[extropy-chat] silent night

[Harvey Newstrom]

My biggest concern with the Iowa market paper is with their 
methodology.  For multi-day polls, they take the average of all days to 
represent the poll.  They then compare this with the last value on the 
last day for the Iowa market.  That means that earlier predictions from 
the poll are being compared with later predictions from the market.  
Assuming that the markets get more accurate as we get closer to the 
election, the Iowa market has an unfair advantage using this method.

Worse yet, we can test this methodology by running an identical set of 
test data through it.  If we were to give this methodology identical 
data for both polls and the market, we would find it claiming the 
market as the winner in most cases, even though they should be a tie in 
all cases.  We would have to rig our test data to give the polls better 
results than the market to get this methodology to declare them even.  
We would have to rig our test data to give the polls even better 
results than that to get this methodology to declare the polls as 
winning.

In short, their methodology is flawed.  It misrepresents the market as 
beating the polls even when it isn't.

--
Harvey Newstrom <HarveyNewstrom.com>
CISSP, ISSAP, ISSMP, CISA, CISM, IAM, IBMCP, GSEC