Ok, faithful fans, we have had our annual wacky Saturday and things were shaken up a bit. I know none of us ever expect a bunch of teenagers to do something erratic or unexpected, but sometimes life just throws these surprises at us. The GCR doesn’t think the craziness is over, but I get ahead of myself.
This is a combined blog because of my personal whacked out schedule so it has a lot in it. First, I’m going to answer a challenge my friend Carl Larsen threw out, then I’m going to talk about my success or lack thereof at picking upsets along with the ones to expect this week (there are a lot more this time), and finally a look into what games to watch this weekend.
Carl asked a wonderful series of questions, as he is wont to do, about recency or trending and the impact on the GCR. Actually, the GCR doesn’t care – by that I mean, as far as the program can understand, all games happen simultaneously, so if a team suffered a bad loss in week 1, it doesn’t recognize that said team has improved since then. On the one hand, there’s a purity in that look. A couple of examples. First, a team is 14-1 and wins the championship and, coincidentally, is first in the final GCR – does it matter if that loss was in week 1, or 6, or 13, or even in their championship game? The GCR says it doesn’t matter – the SOS and Performance Scores take 14 wins by x amount and 1 loss by x amount, and, along with the other factors calculates a Total Score. Done. Pure. But, what if they lost the final game (so did not win the championship) and were the only 1-loss team in the country – the GCR might put them at #1, if, if, if, but the polls would say #2. Another scenario is what if a Power 5 team lost to a bad FCS team in week 2 and then was 9-1 after that? Should the ranking be different if that loss to that bad FCS team happened in the penultimate week if both scenarios end up 10-2? The GCR says no, but not only would the polls have a bigger hit against the team with the late lost (usually), human nature says bowl committees might knock them down a peg as well. The GCR has no “memory” because everything is “now”. Is that fair or accurate? Last scenario – Auburn beat Oregon is the first game this season with both teams ranked in the preseason AP (silly thing – read that Exploration for some choice words about preseason foolishness). Oregon is currently 5-1 and leading the Pac 12. The GCR sees that game as occurring “today”. Should it count as much at the end of the season as the Auburn/Alabama game? If the game was played just before rivalry week, should there be more value on it than in week 1? I really don’t think so – seeing things in the “now” provides an “every game really is important” mentality. That said, Carl’s question about recency or trending is sound? So I dusted off the old algebra skills and came up with a weighting system – untested, certainly, but something I’ll watch for potential future season’s reporting. It did have a couple of interesting moves. See the table below. For the most part, the rankings were within +/- 5 spots with the adjustment.
| Team | GCR | Recency | Difference |
| Arizona | 17 | 33 | -16 |
| Boise State | 20 | 25 | -5 |
| Florida | 3 | 17 | -14 |
| Kansas State | 28 | 12 | 16 |
| Miami | 18 | 47 | -29 |
| Navy | 27 | 13 | 14 |
| Nebraska | 59 | 76 | -17 |
| Washington | 38 | 64 | -26 |
Suffice to say there’s enough interesting data that I’m going to continue to do the analysis and see what it shows.
Switching gears: Last week I picked 8 games for which the GCR and Vegas picked different winners – I’m not following games that the favorite won, but the underdog beat the spread, aka Vegas Upsets. Before I talk about how I did with those. There were 53 games with an FBS team playing and 54 where the highest level was FCS. 2 of those were against non-Division I schools – I don’t calculate those. I calculate a % to win for all games, except those mentioned. I started tracking in categories: first cut FBS and FCS, second cut by decile – the 50%’s, 60%’s etc. and then a total. Here’s that result:
| Decile | FBS | FCS | Total |
| 50.0-59.9% | 15-3 (.833) | 14-8 (.636) | 29-11 (.725) |
| 60.0-69.9% | 11-5 (.688) | 8-8 (.500) | 19-13 (.594) |
| 70.0-79.9% | 6-2 (.750) | 5-4 (.556) | 11-6 (.647) |
| 80.0-89.9% | 4-1 (.800) | 3-2 (.600) | 7-3 (.700) |
| 90.0%+ | 1-0 (1.000) | 4-0 (1.000) | 5-0 (1.000) |
| Total | 39-14 (.735) | 33-19 (.635) | 72-33 (.685) |
The results show the GCR was above expectations in close games (would expect the 50’s to be about 55% right), and below for the 60-80 ranges – it was upset week after all so that makes sense. I’m going to add this bit of data to the analysis going forward, because it’s fun if no other reason, and see how it goes. Using the expected values, the GCR did better than expected in the FBS. A weighted average should have yielded 31.2 wins or .589. The FCS was under as the weighted average should have yielded 36.85 wins or .709 (4 “locks” pulled that weighted average up). Overall 68 wins expected, 72 achieved – not too bad.
