Early power ratings are notorious for looking weird and ridiculous, with teams that will never pan out popping up in the top five while obvious favorites linger around #50. This is a well-known phenomenon among makers of power ratings and almost everyone has adapted to it by merging early ratings with pre-season (or last season) scores, so that the first several weeks don't get made fun of. Ken Pomeroy and Jeff Sagarin both used to have very interesting early rankings that were unmarred by prior expectation, and they were a lot of fun. Now both are streamlined and sensible with data this is not garnered from this season's scores alone.
The NCAA's NET rankings got a huge blowback when they were first introduced just a couple weeks into the 2018-2019 season. Ohio State was 6-0 and #1 and the basketball world lost its collective shit, apparently not knowing how much prior years influence our own perspective on things, or how hard it is to start from no information and make sense of the college basketball world after a handful of games. Even people who should know better, like Jerry Palm and Nate Silver, went ballistic on the NET, calling for its destruction. Not that the NET is a great system by any means, but anyone who actually makes power ratings knows what they'll probably look like after 6 basketball games. We had Ohio State #1 too at the time, and we knew that in a few weeks things would start to right themselves.
Our rankings are all no-prior rankings: no pre-season expectations, nothing from last year. Hence, it takes a while before they "make sense." Here is our top ten (Combined rankings of Strength and Success) after a couple of weeks:
# Team record Rating 1. Seton Hall 3-0 44.49 2. Arizona 5-0 40.58 3. Alabama 4-0 37.28 4. Purdue 5-0 35.74 5. Gonzaga 4-0 34.07 6. Kansas 3-0 33.82 7. USC 3-0 33.75 8. Washington St. 4-0 32.95 9. Loyola Chicago 4-0 32.63 10. Duke 5-0 32.30
Some odd choices, wouldn't you say? Seton Hall at #1, when they're just #21 in the AP? #17 Arizona at #2, #10 Alabama at #3, #24 USC at #7, Washington State at #8?
Purdue, Gonzaga, Kansas, and Duke are all in the AP top five so it isn't all weird, but #2 UCLA is missing as is #6 Baylor.
So it goes. Some hits and some misses at this early stage. It takes a while before the scores really tell the story that Gonzaga is a clear #1, not #5. Sometimes the early power ratings are on to something, but usually the small sample is in error.
Our Markov Chain rankings, however, seem to have things sorted out pretty well already:
# Team record Rating 1. Baylor 4-0 51.95 2. Gonzaga 4-0 43.28 3. Kansas 2-0 37.44 4. Purdue 5-0 29.01 5. Duke 5-0 26.60 6. Virginia Tech 5-0 24.81 7. Villanova 3-2 16.52 8. Connecticut 4-0 15.59 9. Arizona 5-0 14.00 10. Alabama 4-0 12.06
#6 Baylor is #1, and the rest of the top 5 is in the AP top 5. #7 Villanova is at #7, and #10 Alabama is at #10.
Arizona is still ranked high, though not at #2, unranked Virginia Tech (#27 by Others Receiving Votes) shows up, and #22 UConn is at #8. But overall it looks pretty solid, especially for such an early ranking.
Still no UCLA though. Maybe the Bruins aren't as great as advertised? They're in the top 25 in both Combined and Markov so they aren't completely dismissed. But AP #17 Arizona ranks in both top tens.
Ultimately the rankings end up pretty close, but it's interesting that it takes the Markov Chain—at least this year—so little time to look "reasonable", to the point of having last year's NCAA tournament winner and runner-up at #1 and #2 while using nothing from last season to get there.