Why you don’t use scores to compare teams

[Jjoey52]

Idaho State beat Nevada
Nevada beat UNLV
UNLV beat Fresno State
Fresno State beat SDSU
SDSU beat Stanford
Stanford beat Notre Dame

So using logic from this Idaho State beats Stanford and Notre Dame.


Sent from my iPad using Tapatalk

[93henfan]

That post is quite transitive.

[kalm]

Idaho State beat Nevada
Nevada beat UNLV
UNLV beat Fresno State
Fresno State beat SDSU
SDSU beat Stanford
Stanford beat Notre Dame

So using logic from this Idaho State beats Stanford and Notre Dame.


Sent from my iPad using Tapatalk

So does Howard.

[JohnStOnge]

You have to use all the scores simultaneously. Then of course you're not going to be 100% correct but you'll get a performance way better than something like flipping a coin. By this point of the season something like the Sagarin ratings "Predictor" ratings use nothing but the set of all margins of victory for the season. It compares them all simultaneously. It's been correct on who the winner would be for 531 of 722 games for which systems could be compared to the performance of the line at http://www.thepredictiontracker.com/ncaaresults.php. Correct 73.5 percent of the time. If you knew nothing at all about any of the teams and were just flipping a coin to flip games you'd know going on that there's a 95% chance that you wouldn't pick winners correctly in more than about 53 percent correct. 383 of 722.

I've also done a thing before where I've just taken teams from a conference and compared each matchup from the season in terms of how each team did as far as average margin of victory over common opponents goes. The team that had the higher average margin won the majority of the time to an extent well beyond what could be accounted for by chance.

[SuperHornet]

93 says it all....