Fanblogs.com - Latest Comments in Benchmarks

Re: Benchmarks

Zac — Sat, 23 Aug 2008 20:50:51 -0000

Sorry, Ben, like I said, it wasn't my idea. No genius to see here. Move along. Move along...

Something I forgot to mention a lot farther above was the idea of margin of victory. To some extent, this lends credence to a given team's strength. Once each team within a conference is rated at their respective strength, that can be used to rate a conferences strength as a whole. The big problem with this is it lends to teams already winning a game handily, by unnecessarily running up the score with their starters and key back-ups, instead of providing valuable playing time for those players used less frequently.

Re: Benchmarks

gatorhippy — Thu, 21 Aug 2008 23:38:51 -0000

You have to look at the entire field equally...

The BCS/Non-BCS moniker has to be removed...

While there may never be a playoff...

Unfortunately, it is the only way to come to a factual conclusion of crowning a NC team...

ANY other "system" is just that...

A "system"...

Re: Benchmarks

gatorhippy — Thu, 21 Aug 2008 23:27:13 -0000

Does it though?

Because when we take a peek at opponents' records the tale tells another story...

Re: Benchmarks

Ben Prather — Thu, 21 Aug 2008 21:34:46 -0000

I meant membership in a BCS conference, as we where talking about using record vs teams in BCS conferences as a measure. This is a statement of fact.

Playoffs are never going to happen. Not a field of 16.

Re: Benchmarks

gatorhippy — Thu, 21 Aug 2008 21:26:52 -0000

The BCS is NOT a statement of fact...

It is particularly riddled in opinion...

Two thirds opinion as a matter of fact and even the computer portion is still biased given it is programmable and is simply computing the numbers in the way which it is told...

Dumping it, applying a 16 team playoff which includes the 11 current D1 conference champs and 5 at large squads admitted on the basis of on field performance (win percentage) and then their opponents' performance (win percentage) leaves a factual benchmark from which to work with and quantify achievements...

All while placing the point of argument at one place...

On field results...

Which is where it should be...

Re: Benchmarks

Ben Prather — Wed, 20 Aug 2008 17:43:43 -0000

NOTE:

My presentation may be misleading.

I am more interested in conference comparison criterion than team comparisons.

Re: Benchmarks

Ben Prather — Wed, 20 Aug 2008 17:41:22 -0000

Not if they are playing patsies.

A comparison between the strength of away vs home teams definitely give a measure of a teams bravado, especially in OOC games. Fresno would be hard to beat in this stat.

The basic question always comes back to how do you balance difficulty of schedule with the win loss record of against those schedules when the gap between is wide? No matter what game set you limit your analysis to this basic problem arises.

Re: Benchmarks

Porcine — Wed, 20 Aug 2008 16:12:44 -0000

Many a times I wanted to see one.

Re: Benchmarks

aVoiceInSC — Wed, 20 Aug 2008 08:27:24 -0000

How about factoring in road records. It seems that some certain teams don't play many difficult games outside of their home stadium at all. How a team fairs at someone else's house should be a great indicator of their strength.

Re: Benchmarks

aVoiceInSC — Wed, 20 Aug 2008 08:19:22 -0000

Wouldn't that be nice? Keep the dream alive, man. Someone certainly needs to.

Re: Benchmarks

Ben Prather — Wed, 20 Aug 2008 00:33:20 -0000

I have already expressed my prefference for Massey's College Football Comparison. Why not combine it with Associate Professor Kuan Chen's grade policy for Thermodynamics. The grades divisions are determined by where the gaps form in the distribution of results. The four largest divisions separate A, B, C, D, and E teams. FCS teams count as a W. To go with the Utah grading system A's will be broken down to A and A-. B, C and D will have + and -. The divisions are determined by the largest gaps in that grade range. An E is an E.

Expect me to clarify this using the current rankings HERE soon.

Each conference could be measured by its GPA.
Each schedule could be measured by its GPA.

Each conference could be measured by its performance vs each grade of team.

Grades using a broad based consensus, mostly computer based, work better than simple BCS membership and wins to compare collections of teams.

Re: Benchmarks

Ben Prather — Wed, 20 Aug 2008 00:07:44 -0000

Conference based all star/sr games

Your a genius!

Re: Benchmarks

Zac — Tue, 19 Aug 2008 21:25:35 -0000

This is not my idea; someone else proposed this many, many threads ago, but in fact it may well work, not to mention extend the season a little.

