PhillyGirl
Member
http://www.gobblercountry.com/2013/...rginia-tech-hokies-football-in-week-9-ap-poll
Foreword: This thread on Reddit was the inspiration for this article. When I first read through this thread I was doing my normal thing - upvoting every pro-Virginia Tech argument and downvoting every anti-Virginia Tech argument. However, it occurred to me that instead of blindly following the arguments, I could actually contribute to it. In this post I will examine and break down every computer poll used to formulate the collective one-third of the BCS formula.
The argument: Does LSU, which has two losses to currently unranked opponents, deserve to be ranked above a one-loss Virginia Tech team in the AP poll?
Opening statements: There are some fundamental flaws with all forms of college football polling. The Harris and Associated Press polls are both human polls, and thus are predicated on subjective logic and reason. This really boils down to the classic "eyeball test" to the pollsters. This allows for preconceived biases towards particular teams or conferences. As most college football fans nationally would attest to, there seems to be somewhat of an SEC bias in college football polls. Similarly, the ACC has been viewed in a negative light, largely being painted as a basketball conference. In a perfect world these biases wouldn't have any influence in the makeup of polls. Humans are far from perfect, however, and biases are likely alive and well in the human polls in one form or another.
At the same time, computer polls have some flaws as well. Life is often not black and white. There are usually shades of gray somewhere in the middle in most situations. Sports can be hard to judge by cold, hard numbers. Each of the computer polls have different formulas, which means they will produce different results. These results cannot account for the human element, weather conditions, home field advantages, injuries, etc. It's only possible to program so many variables into one equation. However, computer polls are very valuable when taking averages of power rankings, strength of schedule and other relevant statistics.
It's important to note that none of the computer polls used in this evaluation have any influence in the AP Poll. The computer polls are used solely for averaging purposes for all teams, because most human polls rank only the top-25 teams in the nation. Therefore, to take an accurate reading on the strength of opponents I had to consult the computer polls. I chose to take averages of all six BCS computer polls to reduce polling bias from individual polls or methods, and to circumvent the influence of potential outliers.
In case you are unsure of how the BCS works, here are some brief excerpts from the site BCS Knowhow,
Hypothesis: I believe the computer polls will show Virginia Tech has played a very similar schedule to LSU in terms of overall opponent strength, though should still be ranked behind them despite having one less loss.
Experiment: As I said before, I decided to consult every one of the six computer polls used in the makeup of the BCS formula. After combing through these polls, I listed each opponent that Virginia Tech has played at the time of this post (week nine of the college football season). I then went through every poll and extracted the ranking for every opponent Virginia Tech has played. I averaged the rankings together from all six polls and determined the mean ranking of each opponent. I then broke down the average ranking of all opponents played up until this point in the season, and subsequently broke those averages down again on the basis of "wins," "FBS only," and "FBS wins only." This breakdown helped to get a better understanding of the data. Below is the chart that I assembled using Virginia Tech's schedule.
NOTE: The lower the number, the better the opponent is. Think of it like golf.
Virginia Tech's Schedule Through Week Nine:
<tbody>
</tbody>
The data shows that Virginia Tech has played a relatively easy schedule, with an average opponent ranking of 83.71. However, Virginia Tech's schedule gets significantly more difficult when Western Carolina is removed from the equation, resulting in an average opponent ranking of 50.42. This is significant because the Sagarin and Peter Wolfe polls are the only polls that include FCS-and-below tier schools. By factoring out the only major drag on Virginia Tech's schedule, the lone FCS opponent, Virginia Tech's average opponent is squarely in the top-half of all FBS teams (there are 125 active FBS teams). When looking at Virginia Tech's average FBS-tier wins, the ranking jumps up to 60.07 due to the factoring out of Alabama as a loss. This still puts Virginia Tech's average FBS win in the top-half of all FBS teams, but barely.
Next, I repeated the process explained above for LSU. Below is the chart containing LSU's schedule data through the same time frame.
NOTE: The lower the number, the better the opponent is. Think of it like golf.
LSU's Schedule Through Week Nine:
<tbody>
</tbody>
As the data shows, LSU's average opponent is more difficult than Virginia Tech's. LSU benefits from not having an FCS opponent on their schedule, instead scheduling "easy wins" against lower-tier FBS programs. LSU has an average opponent ranking of 56.58, and an average win ranking of 67.44. LSU's "FBS only" and "FBS wins only" averages are the same as the "average" and "wins only" because there are no FCS schools worth factoring.
