TOP 25

Started by short, July 11, 2008, 10:56:29 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

ADL70

7, 9, 14, and 22 lose and 6 just survives.
SPARTANS...PREPARE FOR GLORY
HA-WOO, HA-WOO, HA-WOO
Think beyond the possible.
Compete, Win, Respect, Unite

redswarm81

Quote from: cwru70 on October 18, 2008, 09:37:46 PM
7, 9, 14, and 22 lose and 6 just survives.

Yeah, no. 9 Capital's loss spoiled my whole bit.  I was going to note that in Week 7, the multiples of 7 lost:


  • 7 UW-Eau Claire;
  • 14 DelVal; and
  • 22 Carleton

thus proving that Carleton had been ranked one position too low.    :D
Irritating SAT-lagging Union undergrads and alums since 1977

CobberFvr

Quote from: redswarm81 on October 18, 2008, 09:50:27 PM
Quote from: cwru70 on October 18, 2008, 09:37:46 PM
7, 9, 14, and 22 lose and 6 just survives.

Yeah, no. 9 Capital's loss spoiled my whole bit.  I was going to note that in Week 7, the multiples of 7 lost:


  • 7 UW-Eau Claire;
  • 14 DelVal; and
  • 22 Carleton

thus proving that Carleton had been ranked one position too low.    :D

Not sure they (Carleton) should have been ranked at all.

retagent

I don't know if it's refreshing or disturbing, that after all the analytical analysis (or at least a reasonable facsimile) over the past few pages, BF comes in and puts his opinion out there without any backup data or explanation. As someone once said, "Opinions are like a$$h0les, everybody has at least one."

K-Mack

Quote from: redswarm81 on October 18, 2008, 10:09:38 AM
Quote from: K-Mack on October 18, 2008, 12:09:03 AM
Quote from: redswarm81 on October 17, 2008, 01:32:29 PM
Still, I don't see how Monmouth has been kept so far away for so long.

Easy. Carroll is not a marquee win anywhere but on your chart, and the Midwest is one of the least-strong conferences, and the numbers don't necessarily account for that. As with teams from other less-strong conferences (see Curry, Trine, etc.), it takes some time and several wins before they creep into the 20s. Generally.

I think the St. Norbert win is the first one that thrust them onto most voters' watch lists.

The numbers are certainly helpful to voters, but it's safe to say a lot of them probably prefer to use common sense to huge charts. Right or wrong, it's likely how it works.

Easy.  Sure.
Quote from: redswarm81 on October 17, 2008, 01:32:29 PMI don't know how to identify or evaluate Marquee Wins, so I included wins against former national champions, pre-season Top 25 ranked teams, and teams who currently have an SoS winning percentage of .800 or higher.

I know (and appreciate!) how to learn the definitions of ironic, coincidental and apropos, but I don't know how Marquee Win is defined.

This is all a heck of a lot of fun, and I understand that it's all based on very inexact science.  However, d3football.com has advertised its Top 25 poll as the result of more exact science than any of the alternatives.

I would guess that the AFCA poll voters believe that they are using common sense.  "Common sense" is more often discussed when it's missing than when it's present, so I don't really know how to recognize when common sense is being used in lieu of huge charts.

However, I do know how to read thoughtful explanations offered by poll voters.  My list was as near as I could get to an objective rendering of Pat's rather thorough explanation of the relative positions of Cortland St. and Case Western after week 5.  Pat mentioned SoS/OWP, Margin of Victory, and Marquee Wins.  Recently it appears that OOWP is implicated, but I'm not sure if OOWP was used in the Week 5 poll.

I can understand the temptation to use perceived relative conference strength, but that would really be a stretch to apply to an undefeated team that has not been in the playoffs recently, e.g. Monmouth.  Sure, you can point out that its conference mates don't do well out of the conference in regular season or playoffs, but that argument can't be applied directly to the undefeated playoff-stranger team in question--not logically, at least. . . . but I understand that to many people, logic is different than common sense.   ;)

I can also understand (sort of) why voters want to look at SoS/OWP and OOWP, but common sense dictates that those statistics are secondary to Winning Percentage.  An 0-5 team could have an OWP/OOWP of .750/550, but does that mean it should be ranked higher than a 5-0 team?