But what about the 8 GCR upsets? I got 4 of them App State beat ULL, South Florida beat BYU, Western Kentucky beat Army, and Navy beat Tulsa. I missed Texas State, New Mexico, West Virginia, and Arizona who lost to ULM, Colorado State, Iowa State, and Washington, respectively. None of my losers beat the spread so 4-4 come Vegas time.
| Home | Away | Vegas | GCR% |
| Pittsburgh (33) | Syracuse (51) | Pitt-3.5 | Cuse 54.4% |
| Iowa St (39) | Texas Tech (31) | IASt-7.0 | Tech 55.7% |
| Kent State (49) | Ohio (145) | Ohio-7.5 | State 53.7% |
| TCU (62) | Kansas St (28) | TCU-3.5 | K-State 65.5% |
| Indiana (47) | Maryland (71) | Ind-5.5 | Mary 51.9% |
| UNC (57) | Va Tech (70) | UNC-3.5 | VT 63.1% |
| Oregon (14) | Washington (38) | OU-3.0 | UofW 52.5% |
| Baylor (10) | Oklahoma St (64) | OKSt-3.5 | Baylor 62.0% |
| Rice (196) | UTSA (170) | Rice-4.5 | UTSA 70.7% |
| Army (89) | Ga State (87) | Army-6.5 | GaSt 61.8% |
| W Mich (109) | E Mich (122) | West-9.0 | East 55.6% |
| Texas A&M (43) | Mississippi (88) | TAM-6.5 | Miss 50.7% |
| Arizona (17) | SoCal (52) | Cal-9.5 | Zona 57.0% |
| Air Force (56) | Hawaii (35) | AF-2.5 | Haw 59.3% |
As a reminder, I don’t do this comparison for FCS games. As a more important reminder, send me a comment for any game this week and I will give you the GCR % to win – absolutely free of charge. Just remember I don’t bet on a bunch of teenagers that people expect to behave in any kind of expected way. Take the information I give with that word of warning, please. It’s all fun and games until somebody loses their shirt.
So that’s 14 games that could prove discussion worthy to some degree, but let’s talk about what to watch.
For the pre-Saturday games (by the way, Troy (83) is up 16-10 over South Alabama (186) in the 3rd as I’m typing this. Troy wins 79.3% of the time in this game), there are a key interesting game. It won’t have much impact but UCLA (130) is playing at Stanford (95) Thursday night. Stanford should win (70.5%) as they are in second place (2 losses behind Oregon) and UCLA is just not good, but this is not the Stanford of recent years – they barely beat Oregon State (138) a couple of weeks ago 31-28. Of course they beat Washington (38) last week 23-13 so it really depends on which Stanford shows up. UCLA’s only win was a 67-63 (no overtime) shootout at Washington State (77). The Vegas line is Stanford -4.5. Should be fun.
Saturday’s early games have a bunch of top 50 teams playing against each other but the one that sticks out is Clemson (9) at Louisville (37). In both the GCR and the AP, Clemson just keeps sliding (SOS related). These two teams are now 1-2 in the Atlantic Division and the winner has some power. Louisville gave Wake Forest (25) its first loss of the season is a crazy 62-59 (no overtime) pre-basketball season game. The Cardinals gave up 28 4th quarter points to make it close after leading 31-21 at the half and 45-31 after 3. Clemson trounced Florida State (42) by 31 and the game was never in doubt. Vegas has the Tigers winning by 24.0, but the GCR shows just 60.7%. I think this will be a bit closer than expected. Clemson should win, but don’t expect the polls to reward them.
The mid-afternoon games look really inviting: Indiana (47)/Maryland (71) should be a close one (in the table above). Duke (26) is visiting Virginia (36) and the Cavaliers are favored by 3.5 (GCR UVA 55.3%). These two teams are 1-2 in the Coastal (see above). But the game of the week for this time slot has to go to Oregon (14) at Washington (38). Oregon is undefeated in the Pac 12 and trying to keep some kind of … quality?…meaning?… to this conference. Every game is already playoff because of the early loss to Auburn (7) and the overall weakness of the Pac 12. Vegas likes Oregon -3 and I do think it will be close, but home field advantage let’s the Huskies (52.5%) pull off the upset.
The first night game pits Florida State (42) against Wake Forest (25). Both lost last week and are fighting for respectability. Vegas isn’t sure about the 5-1 Demon Deacons and sets them at -2. I think they will beat that by a lot (74.4%). But the game of the week goes to Michigan (13) at Penn State (4). These teams are a combined 11-1 (Michigan lost to #8 Wisconsin, but did score on them). Penn State is in the middle of at @Iowa (23), Michigan, @ Michigan State (41) stretch. Vegas likes the Nittany Lions -8.5. I think that’s a hair steep with the GCR saying 68.4%. One score sounds closer to expectations here.
That’s it on double up Wednesday. Tune in Sunday for the impact of the 117 games on the slate this weekend. (Troy is up 23-10 now).
Please send comments, questions, challenges, and share with friends – let’s grow the community!! Thanks, G
Okay, I’m really confused now. (I’m looking at the upset prediction table above – the last table.) There seems to be a disconnect between the GCR ranking and the GCR win % prediction probability. About half the time the better team is predicted to win (more often), which makes sense, but why is Syracuse predicted to win 54.4% of the time when they are ranked worse by the GCR? Why is Mississippi predicted to be about even with Texas A&M when they are ranked 45 slots worse than Texas A&M? Indiana vs. Maryland and UNC vs. VA Tech are other mysteries. Also, why is GA State predicted to be beat Army more than 60% of the time when the GCR only has them ranked two slots apart? The lower the GCR number, the better the team is supposed to be, right? And the better team wins more often, right (>50% of the time)? And the wider the disparity in ranking, the higher the win percentage should be, right?
Okay, no to all of the above apparently.
Great observation Carl. There are two potential factors. First, the most straightforward: home field advantage. That gives a team roughly 2%. That is based on regression analysis and is a constant for all games (there just aren’t enough games to calculate specific HFA for each team individually). The second factor is interesting. When I ran the regression the relevant data was the Performance Score. The SOS and Total Scores were not statistically significant to predict games (p-value over .05). So the predictor only uses that number and ignores SOS. Because the overall rankings do include SOS, we sometimes see what looks like a disconnect. Hope that helps, G