Tuesday, November 15, 2016

Betting the numbers

Just, the generic issue of betting GDP. I set the rules and say. the initial GDP report is taken to be an integer to the nearest 500 billion, and I round that off to a 8 bit integer, and betters bet the 256 bins.

Now, we expect that better would bet a proabilitity distribution about themean estimate.

How to pay that off?
Say we run the bet and get 1,000 bet on 127 and 1,000 bet on 128; but the number comes in at 123. OK, the bets only  have one bit of precision, and they are two bits away from the result, so 3/4 of the money is mutual independent guesses, betters take it home.  The remainder goes to the closest.

In other words, skip the guesses, pay those with good inside information, penalize those with bad inside information.

No comments: