betting the numbes assumes the winning number is selected at random, so its a useless bat in that it does not collect insider information. But it is simply a matter of collecting bets and building a Huffman encoder/decoder based on probability. The difference being that the precision of the code if fixed to, say 256 number. When all the bets are in, then a number is selected and decoded. As it descends the decoding tree, losing branches pay up and the winning branches collect. The decode can have fewer bits than the encode, allowing winnings to be shared among a descending branch not taken. Finite precision Huffman coding is a good invention, a good place to start.
Now change the selected number to something like the next GDP NowCast value. Now the encoding and decoding graphs are shaped to accommodate the NowCast belief function, the betting collects inside information.
No comments:
Post a Comment