Reply to thread

Message: <blockquote data-quote="stinger78" data-source="post: 1082616" data-attributes="member: 6771"><p>It creates a positive feedback loop in the SOS metric. It’s not human bias (necessarily), is algorithmic error which results in bias. When an algorithmic output then becomes input for the next run the risk of a positive feedback loop is created. The SEC had a very good week vs. the ACC early on and earned a lot of higher rankings and strength ratings (though a lot of those ACC teams turned out to be crappy teams). Then as they played each other, the strength outputs were used as inputs and so they got stronger and stronger and stronger SOS ratings, which moved them from 5/6 to 9 teams in the top 25 SOS by end of Feb, just playing with themselves. This SOS metric was a factor in the multiplicity of higher ranked teams and selections. It evened out again somewhat after the NCAAT, back down to 7.</p></blockquote><p></p>