Is Franklin a Marquee win for Trine?  If so, why?  And if Franklin is a Marquee win for Trine, why isn't Carroll a Marquee Win for Monmouth?

My head is starting to spin--and this stuff is a LOT easier to try and figure out than Empire 8 what-ifs.   :D

Well,
Without a doubt polling is an inexact science.

With regard to the remark about D3football.com boosting it's brand of inexactness as more exact, I think that comes from the pollsters' focus on strength of schedule and strong conferences; leading to the theory that not all undefeated records are the same, for instance.

Monmouth has made the playoffs recently, they lost 62-3 to St. John's. I was going to add "but that's beside the point," but it's exactly the point. If you look at the history of results between top MWC teams and top 25 teams, the scores generally aren't even close, although St. Norbert has had a couple good playoff showings.

In that sense, I don't blame people for going beyond the numbers a bit to make sure their ballots accurately reflect how they feel about the strength of the teams they rank. On the other hand, too much subjectivity on conference strength is why I hated the old 16-team playoff system. At least in the 32-team system, every conference champion gets a shot (at least now that the NWC AQ kicks in) so teams aren't prohibited from winning the championship because of the perception of their league's strength. But their top 25 ranking can be affected, and with reason.

QuoteOkay, so it appears that W-L Record, Strength of Schedule, Margin of Victory, and Marquee Wins (bonus for road marquee wins) are useful factors in poll evaluations.

Everything you mentioned is a factor. But there's also the human element, which is on one hand a welcome dose of inexactness. We've seen what numbers by themselves can produce in the BCS, and in Massey Ratings, Laz Index, et. al. Numbers alone can be very helpful, but also can be manipulated, which is why there are usually safeguards for margin of victory, etc., in mathematical ratings.

On the other hand, if you try hard enough, you can find flaws in the mash-up of numbers and human reasoning that constitutes the poll. There are 25 voters, and we probably don't all use the same system, so it's hard to pinpoint sometimes why certain movements in the poll happen. Maybe 11 of us are thinking one way and the other 14 aren't. Its not really possible to know without going way too far in depth.

My personal rule is to have a reason for why each team is where they are relative to someone else. It might not bear out over the course of the season, but at least I can rest comfortably knowing that at the time, I was using something and not randomly slotting teams whereever or on "gut feel" alone.
Former author, Around the Nation ('01-'13)
Managing Editor, Kickoff
Voter, Top 25/Play of the Week/Gagliardi Trophy/Liberty Mutual Coach of the Year
Nastradamus, Triple Take
and one of the two voices behind the sonic #d3fb nerdery that is the ATN Podcast.

redswarm81

Quote from: K-Mack on October 19, 2008, 12:29:28 AM

Without a doubt polling is an inexact science.

Inexact science is one thing.  Art dressed up as science is another.  I'm simply trying to figure out where the science ends and the art begins--and I recognize that it's a grey area, not a bright line.  But I am trying to determine if the grey area extends over the entire field of study.

Quote from: K-Mack on October 19, 2008, 12:29:28 AM
With regard to the remark about D3football.com boosting it's brand of inexactness as more exact, I think that comes from the pollsters' focus on strength of schedule and strong conferences; leading to the theory that not all undefeated records are the same, for instance.

I think I've mentioned before that it's dangerous to focus on strength of schedule ahead of winning percentage.  SoS is necessarily secondary to winning percentage.  If I were to get all statistigeekical about it, I suppose I'd try to describe how W-L records and SoS each have their individual bell curve distributions, and there's considerable overlap, but that the mean "value" of W-L record is higher than the mean "value" of SoS.  By "value" I mean usefulness in gauging relative team strength.

If you assign SoS a higher priority than winning percentage, then a team with a losing record is likely to be ranked higher than a team with a winning record, by virtue of a higher SoS.

(This is math - the average SoS of teams with losing records is necessarily higher than the average SoS of teams with winning records.)

Thus SoS has its place, but its best place is in breaking ties between teams with identical records.