Many of us want some sort of play-off (Relax, Regan; hear me out this once.). We also tend to debate which is the better conference. Given each conference at the end of each regular season announces its 1st, 2nd, and 3rd teams, why not have a play-off involving all-conference teams?

Think about it. There are 11 conferences. Rank em; take the Top-8; play em off. The NC is played January 8th. The 1st round could be played January 24th; the 2nd round could be played January 31, and the final could be played February 7th. That's right! After the Super Bowl!!! It would give us die-hards something a little more interesting than the N/S, Blue/Gray, E/W-Shrine, and Senior Bowls. (Not that I would not watch any of them.) Just a thought.

Re: Benchmarks

Zac — Tue, 19 Aug 2008 20:55:44 -0000

Oh, pooh! Ya thilly...

Re: Benchmarks

Zac — Tue, 19 Aug 2008 20:55:17 -0000

Regan, 1B would be better served if there weren't a restriction on the # of teams a conference could place in BCS bowls. An alternative to consider would be the national ranking following the regular season, as compared with the national ranking after all the bowls are played.

Re: Benchmarks

Regan — Tue, 19 Aug 2008 20:08:15 -0000

TEAMS:

#1: My Teams are Better than Yours.

Re: Benchmarks

Regan — Tue, 19 Aug 2008 20:07:18 -0000

CONFERENCE:

#1: OVERALL BOWL RECORD:
Probably the best measuring stick of a conference’s strength, Bowl Games tend to pair-up teams that are (should be) evenly matched, and the games are at neutral sites. Especially when taken over a five-year period, this is a great way to examine a conferences’ success on the field against others.

#2: BCS RECORD (Over the last 5 years) (2x Benchmarks)

> 1A: BCS RECORD WIN PERCENTAGE:
I believe that the Mandelian model of Conference Strength applies (that being that Conference Strength is cyclical). This being the case, I believe that the record of a conference in the BCS is very relevant, but only in five-year intervals.

Nowadays, with new coaches coming every five years unless they are seen as successful, this seems to be a good “test sample period” for measuring BCS success.

> 1B: NUMBER OF BCS GAMES PLAYED:
Also, the BCS Record could be weighted to reflect the strength of conferences that are able to land two teams in the BCS, since usually the stronger conferences each year are able to do so.

#3: CONF. RECORD VS. BCS-CONF. TEAMS:
Each team gets 4x OOC games/year (outside of the Bowls). Often, one or two are cupcakes, but BCS-Conf. Team vs. BCS-Conf. Team are on the docket as well. These can be mismatches, of course, but each conference has strong and weak teams, so it could and should have a role.

#4: CONF. RECORD VS. TEAMS WITH WINNING RECORDS:
The SEC often plays light OOC schedules because the Conference slate is so tough, whereas the Big East is starting to go for heavier OOC slates for the opposite reason. A Benchmark like this one works across the board, regardless of Conference strength for this reason.

If a team’s overall schedule is light, then they are likely to have fewer Winning-Record Opponents, but have a better Record themselves against their opponents. Right there are two sets of potential qualifying statistics.

#5: NUMBER OF TEAMS IN THE FINAL AP/COACHES POLLS:
Like it or not, human voters after all is said and done are quite good at listing teams in relative strength, and will reject pretenders, especially at this stage. The more teams a conference has in the Top 25 when all is said and done, the stronger it was that year all things considered.

Take this record over the past 5 years for each conference and that should be a good indicator of how strong that conference has been (relative to others) over each of the last five years.

Re: Benchmarks

Ben Prather — Tue, 19 Aug 2008 18:30:46 -0000

Adding to the list is the BCS criterion of conference comparison:

1) Top ranked team for the the last four years.
2) Rank of all teams in each conference in the BCS computers
3) Number of teams for each of the last four years.

2 will be approximated with the z-score method and Massey's College Football Comparison.
1&3 will be averaged over the four years will be used.

Re: Benchmarks

Zac — Tue, 19 Aug 2008 13:39:27 -0000

Kev, you are spot on, when you say "generally speaking". What I've proposed may be better, but there have to be even better ways that that.

Re: Benchmarks

Ben Prather — Tue, 19 Aug 2008 08:47:32 -0000

I had an instructor in college who used gaps in the grades to determine where the grade divisions fell. Could we use such a system to divide the teams into tiers and measure schedules by the level of the tiers present?

Massey's College Football Comparison with gaps of 4 spots in the mean appears to be a reasonable scale. This shows a broad consensus that a gap exists between teams above and below that line.

