I think the way to model this is by looking at the probability of a change from a no period to a yes period. Just looking at years rather than at half years because I'm a bit lazy

2000: [no, starting]

2001: no

2002: no 

2003: no

2004: no

2005: no

2006: yes

2007: no

2008: no

2009: no

2010: no

2011: no

2013: yes

2014: no

2015: no

2016: yes

2017: yes

2018: no

2019: no

2020: no

2021: no

2022: no

2023: no

2024: no (so far)

What's the probability that a no switches to a yes, per year

No after a no: IIIII IIII I IIIIII = 16

Yes after a no: III = 3

Probability of a switch: 3/19 = 15.7% per year

  - Approximately: 

    - 1-(1-x)^(12/5) = 3/19 => x = 7% per five months

    - 1-(1-x)^(12/4) = 3/19 => x = 5.6% per 4 months

    - 1-(1-x)^(12/3) = 3/19 => x = 4.2% per 3 months

    - 1-(1-x)^(12/2) = 3/19 => x = 2.8% per 2 months

    - 1-(1-x)^(12/2) = 3/19 => x = 1.4% per 1 months

Probability until december. 

"Baserate": 4.2%

Accounting for the US election => 7%? idk man

Accounting for North Korea wanting something specific, and thinking that it can get the US/China/etc. to give it to them/different context => not particularly => 6%

Probability december - April (5 months)

Baserate: 7%

Proability april-august: 

Baserate:7%

Later:

100-6%-7%-7%

Ignoring GJOpen scoring bug.