Examining The Data: On average, LSU has played a tougher schedule than Virginia Tech. LSU's average opponent has a ranking of 56.58, while Virginia Tech's average opponent has a ranking of 83.71. Similarly, LSU has beaten tougher teams on average. The Tigers have beaten teams with an average ranking of 67.44, while the Hokies have beaten teams with an average ranking of 98.06. Based off strength of schedule alone, LSU deserves to be ranked above the Hokies.
However, things get interesting when Western Carolina is removed from the equation and focus is placed solely on the FBS teams played. The data shows that Virginia Tech has played tougher FBS teams than LSU on average. Virginia Tech's average FBS opponent has a ranking of 50.42 compared to LSU's 56.58. Virginia Tech has also beaten tougher FBS teams on average, with an average FBS win ranking of 60.07. LSU has an average FBS win ranking of 67.44. Based on this examination of strength of schedule alone, Virginia Tech deserves to be ranked above the Tigers.
What can we take from all of this? On the whole, Virginia Tech and LSU have played very similar schedules. On the surface, LSU has played a tolerably tougher schedule than Virginia Tech. However, this is mostly due to the data skewing on the part of Western Carolina, which is a statistical outlier. Virginia Tech has played a tougher FBS schedule than LSU, and has one loss compared to LSU's two losses.
Conclusion: With all of this in mind, the question is posed again: Does LSU deserve to be three spots ahead of Virginia Tech in the AP poll? Probably not, but it's close enough to where I don't have a major issue with it. When we examine the entire body of work from both teams, LSU has statistically played the tougher schedule. They also have better wins than Virginia Tech statistically. Because of this, LSU does deserve to be ranked ahead of Virginia Tech. However, three spots may be a bit much considering LSU's two losses, the value of Virginia Tech's lone loss to a consensus top-3 team in Alabama, and the fact that Virginia Tech has played a tougher FBS schedule than LSU. None of that is really scientific though, and is purely my opinion.
The experiment proved my hypothesis was both right and wrong. As it has been stated, ad nauseam, Virginia Tech did play a weaker overall schedule than LSU by a decent margin. The difference between the average ranking of opponents between the two programs is almost 30 spots. That's pretty significant. It also fails to prove my hypothesis that Virginia Tech and LSU had played similarly difficult schedules. However, when one outlier was removed, the schedules are much closer together. In fact, Virginia Tech has played a tougher schedule than LSU when Western Carolina is removed, but still only by about six spots. While that is a 36 point shift, a very statistically significant figure, it does take select data examination to achieve that result. One cannot simply choose to exclude data because it may affect the outcome in a way the researcher doesn't like. I simply think it's something interesting to note.
Extra Stuff: For fun, I calculated Virginia Tech's average opponent ranking if they played a bad FBS team like UAB instead of an FCS team like Western Carolina. Virginia Tech's average opponent ranking drops from 83.71 with Western Carolina to 60.69 with UAB in their place. UAB would also drop Virginia Tech's average win ranking from 98.06 to 70.45. This move would raise Virginia Tech's "FBS only" and "FBS wins" averages to 70.45 from 50.42 and 60.07 respectively. These averages wouldn't be better than LSU's averages, but the important average (the overall opponent average aka strength of schedule) would be much closer. This is important because the computers really only value the average opponent ranking to calculate strength of schedule. In short, I have concluded it's probably better for teams to schedule bad FBS teams instead of FCS teams. There is slightly more risk involved, but there is more reward. If Virginia Tech, had played UAB instead of Western Carolina, it's probable the Hokies would be ranked ahead of LSU in the BCS standings.
This may not matter too much this year seeing as Virginia Tech is likely out of the national championship picture barring some wild and wacky circumstances. However, for the Hokies and other teams moving forward, this will definitely be something to keep an eye on with the adoption of the BCS Playoff System. It's likely the selection committee will look into strength of schedule to help justify their dialectic. Consider this my personal letter to Jim Weaver urging him to abandon the FCS schools in favor of the Georgia State's and Eastern Michigan's of the world.
