re: OFFICIAL BCS THREAD - LSU #6

Korkstand - I will send an email and would be happy to do that. Ditto slackster. Do you have an email address I can send?

quote:

---

Korkstand

---

I would imagine, based on our putrid offensive marks, we are not ranked very high in your poll.

quote:

---

I would imagine, based on our putrid offensive marks, we are not ranked very high in your poll.

---

Well I haven't run all the sims yet for this week, but I expect LSU to be pretty high, possibly even #1. The power ranks are a different story, but LSU will still be quite high thanks to the defense.

Here's the rundown:
Each team's power rank score reflects their chance of defeating an "average" team (or, in other words, their predicted win percentage playing every team a few thousand times each).

The earned ranks are then calculated like so:
Team A defeats Team B, Team A "earns" Team B's power rank score, and Team B loses the complement of Team A's score.

Example:
Team A's power rank score is 75%
Team B's power rank score is 55%

Team A wins, earns 55 points
Team B loses, loses 100 - 75 = 25 points

Basically you earn more points for defeating quality teams, and you lose fewer points for losing to quality teams. Likewise, losing to a bad team will really hurt your score.

Each team's average points earned (or lost) per game determines their position in the poll. As you can probably imagine, a team like Boise going undefeated against a bunch of teams with power rank scores in the 20-30 range won't earn them much respect here, like it should be.

Well, I just finished my first computer poll. I borrowed a few tips from the Colley Matrix, mainly the 1/2 tidbit regarding winning percentages. I like how it affected things. Also, I assume all FCS teams are one team and that team is 0-week, so in Week 7, the FCS team is 0-7.

The interworkings of the poll are as follows:

When a team wins, they receive 10 points. A loss is -10. Lets call the +/-10 points "r". Next a team receives 10*Adjusted Winning % (using CM idea). Now I calculate my first, preliminary rating using this formula:

r+r/2+10*AW% = S, S standing for subtotal.

Next, all S scores are added from Week 1-Week K, in this case, k=7. The total is then divided by the number of games a team has played to arrive at an average S score per week. All average S scores are then assigned a percent rank. (Team #1 has Percent Rank score of 1.000, Team #121, has a Percent Rank score of .000, and everyone else falls in between.) This Percent Rank of S, or PR-S, is important for the next ranking. Let's use LSU and North Carolina as an example.

LSU beats North Carolina and receives 10 points, r=10. NC at this time, is 4-2, but adding 1 to their wins and 2 to their total games, their adjusted winning percentage is 5/7, or 62.5%. 10 times this adjusted winning percentage is 6.25. Plugging back into the formula we get LSU's S score for Week 1 as 10+5+6.25, or 21.5. UNC's S score for Week 1 is -10+-5+10*88.9% (LSU's adjusted winning %), or -6.11.

Now, LSU's S score has a percent rank of 98.4% through Week 7. UNC's S score has a percent rank of 85.2%. The percent rank of an opponent's S score is multiplied by 10 and added to prior total, less the original +/- 10. It looks like this:

r/2+10*(AW%)+ 10*PR-S = "C" subtotal

LSU receives an additional 8.52 points for playing UNC, bringing its total to 19.77, and UNC receives 9.84 points for playing LSU, bring its score to 13.73. The result is a new subtotal score. Let's call that total "C".

I then add all "C" subtotals from each week and divide by the number of games played. Teams are once again ranked by their "C" total average and the percent rank of those scores is again computed. LSU's has a PR-C of 97.5% and UNC has a PR-C of 94.2%. This now finds its way into the Week 1 ratings of the LSU win.

I then used those ratings to derive an "O" score, which is as follows: If a team wins, and the opponent has a PR-C score between .0 -.3999, you their added benefit will be 10* the opponent's PR-C score. If a team beats an opponent with a PR-C score from .4 to .7999, the added benefit will be 2+10*PR-C. If a team beats an opponent with a PR-C score .8 or higher, the added benefit will be 3+10*PR-C. If you lose a game to an opponent with a PR-C between .0 and .3999, you have a benefit of -1*(0+10*(1-(PR-C))), for scores between .4 and .7999, you receive -1*(-2+10*(1-(PR-C))), and for scores .8 and higher, you receive -1*(-3+10*(1-(PR-C))). This step up rewards playing significantly tougher teams, whether you win or lose.

In LSU's case, their "O" is 3+10*.942, or 12.42 points, for beating UNC. UNC's "O" is -1*(-3+10(1-.975))), or 2.75.

Now, it all comes together as follows:

Each team has a weekly total that is added up and divided by the number of games played. Teams that do not play a particular week do not get a score. The weekly totals are computed like this:

"C" + "O" + .75 * "S" = Total Weekly Score

or

r/2 + 10 * (AW%) + 10 * PR-S + "O" + .75 * (r + r/2 + 10 * (AW%))

simplified as 1.625 * r + 17.5 * (AW%) + 10 * PR-S + "O"

WHEWWWWWWWWWW

Note: Keep in mind that the rankings are fluid, and LSU's win over UNC will look better as UNC wins, but if UNC loses out, that win will continuously lose value.