Pool C

[fantastic50]

I'm not convinced that the NESCAC's lack of a double-round-robin (or something close to it) is quite that big a deal.  Consider that about half the league plays two extra rivalry games against NESCAC opponents (Amherst/Wililams/Wesleyan and Bates/Bowdoin/Colby), and that their tournament has a quarterfinal round.  This means that if Amherst or Williams makes the semifinals, they would have played 14 games against other NESCAC teams.  That's the same number as a UAA team, and only one less than a WIAC team with a first-round bye.

Meanwhile, a team from another 11 team conference like Benedictine will have only played 5 non-conference games instead of 13/15, giving them far less of a chance to build up a gaudy SOS in what is historically a weak conference to begin with.

To me, that is quite a difference.

There are 43 conferences in Division III and you found two that half the NESCAC comes close to? Seems to make the argument in the other direction than where you tried to make it.

With no connections to the NESCAC or the northeast, I have no dog in this fight, but being a numbers guy, I don't think that the numbers lie to the degree that is being suggested.  I picked the WIAC and UAA for comparison both because they have 8 teams (and conference size is a big part of the scheduling format issue), but also because they are powerhouse leagues, both this year and perennially.  This discussion is only relevant for very strong conferences that routinely contend for multiple bids; no one really cares about the format of a large conference who is only sending their tournament champ.

[sac]

Oh gosh, lets take Williams for example.  If the NESCAC played a double-round robin and Williams lost the same games to the same opponents  in a double-round robin they'd potentially have as many as 11 losses, way out of Pool C or ranking contention.

The single round robin makes a huge difference.


With the 11 teams, 10 more conference games for each spreads 55 more losses across the conference.  That's a huge amount of losses they can schedule around with their current format.


Compare the 8th place NESCAC team with 8 total losses 6 in conference to the MIAC's 8th place team with 13 total losses with 10 in conference.  Big, big difference in perception.

[toad22]

Oh gosh, lets take Williams for example.  If the NESCAC played a double-round robin and Williams lost the same games to the same opponents  in a double-round robin they'd potentially have as many as 11 losses, way out of Pool C or ranking contention.

The single round robin makes a huge difference.


With the 11 teams, 10 more conference games for each spreads 55 more losses across the conference.  That's a huge amount of losses they can schedule around with their current format.


Compare the 8th place NESCAC team with 8 total losses 6 in conference to the MIAC's 8th place team with 13 total losses with 10 in conference.  Big, big difference in perception.

The NESCAC is the dominant team in the region. Most of the best teams in the region really are in the NESCAC. If 20 of 24 games in the regular season are played against each other, plus 1-3 league championship games, sure, there will be more losses spread around the league. Going to that system would be stupid, in my opinion. I would be very happy to go to some modified system where teams play say 14-16 league games. The Presidents don't want extra travel and time away from class. It certainly would be possible, but I don't think the powers that be view this as important.

[Gregory Sager]

There are ways around that, though. If the NESCAC presidents are adamantly opposed to long-distance league games during weeknights, and the schedule has to accommodate three teams in Maine and a team in New York's Mohawk Valley, then pair up the teams into travel partners and play a Saturday/Sunday league schedule.

Even the modified system you suggest would be a vast improvement. Create two divisions, with intradivisional double round-robins and interdivisional single round-robins, with a six-team division thus playing 15 NESCAC games apiece and a five-team division playing 14 games apiece. You could put the Maine triad in the five-team division with Tufts and Middlebury, which would cut down on the travel for the Mainers. Problem solved.

I suspect, though, that nobody from the NESCAC really views it as being a problem at all. They're either indifferent to the ramifications of the single round-robin polity (the NESCAC presidents) or they're aware of those ramifications and see them as a competitive advantage that shouldn't be surrendered simply as a gesture of goodwill towards the rest of D3. That's why I'm not holding my breath on the NESCAC abandoning the single round-robin, as maddening as it is for the rest of us who follow D3 basketball.

[Mr. Ypsi]

I'd agree with Greg on the terminology, especially in the current climate as it relates to that word. If I were the mouthpiece for the concept, I'd really choose a better title for it and try to define it more succinctly.

It is the word they have used for two or more years... not one I created.

I understand that. But you're the conduit for the committee's voice and terminology to be heard -- feel free to speak in terms that benefit the audience, aka, the fans. You don't have to call it something just because they do.

... though, at the same time I don't want to change a term they are using around and either add to confusion with different terms OR screw up their intent. I would be happy to chat with them about the term, but not sure I want to change it when they may say it another way. And we have been saying it for two-plus seasons.