Re: Benchmarks

Kevin @ Fanblogs — Tue, 19 Aug 2008 07:43:08 -0000

Generally speaking, Zac is correct in that his method would produce more reliable data. Taking any single dataset in a vacuum is never the best idea.

Re: Benchmarks

Zac — Mon, 18 Aug 2008 22:05:33 -0000

"...a conference ranking can be made each year based on the data from that year alone."

That can only be true if there is sufficient empirical data to support the premise. If the teams of a given conference predominantly play teams from two or maybe three conferences, the resulting data only provides a comparison of that conference to the other two or three, not the entire population. The only way your hypothesis can work is if the following occurs:

Assumption: A 12 team conference w/two equal halves; each team plays 4 non-conference games. That's 48 games total. The teams from one half would have to play from top to bottom corresponding teams of four different conferences of like configuration (See assumption.). Each team of the other half plays corresponding teams of other conferences. The resulting data may provide a reasonable comparison of our example against 8 other conferences, provided their configuration is the same. It would not be as easy for the Pac-10 (only 30 non-conference games) to compare itself to the ACC, Big-12, CUSA, MAC, or SEC; it would actually be easier for the Big East (40 non-conference games compared with 48). It's for this reason I recommend a number of years. The additional data is needed to strengthen the premise.

Re: Benchmarks

Ben Prather — Mon, 18 Aug 2008 21:26:27 -0000

Thank you for the thoughts.

I don't plan on a ranking but a collection of data based on the input here.

I disagree with the conference rankings requiring years of data to build. I do agree that BCS eligibility should be expanded over 5 years or more, but a conference ranking can be made each year based on the data from that year alone.

Conference's OOC record against bowl eligible teams is a good idea.

Re: Benchmarks

Zac — Mon, 18 Aug 2008 19:48:38 -0000

The best way to determine the strength of a team or the strength of a given conference is about as subjective as any given poll. It’s a matter of opinion. While it is true, there are parameters which can be used, which help to achieve this end, one usually determines (by one’s own opinion) which parameters should be used and which to discard. Therefore, statistically speaking, the more varied opinions with which one is provided, the more accurate the resulting data becomes. That last statement may seem far-fetched, given the factors which go into an opinion such as a college football ranking poll. Factors such as up-brining/environment, education (HS & College), civic or professional affiliations, spousal preferences, supervisory influences (in addition with the media to which one is exposed or the politics to which one is subjected) will certainly impact one’s opinion. Even something as simple as the location (this great country of ours) within which an individual eventually finds themselves, can impact one’s opinion. Therefore, the best means (not necessarily the only means) with which to counter objectivity is through logical mathematical progression. Which, as we all know, is about as dry as a popcorn fart!!! All the same, I’ll give this a try.

With regards to a final poll, the following are recommended for ranking a given team:

Won/Loss record of team in question (bowl eligible? bowl participant? won/lost?)
Won/Loss record at Home
Won/Loss record Away
Won/Loss record against Ranked teams (Top-40-11 & Top-10, both in final poll)
Aggregate Won/Loss record of that team’s combined schedule
Aggregate Won/Loss record of bowl teams played (based upon which teams on a given team's schedule participated in a bowl at the end of that season)
General Ranking of Schedule (based upon the final ranking of each team on a given team’s schedule)

If one were doing a pre-season ranking, which would be similar to playing darts outside on a windy day, one would have to factor in roster changes due to graduation, due to the incoming class, and coaching change impacts.

Ranking conferences is a little tougher. Historical data must be taken into account. As discussed in a previous thread, not all conferences are good about playing the middle to upper middle of a better conferences bracket. For this to work, one must assume that the existing conference affiliation is now as it was 50 years ago. This allows for the following means, with which to compare conferences head-to-head:

Compare each conference's non-conference won-loss record over 5 consecutive years.*
Compare each conference's bowl record, including elite bowl (BCS) participation and National Championships over 5 consecutive years.*
Compare the Historical Aggregate win percentage of each conference, as they currently exist, as if they existed that way from the beginning (hence the assumption above).

*Where to begin and where to end a given 5 year period is open to opinion. However, the more 5-year-periods used would help to increase accuracy.

Fellas, that's all I'm good for on this...for now.

Re: Benchmarks

Ben Prather — Mon, 18 Aug 2008 17:01:44 -0000

BCS and not BCS is a statement of fact and no one would argue it does not make a difference.

What about conference teams with 10+ wins?
or conference record vs teams with 10+ wins?

For conference rankings number of teams in the top 25 may be appropriate. Top 5 teams could even receive double weight. Using polls adds opinion, but as I don't contribute to the polls it is an impartial measure on my part.