Ganch - Spread/ML converter Fix?

Ganchrow · 02-01-08, 09:59 AM

Originally posted by OWNED

http://www.sbrforum.com/Betting+Tool...Converter.aspx

NFL spread of 16.5 at 99.81%

17 @ 102.09%

17' @ 104.36%

18 @ 105.52%

Safe to assume the lower spread conversions are incorrect or does the converter just need to be fixed in regards to the higher numbers?

As I wrote in a PM to another poster:

Originally posted by Ganchrow

To be completely honest I have very little confidence in [the ML/Spread converter] at all. It's probably reasonably accurate for smallish spreads but as the spread increases the accuracy falls off pretty quickly.

The problem is that the utility implicitly assumes that a bijective correspondence exists between spreads and win probability when in reality that's simply not supported by either theory or data. For sufficiently small threads, the difference between a SU win and a spread cover is little more than an issue of score distribution microstructure but as spread increases other factor becomes increasingly more important in determining the relationship.

The truth is that ... while the spread/ml converter ... might be an interesting toy for the sports betting dilettante it lacks sufficient predictive power to be of much use to the professional in all but the most trivial of cases.

moses millsap · 02-01-08, 10:11 AM

Thanks Ganch, I thought you were going to chime in over at the other place when I saw the thread come up concerning this tool, but never saw you post anything, so wanted to get it straight from you. Thanks again.