College Basketball or Is the GCR Consistent with the Polls?

Welcome back, college sports fans, as we take a look into how “accurate” the GCR is, or put another way, how “accurate” the human polls are. We truly do not mean to demonstrate any hubris by the second half of that sentence. Human polls have been around for decades, much longer than any computer poll, much more so this computer ranking system. For the most part, the polls are pretty close, and like computer algorithms, get better as the season progresses. There are only two faults with the polls and they are somewhat related – it’s all about bias. First, polls have the preseason Top 25. This starts the season with a bias that requires teams who were expected to be highly ranked to lose often or really badly to change the mind of the pollsters and teams were not ranked to win a bunch and really big to inch into the polls. We have written many times about preseason bias and the impact it has on the season. In the women’s game a 20-3 UCONN will still be in the Top 10 while a 20-3 Butler would be in the Top 20 – even though they are in the same conference. The preseason poll not only sets expectations, but the sheer human nature of the PS poll (based on previous season’s results) also creates an “I don’t want to be wrong bias.” For Butler to make the Top 25, a team in the preseason poll would have to drop out of the poll – an admission of incorrectness. This inertia, almost forgiveness to the “good” teams and almost excuses for the “mediocre” teams, is the real fault. The second fault is that in reality, few, if any, of the pollsters actually watch enough of the games to make an informed decision. They look at available information such as records and scores (for the teams they consider worth looking at – certainly not all 360+) and create a list. There are no criteria – no standard way of ranking. By definition, each voter makes his or her choice without explanation.

…..

The GCR has no bias. Every team starts out tied for first or last, depending on your viewpoint. The algorithm looks at whether a team won or not, the quality of that win/loss (opponent’s rating or winning percentage depending on the metric), the quality of that opponent (opponent’s opponents’ winning percentage). If two teams are 20-0, the team with the higher SOS will be ranked higher – and, if the SOS is exceptionally low, may not be in the Top 25 at all (see Miami OH in the Men’s comparison below). The GCR recalculates after each game. Now that we are more than half-way through the season, including the tournaments, major shifts become less likely – not because the GCR has a bias toward that team but rather because the denominator (games played) makes it less likely that a given team’s numbers will shift a great deal. That said, the South Carolina men were ranked 90th before losing to Florida 95-48 at home, which caused them to drop to 103rd. Shifts can happen, but we think everyone can agree, that was an egregiously bad loss which justified the shift. Once the data is input into the algorithm, we simply publish, occasionally try to explain why team A is ranked higher or lower than team B, but we keep our opinions out of the actual numbers. Our human nature is to excuse our favorite teams or the teams we think are better than their records just like a pollster. The difference is that human nature bit has zero impact on the GCR. We really believe the great majority of the pollsters try to get it “right”, whatever that means.

…..

That raised the question to us about how similar the GCR was to the AP and Coaches polls. We looked at both the Men’s and Women’s most recent polls not just for the Top 25 but for any school receiving votes. We then compared the latest GCR numbers (as of the games through January 29). There are some teams in the polls that are not in the same range (Top 33 or Top 37) for the GCR which means the opposite is also true. Both tables are built the same way. The first two columns are the AP Rank and team name, the third is the Coaches rank for the same team, the fourth is the GCR rank for the same team, and the fifth is for notes when a team is unique to the polls or the GCR, OR if a wide discrepancy needs an explanation. An asterisk next to a GCR rank highlights a team with votes in at least one poll, but outside of the range in the GCR. First, the men.

TeamAPCoachGCRNotes
Arizona111
UCONN234
Michigan322
Duke443
Nebraska555
Gonzaga6623SOS 140th
Michigan St787
Iowa St8910
Illinois9109
Houston10713SOS 79th
Texas Tech11116SOS 2nd
Purdue121211
BYU13138SOS 23rd
Kansas141414
Arkansas151618
North Carolina161816
Virginia171717
Vanderbilt181513SOS 41st
Florida192122
Louisville2020316 losses
St Louis212225
Clemson221921
Alabama232320
Miami OH242556*SOS 352nd
St Johns2524
Tennessee262730
Kentucky273239*7 losses
Georgia282632SOS 80th
Iowa292434SOS 75th
Texas A&M302841*SOS 134th
Auburn3126SOS 3rd
NC St3236
SMU333119SOS 7th
St Marys3430283 losses
Utah St35t3342*SOS 144th
Villanova35t2927SOS 40th
Wisconsin3738*similar
Liberty3457*SOS 252nd
UCF1516-4 (SOS 20th)
Virginia Tech2916-6 (SOS 34th)
Stanford3314-7 (SOS 11th)
George Mason3519-2 (SOS 184th)
Miami FL3717-4 (SOS 142nd)

…..

A couple additional notes. Kentucky receiving votes could be simply because they are, well, Kentucky. UCF not being on a single ballot while the GCR has them in the Top 15 could be simply because they are, well, not Kentucky (or similar blue blood).

…..

Now let’s shift to the women where we see a bit more variation.

TeamAPCoachGCRNotes
UCONN111
UCLA222
South Carolina335
Texas443
Vanderbilt564
LSU6519SOS 210th
Louisville7710
Iowa886
Michigan9913
Oklahoma101124 4 losses, SOS 75th
Ohio St1112183 losses, SOS 143rd
TCU121072 losses, SOS 51st
Michigan St131315
Baylor141493 losses, SOS 28th
Tennessee151512
Maryland1616315 losses
Mississippi171714
Kentucky181822
Princeton1919111 loss, SOS 76th
Duke202620
Texas Tech2120163 losses
West Virginia222117SOS 25th
Georgia232532SOS 167th
Alabama242283 losses, SOS 24th
Washington252426
North Carolina262333SOS 96th
Rhode Island2729232 losses, SOS 184th
Nebraska283029
Minnesota2930t44*6 losses, SOS 71st
NC St3027
Illinois313244*6 losses, SOS 98th
Fairfield32102*SOS 350th
Oklahoma St332838*SOS 139th
Iowa St2743*SOS 170th
Arizona St2118-4, SOS 70th
Syracuse2517-4, SOS 77th
Villanova2816-5, SOS 42nd
Utah3015-6, SOS 15th

…..

How Arizona St has not received any votes is very confusing. Utah should have been on at least one ballot. Fairfield could be a “favorite” for a single voter because at 17-4 with their schedule is good and should get them an NCAA appearance (assuming they aren’t upset in the MAAC tourney), but hardly Top 25 level – not with the 14th easiest schedule of any school. The Iowa St women are an example of what we talked about in the opening – they were preseason ranked #14 in the AP poll, played a ridiculously easy schedule going 12-0 including 2-0 in the Big 12 and attaining a Top 10 rank, then dropped 5 straight games. They are 4-1 in conference against teams who are ranked lower than they are in the GCR and 1-4 against better teams. We are certain they are on some ballots because of the early “success.”

…..

That’s all we have for today. Next week, we have 4 posts. Monday – WBB Top 363. Tuesday – WBB Conference Comparisons. Wednesday – MBB Top 365. Thursday – MBB Conference Comparisons. We are always open to suggestions for future analysis from any and all readers. Let’s close with a hearty Happy Birthday to JoJo!

Leave a Reply

Your email address will not be published. Required fields are marked *