Comments on: Geometry – Day 66 – Review Period

By: Brian

Brian — Mon, 23 Jul 2007 06:00:17 +0000

Dan–I know the regression model that works, I just don’t remember where I learned it from. If offer = banker’s offer, round = round number and average = average of remaining cases, then a regression model of the form offer = b0 + b1*round + b2*offer + b3*round*offer will work really well (R^2 > .9). I can send you some data if you want to try it out.

By: Tony Lucchese

Tony Lucchese — Sun, 22 Jul 2007 22:42:34 +0000

I hate have my analysis trumped so completely, but a search for “regression model deal or no deal” yields this link.

http://econ.ucsd.edu/seminars/0607seminars/thaler_paper.pdf

I haven’t had a chance to test it out yet, but I can’t find any obvious flaws in reasoning.

By: dan

dan — Sat, 21 Jul 2007 18:59:38 +0000

Brian, you’re a tease, man. Hinting that the regression model exists is only gonna keep me searching.

Thanks for the intro, anyway, and the link to your site, which has some fun stuff.

By: Brian

Brian — Sat, 21 Jul 2007 17:18:19 +0000

I used Deal or No Deal as an example when I was a teaching assistant for regression courses. It turns out (and this was not my idea–I found it somewhere on a web site and I can’t find the site any more) that a regression model can predict the banker’s offer very well (R^2 in the high .90s) given an interaction term between the average of the remaining cases (expected value) and the round number (1 = first time the banker makes an offer, 2 = second time banker makes an offer, etc). The R^2 is so high that I think the banker basically uses an equation similar to the regression equation and then just adds in some random noise.

P.S. I was at your OTF presentation this week–a lot of great stuff there (and here on your site)!

By: Tony Lucchese

Tony Lucchese — Mon, 09 Apr 2007 21:55:37 +0000

I’ll see what I can do.

By: dan

dan — Mon, 09 Apr 2007 19:43:37 +0000

Tony, I can’t pull a comment form up on your blog. So, anyway:

I’d love a good probabilistic run-down of an episode of Deal Or No Deal if you felt like providing one. The first episode I saw, I pulled open my laptop at a friends’ house, fired up Excel, and punched in a formula for the suitcase’s expected value. I’d predict what the banker would offer, tell the teevee person when she oughtta take the money and when she should stand, and generally make a nuisance of myself among polite company.

They were good times, but I could never predict the Banker’s strategy with sort of reliability or accuracy. Someone oughtta handle this, in any case.

And fwiw, it’s pronounced “dee why dan.”

By: Tony Lucchese

Tony Lucchese — Sun, 08 Apr 2007 22:40:28 +0000

As long as we’re ‘shipping probability, let’s give them change for a dollar and see what odds our aerial coin-tossers can get.