Decision Model: “Determine the Risk of Meeting a Werewolf” Solutions
Trying to determine the risk of meeting a Werewolf based upon two factors: Phase of the Moon, and Distance from Graveyard. Here are the rules:
Fact Type: Phase of the Moon
Values: New Moon, Half Moon, Three-Quarter Moon, Full Moon
Rule Family: Phase of the Moon Risk
Conditions | Conclusions
New Moon | .01
Half Moon | .25
Three-Quarter Moon | .50
Full Moon | .75Fact Type: Miles from Graveyard
Values: < 5 miles, between {5 miles, 25 miles}, > 25 miles
Rule Family: Miles from Graveyard Risk
Conditions | Conclusions
<5 miles | .75
between {5 miles, 25 miles} | .25
> 25 miles | .01Fact Type: Werewolf Risk
Values: Phase of the Moon, Miles from Graveyard
Rule Family: Werewolf Risk Weights
Conditions | Conclusion
Phase of the Moon | .75
Miles from Graveyard | .5Calculate Werewolf Risk:
(Phase of the Moon weight * Phase of the Moon Risk) + (Miles from Graveyard Risk * Miles from Graveyard Risk) = Werewolf Risk
Proposed By: Brian Bouvier
Solutions:
- OpenRules – submitted by Jacob Feldman
- Sparkling Logic – submitted by Carole-Ann Matignon
- BiZZdesign – submitted by Antonio Plais from CENTUS
- IBM ODM – submitted by Francisco Manso from DECIDE
dmcommunity.org 
I was intrigued by this problem, but slightly suspicious of the structure of the Werewolf Risk Model, which seemed to imply that there are two independent populations of werewolves: “activated by the moon werewolves” and “loitering around graveyards” werewolves. My understanding was that there is in fact only one population of werewolves, whose frequency of observation varies according to phase of moon and proximity to graveyards. In particular, I thought one might expect the Phase of Moon Risk and Miles from Graveyard Risk variables to interact, since moonlight improves visibility, and werewolves can therefore travel farther from their graveyards when the moon is full.
I thought the best way to investigate this was empirically, so I prevailed upon FICO’s Analytics Group to conduct a study and model the results. We selected 224 locations in the US, identified by adding small pseudo-random numbers to the coordinates of FICO offices. This ensured that the locations were random but accessible. Brave volunteers from the Analytics Group then visited these locations over a period of 28 days (eight in each night), to ensure all phases of the moon were covered. Each volunteer was equipped with a night-vision camera, a notebook, and a vial of holy water. At each location the volunteer watched for one hour, beginning at midnight, and recorded the number of werewolf sightings (any sightings before 12:00, or after 1:00, were ignored).
After a few days’ analysis using our most powerful modelling tools, I can now reveal the results. The group tells me the best fit model is described by the following equation: Werewolf Risk = 0.