George Kahrimanis commented on my paper on the boxer and the wrestler, writing:

First, a matter of presentation. The scenario that accounts for the condition `X=Y’ is implausible as a story. That is, the friend knowing the results but only telling us that X=Y. Imo the impact of your argument will increase if you simply assume a conditional bet: “(1,1) vs. (0,0)” (the other two possible outcomes just void this bet).

The second remark is also about strengthening the impact of your argument, by means of describing some problem context for which the robust Bayes framing is appropriate, and another problem context for which the Dempster framing is appropriate.

If you are interested, I shall show you two examples of problem contexts.

My response:

1. I’d be interested to hear the two examples, of course.

2. Another way of framing the boxer-wrestler paper–perhaps a better way to frame it–would be to line up the two examples:
(a) betting on the boxing/wrestling match
(b) betting on a coin flip.
Each of these is 50/50 bet (assuming complete ignorance about the fight, as discussed in the paper); the puzzle is, how can they be distinguished operationally. (The answer, as discussed in the paper, is by considering what would happen by conditioning on more information.)