Foreword: This thread on Reddit was the inspiration for this article. When I first read through this thread I was doing my normal thing - upvoting every pro-Virginia Tech argument and downvoting every anti-Virginia Tech argument. However, it occurred to me that instead of blindly following the arguments, I could actually contribute to it. In this post I will examine and break down every computer poll used to formulate the collective one-third of the BCS formula.
The argument: Does LSU, which has two losses to currently unranked opponents, deserve to be ranked above a one-loss Virginia Tech team in the AP poll?
Opening statements: There are some fundamental flaws with all forms of college football polling. The Harris and Associated Press polls are both human polls, and thus are predicated on subjective logic and reason. This really boils down to the classic "eyeball test" to the pollsters. This allows for preconceived biases towards particular teams or conferences. As most college football fans nationally would attest to, there seems to be somewhat of an SEC bias in college football polls. Similarly, the ACC has been viewed in a negative light, largely being painted as a basketball conference. In a perfect world these biases wouldn't have any influence in the makeup of polls. Humans are far from perfect, however, and biases are likely alive and well in the human polls in one form or another.
At the same time, computer polls have some flaws as well. Life is often not black and white. There are usually shades of gray somewhere in the middle in most situations. Sports can be hard to judge by cold, hard numbers. Each of the computer polls have different formulas, which means they will produce different results. These results cannot account for the human element, weather conditions, home field advantages, injuries, etc. It's only possible to program so many variables into one equation. However, computer polls are very valuable when taking averages of power rankings, strength of schedule and other relevant statistics.
It's important to note that none of the computer polls used in this evaluation have any influence in the AP Poll. The computer polls are used solely for averaging purposes for all teams, because most human polls rank only the top-25 teams in the nation. Therefore, to take an accurate reading on the strength of opponents I had to consult the computer polls. I chose to take averages of all six BCS computer polls to reduce polling bias from individual polls or methods, and to circumvent the influence of potential outliers.
In case you are unsure of how the BCS works, here are some brief excerpts from the site BCS Knowhow,
The BCS consists of two human elements, or polls from which two-thirds of the BCS formula is taken. The two polls are the Harris Interactive College Football Poll and the USA Today Coaches’ Poll.
...
The third and final part of the BCS calculations is derived from six computer rankings posted weekly. The six computer polls, managed by people, newspapers and the BCS itself are: Anderson & Hester, Richard Billingsley, Colley Matrix, Kenneth Massey, Jeff Sagarin’s USA Today and Peter Wolfe.
...
Again, the computer rankings and BCS polls have no influence on what the AP voters do. This is simply for educational purposes, and curiosity as well. Perhaps it will also dispose of some biases toward the SEC and ACC. I don't yet know the results of the test at the time of my writing this. Plenty remains to be seen....
The third and final part of the BCS calculations is derived from six computer rankings posted weekly. The six computer polls, managed by people, newspapers and the BCS itself are: Anderson & Hester, Richard Billingsley, Colley Matrix, Kenneth Massey, Jeff Sagarin’s USA Today and Peter Wolfe.
...
The BCS aggregates the standings of each of the six computer polls and awards teams 25 points for a first-place standing on a computer poll, 24 for a second, 23 for a third, and so on, just as in the human element, until one point is given for a 25th-place vote.
The BCS takes the six point values and removes the lowest and highest point values given to a specific team. Four point values are left, which the BCS adds together.
Because a perfect score would mean receiving four first-place votes, or four 25 point values, the BCS divides the team’s four computer ranking point values by 100 to arrive at the composite computer ranking value.
...
(Harris Poll % + Coaches’ Poll % + Computer Poll %)/3 = BCS Ranking
Hypothesis: I believe the computer polls will show Virginia Tech has played a very similar schedule to LSU in terms of overall opponent strength, though should still be ranked behind them despite having one less loss.
Experiment: As I said before, I decided to consult every one of the six computer polls used in the makeup of the BCS formula. After combing through these polls, I listed each opponent that Virginia Tech has played at the time of this post (week nine of the college football season). I then went through every poll and extracted the ranking for every opponent Virginia Tech has played. I averaged the rankings together from all six polls and determined the mean ranking of each opponent. I then broke down the average ranking of all opponents played up until this point in the season, and subsequently broke those averages down again on the basis of "wins," "FBS only," and "FBS wins only." This breakdown helped to get a better understanding of the data. Below is the chart that I assembled using Virginia Tech's schedule.