Dave, take a cue from Kellyanne Conway - the NESCAC uses "alternative SOS".

[Ryan Scott (Hoops Fan)]

There are ways around that, though. If the NESCAC presidents are adamantly opposed to long-distance league games during weeknights, and the schedule has to accommodate three teams in Maine and a team in New York's Mohawk Valley, then pair up the teams into travel partners and play a Saturday/Sunday league schedule.

Even the modified system you suggest would be a vast improvement. Create two divisions, with intradivisional double round-robins and interdivisional single round-robins, with a six-team division thus playing 15 NESCAC games apiece and a five-team division playing 14 games apiece. You could put the Maine triad in the five-team division with Tufts and Middlebury, which would cut down on the travel for the Mainers. Problem solved.

I suspect, though, that nobody from the NESCAC really views it as being a problem at all. They're either indifferent to the ramifications of the single round-robin polity (the NESCAC presidents) or they're aware of those ramifications and see them as a competitive advantage that shouldn't be surrendered simply as a gesture of goodwill towards the rest of D3. That's why I'm not holding my breath on the NESCAC abandoning the single round-robin, as maddening as it is for the rest of us who follow D3 basketball.

I don't think most NESCAC coaches would be opposed to 15 league games.  From what I gather, they'd like more than 10, but not 20.  I also get the impression the coaches aren't driving that ship.

[fantastic50]

Warning: long, nerdy post!

I analyzed the data regarding Pool C selections over the last four seasons (2013-2016).  This graph shows WP & SOS of teams that were candidates for Pool C, and whether or not they were selected.  Moving up or to the right indicates a stronger resume.

The top left represents teams with good records but soft schedules, and the bottom right includes teams with so-so records versus tough schedules.  We could find lots more mediocre teams that didn't make the field to fill out the bottom and left portions (except for the extreme top-left), but I included only about 40 per season. 


If the image doesn't display above, it's located here: http://imgur.com/a/5UxNp

The solid line is the generic bubble mark (50% shot at getting a Pool C berth), and the dashed lines above and below represent 90% and 10% chances, respectively (without considering the number of qualified bubble teams that year)  The moderately-sloped portions toward the center of the 50% and 90% lines represent the "2 wins [or .080 WP] equals .030 SOS" principle.  Moving toward the edges, the slopes change in ways that represent that the committee values a balanced resume more than one with either a weak SOS (and a great record) or a weak WP (against a tough schedule). 

For any engineers, etc. interested in the gory details, the formula is as follows:
P = SOS - 3/8*(1-WP) - 1.2*max(.720-WP,0) - 1.5*max(.510-SOS,0)
The first two terms are .030 SOS = .080 WP, the next term penalizes WP<.720 and the last penalizes SOS<.510. 
Fitting a single-variable logistic regression model to the outcomes (bid or no bid) yields the probabilities.

The outliers of teams that would appear less qualified but got in were as follows:
Randolph '13 (.714, .520, 4-5), which the model gives only a 5% chance
Bowdoin '14 (.792, .503, 1-3) 10%
Wittenberg '14 (.750, .517, 3-6) 21%

On the flip side, the teams that appear deserving but didn't get in were:
Albright '13 (.769, .546, 3-1) 88%
Buena Vista '13 (.720, .563, 1-2) 87%
Carroll '16 (.800, .530, 0-3) 83%
Thomas More '13 (.846, .512, 1-3) 82%

Overall, it's interesting that there were very few outliers in the last couple of years, perhaps indicating a shift to a more quantitative approach by the national committee.  This model does reasonably well despite not including results versus regionally ranked opponents (and Carroll's lack of a vRRO win last year helps explain why they got left out), although my current one does consider that.

Edited to fix image embed -- you almost had it, but you need to have the actual filename in there, not the page it lives on. Click the image to enlarge./pc

[Flying Dutch Fan]

Warning: long, nerdy post!

I analyzed the data regarding Pool C selections over the last four seasons (2013-2016).  This graph shows WP & SOS of teams that were candidates for Pool C, and whether or not they were selected.  Moving up or to the right indicates a stronger resume.

The top left represents teams with good records but soft schedules, and the bottom right includes teams with so-so records versus tough schedules.  We could find lots more mediocre teams that didn't make the field to fill out the bottom and left portions (except for the extreme top-left), but I included only about 40 per season. 


