomed (since the default chance was in a range of 0,5% or something)
I just hope it's not something crazy like 50% Fascist /50% Kaiser.
I prefer the games without historical focus to be like a more dynamic, randomized gameplay with somewhat plausible outcomes.
Not an explosion of releasables, ideology flips, weird factions, and civil wars that make no sense every game.
Britain is already decolonizing almost 90% of time in non-historical and it's really annoying. It causes a lot of lags, and doesn't bring anything at all to gameplay (100 AI releasables with 0 factories and 0 army are not useful in my opinion).
Germany's ambitions are also a bit too automatically opposed by Austria and Czechoslovakia in non-historic, making it really weak.
I'd prefer 65% fascist 25% nonaligned and 10% democratic germany. At least until the Soviets get revamped and can be a replacement for Germany's as the "pace maker" of the game
As there is bound to be a heated argument about this change, here are some numbers. The AI weight for Oppose Hitler is now 7 (from 1); other weights remain unchanged.
Bringing back my wildly unpopular dartboard visualization, the effective first pick probabilities change as follows (yellow: 4 YP, green: Rhineland, blue: Oppose Hitler; area is proportional to probability): ->
It is pretty clear there will be no issue (by restarting a couple of times) to get Oppose Hitler. For the persistent reloader it should even be manageable now to put Germany on Oppose Hitler plus getting another specific semi-unlikely choice elsewhere. So I'll focus on the opposing view - how many runs are spoiled (or however you prefer to call it) by AI Germany picking Oppose Hitler eventually, tracing the most likely paths:
4YP: 51.5%
... followed by Rhineland: 20.2%
... followed by Oppose: 6.1%
... followed by Autarky: 15.0%
... followed by Rhineland: 3.7%
... followed by Oppose: 0.8%
... followed by Westwall: 10.2%
... followed by Rhineland: 4.7%
... followed by Oppose: 1.8%
Rhineland: 34.8%
Oppose: 13.6%
That's 63.4% Rhineland versus 22.3%. The remaining 15% are covered by numerous less likely paths. Roughly, AI Germany should be expected to Oppose Hitler in about 25% of games.
Considering the large global impact of the focus, I'd have preferred something lower, like weight 4, putting the overall probability in the upper single-digit percentage range.
Bringing back my wildly unpopular dartboard visualization, the effective first pick probabilities change as follows (yellow: 4 YP, green: Rhineland, blue: Oppose Hitler; area is proportional to probability):
I have absolutely no idea what those graphs are intended to represent. Even after reading the thread you link to. And the thread linked to in that thread. Though it did help me to understand the decision making behind the obvious lie that ai_will_do is.
They were a (clearly failed) attempt to draw a visual connection between how the die rolls can be grouped (by analyzing the value ranges shared between focuses) and the resulting probabilities. E.g. the full yellow outer ring covers the 1/6th of cases where 4-Year plan is chosen by an unbeatable dice roll between 10 and 12. The next inner layer covers the cases of the highest die roll being between 7 and 10, shared between 4YP and Rhineland, etc. I'll shut up about it now , you can take a look at the attached Python script for more details.
It very likely is the simplest possible algorithm that technically fulfills the requirement "random selection based on weights". And isn't technically correct the best kind of correct?
If you search for "Berlin-Moscow" in the Steam workshop, there are four mods with matching names. All of them are very outdated though and none changes just the AI. There does not seem to be an easy way to find mods that are current and just change particular files or directories.
In my experience it is easier in such a situation to write the mod from scratch rather than extract/update/filter/fix/customize what the existing mod does.
For the same reason you should try to use new file names in your mod wherever possible instead of modified versions of vanilla files (i.e. common/ai_strategy_plans/MYMOD_berlin_moscow_axis.txt instead of common/ai_strategy_plans/GER_alternate.txt). Makes it much easier to tell what your mod actually adds and to update it when necessary.
->