Strong conferences.  If I recall your recent re-evaluation of the conferences, you described a big middle 14 or so conferences whose strengths were so close as to be indistinguishable as a practical matter.  I doubt that's the analysis that poll voters use if they're factoring in conference strength.  Rather, I suspect that they're applying the results of different teams playing in different years, in order to evaluate teams this year.  That might be art, but it ain't pretty.

Quote from: K-Mack on October 19, 2008, 12:29:28 AM

Monmouth has made the playoffs recently, they lost 62-3 to St. John's. I was going to add "but that's beside the point," but it's exactly the point. If you look at the history of results between top MWC teams and top 25 teams, the scores generally aren't even close, although St. Norbert has had a couple good playoff showings.

Here's where I'll frustrate many, including myself, for being so damned literal.

If it is indeed exactly the point today, that four seasons ago St. John's beat Monmouth, then Monmouth should be ranked behind St. John's today.  I don't think anyone would defend such a nonsensical analysis, but what do you know: there's 4-2 St. John's with 9 poll votes, the same total as 6-0 Monmouth.

St. Norbert has had a couple good playoff showings, but Monmouth isn't entitled to St. Norbert's successful record, despite the fact that Monmouth beat St. Norbert this season?

I appreciate your insights, but you've posited contradictions that can't be reconciled.

How Monmouth did against anyone four seasons ago really has nothing to do with how strong Monmouth is today relative to anyone else today.  How non-Monmouth teams in Monmouth's conference performed against Top 25 teams in different years is even less relevant to Monmouth's relative strength today.

Quote from: K-Mack on October 19, 2008, 12:29:28 AM

I don't blame people for going beyond the numbers a bit to make sure their ballots accurately reflect how they feel about the strength of the teams they rank. . . . Everything you mentioned is a factor.  But there's also the human element, which is on one hand a welcome dose of inexactness. . . . at the time, I was using something and not randomly slotting teams whereever or on "gut feel" alone.

I don't blame people either.  I'm just trying to figure out where the"using something" ends and where the feeling starts.  (Wait--"where the feeling starts" . . . I heard that at a rock concert, didn't I?  Foo Fighters?  Rush?  Wyclef Jean?  :P )
Irritating SAT-lagging Union undergrads and alums since 1977

Pat Coleman

Quote from: redswarm81 on October 19, 2008, 01:39:36 AM
Thus SoS has its place, but its best place is in breaking ties between teams with identical records.

I disagree. Similar records, perhaps. Definitely not identical records.

What would Monmouth's record be if it had played UW-La Crosse's schedule? How about if it had played Hardin-Simmons' schedule? Or even St. Norbert's schedule? Would Monmouth have beaten Wartburg the way it beat Loras?

These two facets are so intertwined that I don't think you can generate an accurate poll if you ignore one or the other.

Who would beat whom on a neutral field: That's what we're trying to measure. And 25 voters may well take 25 different paths to fill out their ballot, but the end result is pretty darn good. If you want a 100% subjective ranking, you need a computer. Computers are alright, but I wouldn't want my ranking solely decided by them.
Publisher. Questions? Check our FAQ for D3f, D3h.
Quote from: old 40 on September 25, 2007, 08:23:57 PMLet's discuss (sports) in a positive way, sometimes kidding each other with no disrespect.

redswarm81

Quote from: Pat Coleman on October 19, 2008, 01:53:05 AM
Quote from: redswarm81 on October 19, 2008, 01:39:36 AM
Thus SoS has its place, but its best place is in breaking ties between teams with identical records.

I disagree. Similar records, perhaps. Definitely not identical records.

I said best.  I think that SoS is good1 for breaking ties/evaluating teams with similar records, but I think it's better for breaking ties between teams with identical records.

1I think that head-to-head competition is better than SoS, and I think that common opponent analysis is better than SoS.  I recognize that h2h and common opponents are less available data than SoS.

Quote from: Pat Coleman on October 19, 2008, 01:53:05 AM
Who would beat whom on a neutral field: That's what we're trying to measure. And 25 voters may well take 25 different paths to fill out their ballot, but the end result is pretty darn good. If you want a 100% subjective ranking, you need a computer. Computers are alright, but I wouldn't want my ranking solely decided by them.

