vote share with LMSR

[Sveinung Arnesen]

Hi,

what are your opinions about using Hanson´s logarithmic scoring rule
for predicting vote share outcomes in political elections? I have not
seen anyone doing it, but I guess that is because the binary contracts
are more popular. Just to get it out in the open: I have done it
already, and I thought it worked fine.

Kind regards,
S Arnesen

[David Pennock]

Agreed, LMSR naturally works for vote share or more generally any number
with a fixed minimum and maximum, in this case 0 and 100. Glad to see
you tried it and it works.

[Sveinung Arnesen]

Yes. My concern was that the price on small parties would more easily
get pushed up than down, given the logarithmic algorithm (price
"wants" to 50). For the particular case we were concerned with, we
multiplied the values by 10 "behind the scenes", so that a party
predicted by the traders to get 5% of the votes were actually read by
the program as 50%. In reality we bounded the party´s potential vote
share between 0-10%, but given the political context, this was a
pretty safe gamble to make.

[Sveinung Arnesen]

Sorry, was a bit quick on the send button there. The problem of
pushing the price down had to do with the cost of shorting a contract
at 5% when the trader had to guarantee s/he could buy back the
contract at 100%.

On Sep 24, 11:57 pm, Sveinung Arnesen <sveinung.arne...@isp.uib.no>
wrote:

[David Pennock]

Ok, thanks for the clarification. If I understand correctly, you're
talking about how much money to hold in reserve to cover the trader's
worst possible loss. The true absolute worst possible loss in your
example is $95 per $100 contract but in reality you (the market
operator) felt safe that there was no way such a trader would lose more
than $5 per contract betting against a small party. That helps liquidity
but exposes you to some additional albeit small risk.

thanks,
Dave