Sagrin?
The team that lost at home to NIU he still has ranked in the top ten.
The team that lost at home to the team beaten by NIU he still has in the top 20.
He has VaTech at #48 even though they got beat by a team he has at #83
etc etc etc
I'm starting to think he is not a reliable source.
I'm not an expert in Sagarin, or any other rating system, but I am aware of the core problem. Keep in mind that we have only 2 weeks of data. It takes 5-6 weeks before all the teams are interconnected. If you limit a rating system to just the data, after week 1, you'd have only two groups, ones that won, and ones that lost, with no way to differentiate the teams in the "won" group, especially since the rules for numerical polls say that they aren't supposed to use the final point spread, because using the point spread would encourage teams to run up the score.
Some ratings systems deal with this by not releasing any ratings at all for about 5-6 weeks. Others deal with it by starting the season with a rating system based on preconceived notions, such as last years results, preseason magazine predictions, or whatever. Then, as actual results come in, they blend the results into the the preseason prediction numbers. After 5-6 weeks, the preseason preconceived numbers are no longer a part of the rating system, but yes, it does take fully 5-6 weeks to get to that result.
Sagarin is in the latter group. They used to be in USA Today, who expected a "Sagarin Poll" to be published each week of the season, so they couldn't just sit at wait for enough data for real results. They had to publish numbers every week, even after just 1-2 games. Are they of any value? That's for you to decide. At this point, probably 70% of the number is based on preseason predictions, and 30% is based on the 2 weeks of data that they have. After week 3, it will probably be 50% on real data, 50% preconceived notions. By week six, all numbers will be solely based on actual results.
The point here is, when you look at numerical polls, whether Sagarin, or some other one, don't expect them to be meaningful during the first 5-6 weeks. They simply don't have enough data yet. In time they will, but not yet...
Last Edited: 9/14/2024 8:17:44 AM by L.C.