NOTE: The lower the number, the better the opponent is. Think of it like golf.
Virginia Tech's Schedule Through Week Nine:
| Opponent | Sagarin | Peter Wolfe | Anderson & Hester | Richard Billingsley | Massey | Colley Matrix | Average |
| Alabama | 2 | 3 | 2 | 1 | 2 | 3 | 2.16 |
| W. Car | 246 | 321 | N/R | N/R | N/R | N/R | 283.50 |
| East Carolina | 55 | 52 | 57 | 61 | 48 | 52 | 54.17 |
| Marshall | 67 | 76 | 68 | 86 | 62 | 62 | 70.17 |
| Georgia Tech | 46 | 35 | 44 | 44 | 36 | 46 | 41.83 |
| North Carolina | 132 | 73 | 94 | 76 | 76 | 99 | 92.67 |
| Pittsburgh | 33 | 38 | 39 | 62 | 41 | 36 | 41.50 |
| ----- | ----- | ----- | ----- | ----- | ----- | ----- | ----- |
| Average | 83.00 | 85.43 | 50.67 | 55.00 | 44.17 | 49.67 | 83.71 |
| Wins Only | 96.50 | 99.17 | 60.40 | 65.80 | 52.60 | 59.00 | 98.06 |
| FBS Only | 55.83 | 46.17 | 50.67 | 55.00 | 44.17 | 49.67 | 50.42 |
| FBS Wins Only | 66.60 | 54.80 | 60.40 | 65.80 | 52.60 | 59.00 | 60.07 |
<tbody>
</tbody>
The data shows that Virginia Tech has played a relatively easy schedule, with an average opponent ranking of 83.71. However, Virginia Tech's schedule gets significantly more difficult when Western Carolina is removed from the equation, resulting in an average opponent ranking of 50.42. This is significant because the Sagarin and Peter Wolfe polls are the only polls that include FCS-and-below tier schools. By factoring out the only major drag on Virginia Tech's schedule, the lone FCS opponent, Virginia Tech's average opponent is squarely in the top-half of all FBS teams (there are 125 active FBS teams). When looking at Virginia Tech's average FBS-tier wins, the ranking jumps up to 60.07 due to the factoring out of Alabama as a loss. This still puts Virginia Tech's average FBS win in the top-half of all FBS teams, but barely.
Next, I repeated the process explained above for LSU. Below is the chart containing LSU's schedule data through the same time frame.
NOTE: The lower the number, the better the opponent is. Think of it like golf.
LSU's Schedule Through Week Nine:
| Opponent | Sagarin | Peter Wolfe | Anderson & Hester | Richard Billingsley | Massey | Colley Matrix | Average |
| TCU | 84 | 57 | 64 | 51 | 53 | 71 | 63.33 |
| UAB | 168 | 137 | 106 | 111 | 108 | 104 | 122.33 |
| Kent State | 176 | 130 | 108 | 80 | 104 | 105 | 117.17 |
| Auburn | 13 | 10 | 7 | 17 | 12 | 7 | 11.00 |
| Georgia | 31 | 14 | 20 | 25 | 19 | 25 | 22.33 |
| Mississippi State | 97 | 55 | 55 | 50 | 50 | 58 | 60.83 |
| Florida | 39 | 18 | 30 | 32 | 27 | 34 | 30.00 |
| Ole Miss | 25 | 13 | 27 | 29 | 28 | 32 | 25.67 |
| ----- | ----- | ----- | ----- | ----- | ----- | ----- | ----- |
| Average | 79.13 | 54.25 | 52.13 | 49.38 | 50.13 | 54.50 | 56.58 |
| Wins Only | 96.17 | 67.83 | 61.67 | 56.83 | 59.00 | 63.17 | 67.44 |
| FBS Only | 79.13 | 54.25 | 52.13 | 49.38 | 50.13 | 54.50 | 56.58 |
| FBS Wins Only | 96.16 | 67.83 | 61.67 | 56.83 | 59.00 | 63.17 | 67.44 |
<tbody>
</tbody>
As the data shows, LSU's average opponent is more difficult than Virginia Tech's. LSU benefits from not having an FCS opponent on their schedule, instead scheduling "easy wins" against lower-tier FBS programs. LSU has an average opponent ranking of 56.58, and an average win ranking of 67.44. LSU's "FBS only" and "FBS wins only" averages are the same as the "average" and "wins only" because there are no FCS schools worth factoring.