I agree.  Who would beat whom on a neutral field.  That's what you're trying to measure.  And yet, the McMillan Paradox occurs every year at some point.  The end result is pretty darned good--not flawless, and some of the flaws can't be explained.

As far as different paths go, you've gone a step farther in explaining your path on occasion, that's great.

I don't want rankings decided solely by computers either, but neither do I want them decided solely by people who are so invested in tradition and history and reputation that they don't have any objective means of explaining their paths.

(I think you meant "if you want a 100% objective ranking, you need a computer.")
Irritating SAT-lagging Union undergrads and alums since 1977

Pat Coleman

Publisher. Questions? Check our FAQ for D3f, D3h.
Quote from: old 40 on September 25, 2007, 08:23:57 PMLet's discuss (sports) in a positive way, sometimes kidding each other with no disrespect.

retagent

Let me ask swarm to consider a team, team "A", that played 6 games against opponents who were all 6 -0 at that point in a season, and had lost to those teams by a margin of 10 points or less. Let's further assume that the teams that those six undefeated teams played were all 5 - 1 at that point in the season. Then consider team "B" that was 6 - 0, but had played teams that were 0 -6, and had a margin of victory of 10 points or less. If then, team "A" faced team "B", who would he pick as a winner. Also, let's say team "A" beat team "B", how would he rank them relative to each other?

jam40jeff

Quote from: Pat Coleman on October 19, 2008, 01:53:05 AM
Quote from: redswarm81 on October 19, 2008, 01:39:36 AM
Thus SoS has its place, but its best place is in breaking ties between teams with identical records.

I disagree. Similar records, perhaps. Definitely not identical records.

What would Monmouth's record be if it had played UW-La Crosse's schedule? How about if it had played Hardin-Simmons' schedule? Or even St. Norbert's schedule? Would Monmouth have beaten Wartburg the way it beat Loras?

These two facets are so intertwined that I don't think you can generate an accurate poll if you ignore one or the other.

Who would beat whom on a neutral field: That's what we're trying to measure. And 25 voters may well take 25 different paths to fill out their ballot, but the end result is pretty darn good. If you want a 100% subjective ranking, you need a computer. Computers are alright, but I wouldn't want my ranking solely decided by them.

I believe you meant 100% objective. :)

Anyways, I think SoS has its place.  It is one of many tools that should be used to determine rankings.  But there are many others, one of which has to be common sense.  Even MoV can't take into account blowouts where the subs played the second half.  Beating a bad team 42-14 where it was 42-0 at halfitme and all starters were pulled and not a pass was thrown in the second half should show more about the strength of a team than a 42-14 win where it was 28-14 in the 4th quarter and two late scores were tacked on by the starters when the team that was behind was playing desperately.  Also, SoS surely has its limitations.  Yes, it means a team has or has not "proven" itself, but a poll should be measuring how good one thinks a team is, not what a team has so far proven.  If Mount Union played a weak schedule would it make them any less of a powerhouse than with their current schedule?  A team is either good or not, regardless of who they have played, and that is where the human judgement must come in.

redswarm81

Quote from: retagent on October 19, 2008, 12:17:43 PM
Let me ask swarm to consider a team, team "A", that played 6 games against opponents who were all 6 -0 at that point in a season, and had lost to those teams by a margin of 10 points or less. Let's further assume that the teams that those six undefeated teams played were all 5 - 1 at that point in the season. Then consider team "B" that was 6 - 0, but had played teams that were 0 -6, and had a margin of victory of 10 points or less. If then, team "A" faced team "B", who would he pick as a winner. Also, let's say team "A" beat team "B", how would he rank them relative to each other?

Ignoring the practical impossibility of your hypothetical situation (and maybe even theoretical impossibility), SoS is a tool.  Tools have proper uses and improper uses.

Milton Friedman once visited India, where construction of a large canal was underway.  The Minister of Public Works proudly showed Milton Friedman the thousands of laborers who were digging the canal with shovels.  The Minister explained that they could use more efficient equipment such as backhoes and skip loaders, but by using shovels, they were able to employ more workers.