If the image doesn't display above, it's located here: http://imgur.com/a/5UxNp

The solid line is the generic bubble mark (50% shot at getting a Pool C berth), and the dashed lines above and below represent 90% and 10% chances, respectively (without considering the number of qualified bubble teams that year)  The moderately-sloped portions toward the center of the 50% and 90% lines represent the "2 wins [or .080 WP] equals .030 SOS" principle.  Moving toward the edges, the slopes change in ways that represent that the committee values a balanced resume more than one with either a weak SOS (and a great record) or a weak WP (against a tough schedule). 

For any engineers, etc. interested in the gory details, the formula is as follows:
P = SOS - 3/8*(1-WP) - 1.2*max(.720-WP,0) - 1.5*max(.510-SOS,0)
The first two terms are .030 SOS = .080 WP, the next term penalizes WP<.720 and the last penalizes SOS<.510. 
Fitting a single-variable logistic regression model to the outcomes (bid or no bid) yields the probabilities.

The outliers of teams that would appear less qualified but got in were as follows:
Randolph '13 (.714, .520, 4-5), which the model gives only a 5% chance
Bowdoin '14 (.792, .503, 1-3) 10%
Wittenberg '14 (.750, .517, 3-6) 21%

On the flip side, the teams that appear deserving but didn't get in were:
Albright '13 (.769, .546, 3-1) 88%
Buena Vista '13 (.720, .563, 1-2) 87%
Carroll '16 (.800, .530, 0-3) 83%
Thomas More '13 (.846, .512, 1-3) 82%

Overall, it's interesting that there were very few outliers in the last couple of years, perhaps indicating a shift to a more quantitative approach by the national committee.  This model does reasonably well despite not including results versus regionally ranked opponents (and Carroll's lack of a vRRO win last year helps explain why they got left out), although my current one does consider that.

Long?  Not so bad

Nerdy? Yup

Loved it all +k

[Smitty Oom]

Warning: long, nerdy post!

I analyzed the data regarding Pool C selections over the last four seasons (2013-2016).  This graph shows WP & SOS of teams that were candidates for Pool C, and whether or not they were selected.  Moving up or to the right indicates a stronger resume.

The top left represents teams with good records but soft schedules, and the bottom right includes teams with so-so records versus tough schedules.  We could find lots more mediocre teams that didn't make the field to fill out the bottom and left portions (except for the extreme top-left), but I included only about 40 per season. 


If the image doesn't display above, it's located here: http://imgur.com/a/5UxNp

The solid line is the generic bubble mark (50% shot at getting a Pool C berth), and the dashed lines above and below represent 90% and 10% chances, respectively (without considering the number of qualified bubble teams that year)  The moderately-sloped portions toward the center of the 50% and 90% lines represent the "2 wins [or .080 WP] equals .030 SOS" principle.  Moving toward the edges, the slopes change in ways that represent that the committee values a balanced resume more than one with either a weak SOS (and a great record) or a weak WP (against a tough schedule). 

For any engineers, etc. interested in the gory details, the formula is as follows:
P = SOS - 3/8*(1-WP) - 1.2*max(.720-WP,0) - 1.5*max(.510-SOS,0)
The first two terms are .030 SOS = .080 WP, the next term penalizes WP<.720 and the last penalizes SOS<.510. 
Fitting a single-variable logistic regression model to the outcomes (bid or no bid) yields the probabilities.

The outliers of teams that would appear less qualified but got in were as follows:
Randolph '13 (.714, .520, 4-5), which the model gives only a 5% chance
Bowdoin '14 (.792, .503, 1-3) 10%
Wittenberg '14 (.750, .517, 3-6) 21%

On the flip side, the teams that appear deserving but didn't get in were:
Albright '13 (.769, .546, 3-1) 88%
Buena Vista '13 (.720, .563, 1-2) 87%
Carroll '16 (.800, .530, 0-3) 83%
Thomas More '13 (.846, .512, 1-3) 82%

Overall, it's interesting that there were very few outliers in the last couple of years, perhaps indicating a shift to a more quantitative approach by the national committee.  This model does reasonably well despite not including results versus regionally ranked opponents (and Carroll's lack of a vRRO win last year helps explain why they got left out), although my current one does consider that.

I wasn't following D3 as close during the '13 season but I would be interested to know what the committees reasons were for leaving Albright out and putting Randolph in. Any of you guys know?

Oohh and Fantastic... as a math major I thoroughly enjoyed this post and graph. Well done!

