I am at work on a new feature for President Infinity – the Polls Editor. This is part of the Campaign Editor.
Here, I am entering info for a Dec. ABC-Washington Post poll for the Republican primaries, which has (among other things) Donald Trump at 38% and Ted Cruz at 15%.
This feature has two main aims.
1. It will allow for starting %s in a game to change based on the start date. Currently, there are only primaries or general election %s.
2. It will make it much easier for a campaign designer to figure out starting %s for a given date. Just enter the polling data you have available for around that date. The computer will fill in any of the blanks automatically, using region issue similarities to find the best fits. No more guessing what Bobby Jindal’s %s are in North Dakota.
You can see that you can enter both general election polls and primaries polls. For the general election, you can also enter generic numbers (‘Republican’, ‘Democratic’) or leader-specific %s for polls that looked at hypothetical or real match-ups (this will become more important when Favorability is implemented – that is set to be implemented next).
You can also see that there are specific organizations associated with a poll – for now, that’s just to help keep track of which poll is which. In the future, this might be expanded to allow for tendencies towards over- or under-sampling by various organizations which lead to skewed results.
Finally, there is a margin of error box for each poll (set to 6 in the above screen shot). The computer uses this when figuring out starting %s, especially when there’s a difference between polling numbers that are relatively close in time.
I like this idea. The thought of having different starting percentages really is amazing.
Another very nice addition. Great job Anthony.
“Finally, there is a margin of error box for each poll (set to 6 in the above screen shot). The computer uses this when figuring out starting %s, especially when there’s a difference between polling numbers that are relatively close in time.”
How does the engine aggregate these results?
What I mean is: when the engine uses polls to determine starting support, which polls does it use, and how does it weight them? If it uses only polls within a certain time frame (say, a week), does it give more weight to a poll that came out the day before? Or is it just a simple average weighted by the number of respondents? And if it uses multiple polls, does it correctly shrink the margin of error as statistics suggests?
Thanks for the feedback everyone!
@Eric,
Pretty simple – take the MoE at face value, and find a value that fits all the polls and that is closest to an average. If there’s no such value, discard outliers. Iterate until a value can be found.
@Eric,
It uses the polls inputted by the campaign designer, and that fall within a certain range of time. This varies based on which polls are available (if there’s only a poll from a month ago, then the algorithm uses that).
Currently, # of respondents is not a variable included, just MoE.
This is brilliant. Good work
Pretty cool, looking forward to the next update.
Still waiting for the popular vote option to create individual races. Been waiting for forever…
@Doug,
Feedback noted – it’s not at the top of the list right now, since it’s essentially a feature for modding, and we have several features for the core game I want implemented ASAP.