Here's the probability distribution for the rate at which individuals hear LDB. i.e., if the peak of the curve is at day 16, you can take that to mean that people, on average, run into LDB once every 16 days (during the game period).

**Prior distribution **

**Posterior distribution **

**Here's the data we have on when folks have struck out**

Name | Date |
---|

**Hold up, shouldn't the rate go up as Christmas appraches?**

Yeah, I think it should. This model trains a single parameter, the rate parameter β of a Gamma distribution. I think maybe it should do a linear regression, finding a positively-sloped line that describes β over time. I guess then we'd be training two parameters, β and dβ/dt. Other ideas more than welcome.

**How are people who won the LDB game (didn't hear it all season) encoded?**

They're not. They're treated as no data. Wah-wahhhhhh. One fairly obvious solution would be to assign them a "default" LDB date sometime after the season, but it seemed really hard to justify the use of any particular date for that.