[Bucket]

[quote author=Ryan Scott (Hoops Fan) link=topic=4232.msg1791109#msg1791109 date=1487171470

I don't think most NESCAC coaches would be opposed to 15 league games.  From what I gather, they'd like more than 10, but not 20.  I also get the impression the coaches aren't driving that ship.
[/quote]

Correct and correct.

Someone once floated the idea of an east and west division, with teams playing double round robin within their division and single outside of it. It would create half a dozen more league games, wouldn't necessitate that much more travel, etc. But that was just message board fodder.

There's certainly appeal to the idea, though I also shudder at a western gauntlet of Williams, Middlebury, and Amherst all being in the same division.

[Dave 'd-mac' McHugh]

Warning: long, nerdy post!

I analyzed the data regarding Pool C selections over the last four seasons (2013-2016).  This graph shows WP & SOS of teams that were candidates for Pool C, and whether or not they were selected.  Moving up or to the right indicates a stronger resume.

The top left represents teams with good records but soft schedules, and the bottom right includes teams with so-so records versus tough schedules.  We could find lots more mediocre teams that didn't make the field to fill out the bottom and left portions (except for the extreme top-left), but I included only about 40 per season. 


If the image doesn't display above, it's located here: http://imgur.com/a/5UxNp

The solid line is the generic bubble mark (50% shot at getting a Pool C berth), and the dashed lines above and below represent 90% and 10% chances, respectively (without considering the number of qualified bubble teams that year)  The moderately-sloped portions toward the center of the 50% and 90% lines represent the "2 wins [or .080 WP] equals .030 SOS" principle.  Moving toward the edges, the slopes change in ways that represent that the committee values a balanced resume more than one with either a weak SOS (and a great record) or a weak WP (against a tough schedule). 

For any engineers, etc. interested in the gory details, the formula is as follows:
P = SOS - 3/8*(1-WP) - 1.2*max(.720-WP,0) - 1.5*max(.510-SOS,0)
The first two terms are .030 SOS = .080 WP, the next term penalizes WP<.720 and the last penalizes SOS<.510. 
Fitting a single-variable logistic regression model to the outcomes (bid or no bid) yields the probabilities.

The outliers of teams that would appear less qualified but got in were as follows:
Randolph '13 (.714, .520, 4-5), which the model gives only a 5% chance
Bowdoin '14 (.792, .503, 1-3) 10%
Wittenberg '14 (.750, .517, 3-6) 21%

On the flip side, the teams that appear deserving but didn't get in were:
Albright '13 (.769, .546, 3-1) 88%
Buena Vista '13 (.720, .563, 1-2) 87%
Carroll '16 (.800, .530, 0-3) 83%
Thomas More '13 (.846, .512, 1-3) 82%

Overall, it's interesting that there were very few outliers in the last couple of years, perhaps indicating a shift to a more quantitative approach by the national committee.  This model does reasonably well despite not including results versus regionally ranked opponents (and Carroll's lack of a vRRO win last year helps explain why they got left out), although my current one does consider that.

I wasn't following D3 as close during the '13 season but I would be interested to know what the committees reasons were for leaving Albright out and putting Randolph in. Any of you guys know?

Oohh and Fantastic... as a math major I thoroughly enjoyed this post and graph. Well done!

Randolph got in thanks to the "once ranked, always ranked" criteria. It doesn't show it in the work produced, but Randolph ended up with like 18 games or more of their schedule against ranked opponents. RMC that year had like 22. It was basically the death-nail to "once ranked, always ranked." The committee knew that Randolph was basically getting in via the backdoor, but with the criteria in front of them they didn't feel they had another other choice. I think the same was thought when NESCAC got a fifth team in a few years ago. That's why the committee every year looks at how they are dealing with the data and the Championships Committee as a whole looks at how their committees are dealing with the data and criteria and tweeks are made on an almost yearly basis.

[iwumichigander]

There are ways around that, though. If the NESCAC presidents are adamantly opposed to long-distance league games during weeknights, and the schedule has to accommodate three teams in Maine and a team in New York's Mohawk Valley, then pair up the teams into travel partners and play a Saturday/Sunday league schedule.

Even the modified system you suggest would be a vast improvement. Create two divisions, with intradivisional double round-robins and interdivisional single round-robins, with a six-team division thus playing 15 NESCAC games apiece and a five-team division playing 14 games apiece. You could put the Maine triad in the five-team division with Tufts and Middlebury, which would cut down on the travel for the Mainers. Problem solved.