Friedman asked "Then why aren't you using spoons?"
Irritating SAT-lagging Union undergrads and alums since 1977

usee

Not to confuse things even further but there are many other things that influence rankings throughout a season. In ranking teams you are trying to "paint the train while its moving". Teams are often not the same during the course of a season. As a Wheaton fan we were reminded of that last year when the team that began the season was completely different (literally) than the one that ended the season. They were ranked #4 and 8-0 but after the 4th week they were without 6 all conference performers (including 7th round draft pick Andy Studebaker). By the time they played their last game of the season they had lost over half of their 22 starters that began the season (only 4 starters remained on offense).

The team that lost to unranked Carthage and vote getting IWU was completely different than the team that beat North Central in week 5, not even close. The team that started the season for Wheaton was the best team they ever had, certainly a top 5 team. The team that finished was not top 25 IMO, but the pollsters didn't know that till they saw results. There was a string of weeks where Wheaton was ranked way too high but were playing teams they could beat easily.

Similarly, teams with young stars are often at their best later in the season when those players have further playing time.

Its an inexact science at best.

K-Mack

I don't know red. I think we're kind of reaching the point where you're mincing just to keep the argument going. Because the longer this back-and-forth gets, the less it seems we're moving toward establishing anything new or answering any questions you might have. Perhaps it might serve you well to ask a few direct questions and I'll respond with direct answers.

Quote from: redswarm81 on October 19, 2008, 01:39:36 AM
I don't blame people either.  I'm just trying to figure out where the "using something" ends and where the feeling starts.

Well, how's it going? Are you figuring anything out? Do you want me to try to speak for the entire process as I view it from afar, or just to how I vote? What have I been saying that's unclear?

Quote from: redswarm81 on October 19, 2008, 01:39:36 AM
I think I've mentioned before that it's dangerous to focus on strength of schedule ahead of winning percentage. 

I don't think anyone has advocated that, yet you keep coming back to it. I certainly don't believe not winning is more important than winning when trying to determine which teams are "best." But I think it's equally dangerous to focus on winning percentage alone as a means of determining which team is best. Using SoS as a factor is helpful in sorting out teams with similar records, especially ones who have no h2h, common opponents or any kind of crossover to speak of. (i.e. Occidental, Curry and Case Western Reserve, for example).

* I was actually thinking about this on my own time this week while compiling my vote. The best way I can explain the effective use of SoS, or at least one use of it, is in the cases of Hardin-Simmons and Montclair State. And UW-Eau Claire and Capital.

In the case of the former two, their losses (by 2 to UMHB and 6 to Cortland State) should not preclude them from being included alongside an undefeated team that hasn't played anyone nearly as strong as UMHB or Cortland. (like the three mentioned above). However, I think when teams begin to get into the territory of their second loss (sometimes third once 9 or 10 games are played) voters stop "excusing" one loss to a powerful team. In the cases of UWEC and Capital, even though their schedules are still stronger than the three teams mentioned, if your winning percentage in those games is not high enough, voters tend not to give you support over a team with a much greater winning percentage.

The opposite argument one must keep in mind is that if a team is undefeated, even against a weak schedule, there is no relative indication of their limits. You might be left with an absence of data instead of conclusive data.

I have to get back to work, so forgive me for being able to touch on these next things only very quickly:

Quote from: usee on October 19, 2008, 03:13:36 PM
Not to confuse things even further but there are many other things that influence rankings throughout a season. In ranking teams you are trying to "paint the train while its moving". Teams are often not the same during the course of a season.

I agree with usee's post almost wholeheartedly, but especially this part. We can only work with the data we have at the time. I assume you understand his post and won't rehash the 'evolving teams over the course of the season' aspect.

But also what happens is sometimes the data doesn't make perfect sense. (you might have mentioned this). Like how come Del Val lost to Iona but beat Wesley who beat Iona. What do you do with that? I'm not a huge home field guy, but some people like to factor that in.

St. John Fisher, Ithaca and Hartwick have played a 1-1 3-way triangle. With their opponents ranging from Mount Union to Lycoming to Salisbury, and then those teams' opponents ranging from Christopher Newport to Capital to whoever, you could work yourself into a frenzy comparing results and their meanings.

And then how much weight to give each individual result, knowing that each team doesn't necessarily play to its potential (up or down) each time out, and that some teams match up better with others.