Examining The Data: On average, LSU has played a tougher schedule than Virginia Tech. LSU's average opponent has a ranking of 56.58, while Virginia Tech's average opponent has a ranking of 83.71. Similarly, LSU has beaten tougher teams on average. The Tigers have beaten teams with an average ranking of 67.44, while the Hokies have beaten teams with an average ranking of 98.06. Based off strength of schedule alone, LSU deserves to be ranked above the Hokies.
However, things get interesting when Western Carolina is removed from the equation and focus is placed solely on the FBS teams played. The data shows that Virginia Tech has played tougher FBS teams than LSU on average. Virginia Tech's average FBS opponent has a ranking of 50.42 compared to LSU's 56.58. Virginia Tech has also beaten tougher FBS teams on average, with an average FBS win ranking of 60.07. LSU has an average FBS win ranking of 67.44. Based on this examination of strength of schedule alone, Virginia Tech deserves to be ranked above the Tigers.
What can we take from all of this? On the whole, Virginia Tech and LSU have played very similar schedules. On the surface, LSU has played a tolerably tougher schedule than Virginia Tech. However, this is mostly due to the data skewing on the part of Western Carolina, which is a statistical outlier. Virginia Tech has played a tougher FBS schedule than LSU, and has one loss compared to LSU's two losses.
Conclusion: With all of this in mind, the question is posed again: Does LSU deserve to be three spots ahead of Virginia Tech in the AP poll? Probably not, but it's close enough to where I don't have a major issue with it. When we examine the entire body of work from both teams, LSU has statistically played the tougher schedule. They also have better wins than Virginia Tech statistically. Because of this, LSU does deserve to be ranked ahead of Virginia Tech. However, three spots may be a bit much considering LSU's two losses, the value of Virginia Tech's lone loss to a consensus top-3 team in Alabama, and the fact that Virginia Tech has played a tougher FBS schedule than LSU. None of that is really scientific though, and is purely my opinion.
The experiment proved my hypothesis was both right and wrong. As it has been stated, ad nauseam, Virginia Tech did play a weaker overall schedule than LSU by a decent margin. The difference between the average ranking of opponents between the two programs is almost 30 spots. That's pretty significant. It also fails to prove my hypothesis that Virginia Tech and LSU had played similarly difficult schedules. However, when one outlier was removed, the schedules are much closer together. In fact, Virginia Tech has played a tougher schedule than LSU when Western Carolina is removed, but still only by about six spots. While that is a 36 point shift, a very statistically significant figure, it does take select data examination to achieve that result. One cannot simply choose to exclude data because it may affect the outcome in a way the researcher doesn't like. I simply think it's something interesting to note.
Extra Stuff: For fun, I calculated Virginia Tech's average opponent ranking if they played a bad FBS team like UAB instead of an FCS team like Western Carolina. Virginia Tech's average opponent ranking drops from 83.71 with Western Carolina to 60.69 with UAB in their place. UAB would also drop Virginia Tech's average win ranking from 98.06 to 70.45. This move would raise Virginia Tech's "FBS only" and "FBS wins" averages to 70.45 from 50.42 and 60.07 respectively. These averages wouldn't be better than LSU's averages, but the important average (the overall opponent average aka strength of schedule) would be much closer. This is important because the computers really only value the average opponent ranking to calculate strength of schedule. In short, I have concluded it's probably better for teams to schedule bad FBS teams instead of FCS teams. There is slightly more risk involved, but there is more reward. If Virginia Tech, had played UAB instead of Western Carolina, it's probable the Hokies would be ranked ahead of LSU in the BCS standings.
This may not matter too much this year seeing as Virginia Tech is likely out of the national championship picture barring some wild and wacky circumstances. However, for the Hokies and other teams moving forward, this will definitely be something to keep an eye on with the adoption of the BCS Playoff System. It's likely the selection committee will look into strength of schedule to help justify their dialectic. Consider this my personal letter to Jim Weaver urging him to abandon the FCS schools in favor of the Georgia State's and Eastern Michigan's of the world.
