Data analysis dump. version 1.18 stats

Way too tired today to keep working to analyze this, but here are the median values for each turn in build 1.18, first 31 turns. Popularity means potential vote rate in this case.
GDP vs Popularity versus ‘winning’ which is an internal metric.


I immediately realized looking at this today that my internal code to balance the game based on how easy/hard you are finding it (winning) is based on the wrong data. Its tracking voting probability, which is linekd to, but actually a side-effect of the actual approval.

as has been much discussed, approval can rise up for ages, showing you are doing better, but only when it passes 50% (in 2 party system by default) will people vote for you. Thus the voting probability (shown above as popularity) is a trailing indicator, and thus too slow to react.

Thus we get a situation, clearly shown above when the simulation flips as approval crosses 50% for many people, and the country is suddenly seen as winning when it was previously losing. This explains that rapid climb in every country just before the first election at the 15-19 turn mark.

I’ll fix this, which will make the balance mechanisms more responsive, and the game less wildly gyrating.

Also it does look like winning is tracking mere popularity too closely. Other factors come into it, such as debt and deficit, but I think they are currently underweighted.
Hopefully these factors combine tom improve the games balance, especially in the second term.

1 Like

interested in how you plan to fix it. It would be nice to add some sort of expanded political system. Something to make changes a little more lumpy and a little more decisive.

When i think of current democracies and democratic parties i think partisanship needs to play a much higher role. Perhaps something like the political quadrants you have in the game. If each voter can be placed in the quadrant they should be more likely to vote for a party that represents their quadrant no matter how happy they are with the current situation.

I’ve recoded the winning detection algorithm entirely today. If I’m right, this rewrite will totally fix the lumpy-response issue

1 Like