Injuries. Growth. Different points in the season.

Win Pct., SOS, OOWP.

You know, I'm not sure I care whether the poll is perceived as science or art. I think you're getting a little carried away with your nitpicking because it's understood that a poll is 25 people's (hopefully well-researched) opinions of the "best" teams in the country.

I could very easily nitpick the definition "best."

Look at a computer rating ... if win pct. and (O)OWP are the only factors, you'll likely see the MIAC and CCIW at the top of the polls. In the Massey Ratings, WIAC and OAC teams often seem to dominate the top of the rankings.

And there is some truth to that. If Huntingdon sweeps the SLIAC and Capital loses three OAC games but they somehow had an h2h game at the end of the season, who would you expect to win? Even if Huntingdon is ranked 25th and Capital is unranked?

I agree you can't get carried away with conference strength (although just for the record, you brought up Monmouth making the playoffs in the past, and we weren't talking about the middle bunch of conferences I referenced in my article, we were talking about one of the best, the MIAC, and one of the worst, the MWC, between which there most certainly is a difference). Conference history has been a reliable factor in predicting future results over the past 8-10 years. I can't say to what percentage it produces but it most certainly is relevant. And at least in the case of Pat and I, a lot of our opinions are formed by the eye test -- seeing the level of play within a certain conference, though it's not humanly possible for us or any other pollster to see every team in D3.

I love to make long posts and mince as much as anybody here, but I think we all more or less agree on the basic point.

It's understood (at least by readers informed as yourself) that a poll is a collection of opinions (and some collections have the potential to be more informed than others) and not infallible gospel. I don't even agree with it each week (I didn't like Wesley ahead of Del Val despite their h2h result). I'm lucky I can agree on my ballot when I submit it, but even then I realize it's my best approximation of reality and not necessarily reality itself. Four or five of my top 25 teams lose each week, I am used to being proven wrong by what happens on the field. I take delight in the fact that my poll vote is used only to produce a best guess as to the "best" teams each week and isn't a determining factor in crowning a national champion.

Call it art. Call it science. Call it artscience if you want. I'm not sure it really matters so long as people keep the poll itself in proper perspective.

P.S. What's the McMillan paradox again? I'm not sure I'm comfortable having a term named after me if I can't remember what it means.
Former author, Around the Nation ('01-'13)
Managing Editor, Kickoff
Voter, Top 25/Play of the Week/Gagliardi Trophy/Liberty Mutual Coach of the Year
Nastradamus, Triple Take
and one of the two voices behind the sonic #d3fb nerdery that is the ATN Podcast.

K-Mack

Maybe I'm being a little too blunt, but don't mistake that for a reluctance to be challenged on the merits of my particular voters or an unwillingness to discuss my particular thought process.

But if this ...

Quote from: redswarm81 on October 19, 2008, 01:39:36 AM
Quote from: K-Mack on October 19, 2008, 12:29:28 AM

I don't blame people for going beyond the numbers a bit to make sure their ballots accurately reflect how they feel about the strength of the teams they rank. . . . Everything you mentioned is a factor.  But there's also the human element, which is on one hand a welcome dose of inexactness. . . . at the time, I was using something and not randomly slotting teams whereever or on "gut feel" alone.

I don't blame people either.  I'm just trying to figure out where the"using something" ends and where the feeling starts. 

... is what you're trying to determine, I don't know how I'm supposed to answer that on behalf of everybody. My guess is it varies from voter to voter.

And I'm not sure that's necessarily a bad thing. We've seen what computer ratings alone can do. Helpful but not without flaws either.

I had a problem with conference reputation and SOS reliance being a major factor in barring teams, especially undefeated ones or ones whose losses came to teams rated higher than another team had played, from the playoffs under the old 16-team system. In light of the AQ system, I advocate the use of those factors in the creation of a top 25 poll.
Former author, Around the Nation ('01-'13)
Managing Editor, Kickoff
Voter, Top 25/Play of the Week/Gagliardi Trophy/Liberty Mutual Coach of the Year
Nastradamus, Triple Take
and one of the two voices behind the sonic #d3fb nerdery that is the ATN Podcast.