The Scumputer is the trademark scumhunting technique of DrippingGoofball. After several significant wagons (either lynching wagons or wagons that came pretty close to lynching) have formed, DrippingGoofball will analyze them and assign ratings to players according to how often they have wagoned. The base assumption is that any significant wagon has some, but not all, of the scum on it, and that the scum are at least randomly distributed.
This analysis method can be beaten by violating its basic assumption. It will perform suboptimally if every member of a scumteam is on a wagon, and perform even worse if every member of a scumteam is off the wagon.
The scumputer works best on Days 3-4-5 of large games.
First, extract all wagons that contain about 20% of the total players. Those are termed MAXWAGONS.
Player A (6): Player B, Player C, Player D, Player E, Player F, Player G
Player B (7): Player A, Player H, Player B, Player C, Player D, Player D, Player F etc.
Prepare a spreadsheet with the following calculations for each MAXWAGON:
- VAR A number players on wagon
- VAR B probability of random player being scum (estimated number of living scum on day of wagon / number of living players on day of wagon)
- VAR C number of scum on wagon if randomly distributed (VAR A x VAR B)
- VAR D number of townies on wagon
- VAR E number of flipped scum on wagon
- VAR F adjusted of scum on wagon if randomly distributed (VAR C - VAR E)
- VAR G number of unknown alignment players (VAR A - VAR D - VAR E)
- VAR H probability of unknown alignment player being scum (VAR F / VAR G)
Next, for each unflipped player, calculate (VAR H * 100) - (VAR B * 100) for each MAXWAGON that they have voted on and add up all values.
Rank the players according to the results. The higher the number, the higher the probability of a player flipping scum.