I suspect, though, that nobody from the NESCAC really views it as being a problem at all. They're either indifferent to the ramifications of the single round-robin polity (the NESCAC presidents) or they're aware of those ramifications and see them as a competitive advantage that shouldn't be surrendered simply as a gesture of goodwill towards the rest of D3. That's why I'm not holding my breath on the NESCAC abandoning the single round-robin, as maddening as it is for the rest of us who follow D3 basketball.

I don't think most NESCAC coaches would be opposed to 15 league games.  From what I gather, they'd like more than 10, but not 20.  I also get the impression the coaches aren't driving that ship.

in a conference with a 46 year history, the coaches are not driving the ship.  The makeup of conference committees includes both a member of athletics and administrators (usually ADs). 
As one who lived in New England, change comes slowly.  As long as the NESAC has success in the NCAA as it has had historically, why should it change? 
A true demographic example, attending my first school budget twin meeting for which a modest increase was being proposed before a packed gym and overflow crowns in cafeteria and classrooms, the town moderator called upon a geriatric resident for who the moderator noted was a fifth generation town resident.  To parapahse the gentleman - 'We have done this (proposal) this way for over two hundred years.  The way we do it is not broken.   So why should we change now because some new people moved into town.  If you new people do not like the way we do things --- Just deal with it or Move Out!  Either works for my family and friends.'

[Pat Coleman]

Regional rankings: http://www.d3hoops.com/notables/2017/02/men-regional-rankings-second

[sac]

Warning: long, nerdy post!

I analyzed the data regarding Pool C selections over the last four seasons (2013-2016).  This graph shows WP & SOS of teams that were candidates for Pool C, and whether or not they were selected.  Moving up or to the right indicates a stronger resume.

The top left represents teams with good records but soft schedules, and the bottom right includes teams with so-so records versus tough schedules.  We could find lots more mediocre teams that didn't make the field to fill out the bottom and left portions (except for the extreme top-left), but I included only about 40 per season. 


If the image doesn't display above, it's located here: http://imgur.com/a/5UxNp

The solid line is the generic bubble mark (50% shot at getting a Pool C berth), and the dashed lines above and below represent 90% and 10% chances, respectively (without considering the number of qualified bubble teams that year)  The moderately-sloped portions toward the center of the 50% and 90% lines represent the "2 wins [or .080 WP] equals .030 SOS" principle.  Moving toward the edges, the slopes change in ways that represent that the committee values a balanced resume more than one with either a weak SOS (and a great record) or a weak WP (against a tough schedule). 

For any engineers, etc. interested in the gory details, the formula is as follows:
P = SOS - 3/8*(1-WP) - 1.2*max(.720-WP,0) - 1.5*max(.510-SOS,0)
The first two terms are .030 SOS = .080 WP, the next term penalizes WP<.720 and the last penalizes SOS<.510. 
Fitting a single-variable logistic regression model to the outcomes (bid or no bid) yields the probabilities.

The outliers of teams that would appear less qualified but got in were as follows:
Randolph '13 (.714, .520, 4-5), which the model gives only a 5% chance
Bowdoin '14 (.792, .503, 1-3) 10%
Wittenberg '14 (.750, .517, 3-6) 21%

On the flip side, the teams that appear deserving but didn't get in were:
Albright '13 (.769, .546, 3-1) 88%
Buena Vista '13 (.720, .563, 1-2) 87%
Carroll '16 (.800, .530, 0-3) 83%
Thomas More '13 (.846, .512, 1-3) 82%

Overall, it's interesting that there were very few outliers in the last couple of years, perhaps indicating a shift to a more quantitative approach by the national committee.  This model does reasonably well despite not including results versus regionally ranked opponents (and Carroll's lack of a vRRO win last year helps explain why they got left out), although my current one does consider that.

and all 9 of those RRO's came from ODAC opponents.  They were chosen over teams that had made the attempt to play other RRO's out of conference.  RRO's were way overemphasized those first couple of years.

[Ryan Scott (Hoops Fan)]

[quote author=Ryan Scott (Hoops Fan) link=topic=4232.msg1791109#msg1791109 date=1487171470

I don't think most NESCAC coaches would be opposed to 15 league games.  From what I gather, they'd like more than 10, but not 20.  I also get the impression the coaches aren't driving that ship.

Correct and correct.

Someone once floated the idea of an east and west division, with teams playing double round robin within their division and single outside of it. It would create half a dozen more league games, wouldn't necessitate that much more travel, etc. But that was just message board fodder.

There's certainly appeal to the idea, though I also shudder at a western gauntlet of Williams, Middlebury, and Amherst all being in the same division.
[/quote]

They'd really need equal numbers for the divisions, though.