Margin of victory is a bad way to seed teams. In a blowout game, a coach is faced with the choice of blowing out their opponents more to increase the margin of victory and playing the younger backups so they grow faster. Not a good idea at all. 

I have asked Seble a dozen times over the past  2 years to work in a "nitty gritty" report that can be seen throughout the season. It would be similar to the one you see on ESPN's posts in the RPI Bracketology area of their website. The report would be fluid or dynamic changing on a daily basis based on where your opponents RPI currently is unlike the vs Top 25 that shows in the standings( which is pretty much worthless for 95% of teh teams  and shouldn't be used for any seeding or projection logic). Here is a sample of how the report could look with my team King's that made it into the NT on the new system vs 2 teams that just missed the NT. 

TEAM	Record	RPI	RPI Rank	SOS	SOS Rank	Vs Top 25 RPI	Vs 26-50 RPI	Vs  51-100 RPI	VS 101-200 RPI	Vs 201-300	Vs 301+	Away
King's	18-11	0.5827	68	0.5714	30	0-8	1-1	6-1	5-1	5-0	1-0	8-6
Shenandoah	23-5	0.5967	50	0.5127	140	0-2	0-2	2-1	10-0	4-0	7-0	11-3
Hobart WSm	24-4	0.5919	57	0.4989	200	0-0	1-0	3-3	9-1	6-0	5-0	11-2

Margin of Victory should be included to some degree but it shouldn't be a major factor. The percentage of games vs top 100 and winning % vs Top 100( or 200 ) should be a factor considered by the engine when deciding on tournament selections. Teams Playing 40-50% of their games vs teams 200-384 in the RPI  should have very little margin for error and a "bad losses" by any bubble team vs teams with RPI's >100( or 150,200 etc) should be factored in more than margin of victory. If a report like this one was used or factored into the selection process you could make a logical argument King's deserved a spot in the NT over the other 2 teams. 60% of their games were against top 100, with a record of 7-2 vs teams rated 26-100. BTW I don't know if my team deserved to be in but I shouldn't have been a #12 seed I thought was too high. 
Seble - I like the new projection breakdown page. Is their anyway to get the table tweaked to included a dynamic/fluid table that changes on a daily basis with some of the W-L breakdown like in the table above? I think it would help when coaches are trying to compare how teams stack up vs each other. Also, can you make the columns sortable similar to the RPI table page? Thanks Again for the hard work trying to improve things.

Posted by girt25 on 12/4/2011 10:46:00 AM (view original):

Posted by seble on 12/4/2011 9:45:00 AM (view original):
I'll take a closer look at the logic using the Naismith data.  The move away from just using RPI was intentional though.  It's a limited metric of how good a team really is, and can be very misleading.  To me there's a big difference between winning/losing by 20+ points vs. winning/losing by 3 or 4 points.  Maybe there's a little too much weight on opponent strength.

I'll also see what I can do about spreading out teams from the same conference, but that's not easy to do when some conferences have 7 or 8 teams in.

seble, I agree that RPI isn't by far a perfect metric, but I think the early returns are that there is significantly too much emphasis being put on who you play rather than how you actually do. A 7-20 Florida State team made the PIT, basically just as a result of getting beat up all season by their ACC brethren. That's not good and really can't happen.

Right now it seems that you've replaced one issue (over reliance on RPI) with another (over reliance on SOS, and perhaps other things, like "good" losses). And so far, the new issue seems worse than the old one -- at least that one was something that people could understand and jibed semi-reasonably with real life. But in real life, a 7-20 team would be thinking about firing their coach, not about the postseason. Ever.

As far as spreading conference teams out within a bracket, can be challenging, but no conference mates should ever be meeting up in the 2nd round. That never happened before, either.

Until its recent change in format, the real life NIT used to require a minimum .500 record to be eligible. Although no longer true to real-life, maybe it'd be a good idea for WIS to implement a similar requirement, since we don't have human selection committees and are relying exclusively on computer-generated rankings. That might also speed up firings at BCS schools, where coaches are currently able to hang on forever with only PIT appearances despite losing records.

Posted by spottratz on 12/5/2011 11:52:00 AM (view original):
I have asked Seble a dozen times over the past  2 years to work in a "nitty gritty" report that can be seen throughout the season. It would be similar to the one you see on ESPN's posts in the RPI Bracketology area of their website. The report would be fluid or dynamic changing on a daily basis based on where your opponents RPI currently is unlike the vs Top 25 that shows in the standings( which is pretty much worthless for 95% of teh teams  and shouldn't be used for any seeding or projection logic). Here is a sample of how the report could look with my team King's that made it into the NT on the new system vs 2 teams that just missed the NT. 

TEAM	Record	RPI	RPI Rank	SOS	SOS Rank	Vs Top 25 RPI	Vs 26-50 RPI	Vs  51-100 RPI	VS 101-200 RPI	Vs 201-300	Vs 301+	Away
King's	18-11	0.5827	68	0.5714	30	0-8	1-1	6-1	5-1	5-0	1-0	8-6
Shenandoah	23-5	0.5967	50	0.5127	140	0-2	0-2	2-1	10-0	4-0	7-0	11-3
Hobart WSm	24-4	0.5919	57	0.4989	200	0-0	1-0	3-3	9-1	6-0	5-0	11-2

Margin of Victory should be included to some degree but it shouldn't be a major factor. The percentage of games vs top 100 and winning % vs Top 100( or 200 ) should be a factor considered by the engine when deciding on tournament selections. Teams Playing 40-50% of their games vs teams 200-384 in the RPI  should have very little margin for error and a "bad losses" by any bubble team vs teams with RPI's >100( or 150,200 etc) should be factored in more than margin of victory. If a report like this one was used or factored into the selection process you could make a logical argument King's deserved a spot in the NT over the other 2 teams. 60% of their games were against top 100, with a record of 7-2 vs teams rated 26-100. BTW I don't know if my team deserved to be in but I shouldn't have been a #12 seed I thought was too high. 
Seble - I like the new projection breakdown page. Is their anyway to get the table tweaked to included a dynamic/fluid table that changes on a daily basis with some of the W-L breakdown like in the table above? I think it would help when coaches are trying to compare how teams stack up vs each other. Also, can you make the columns sortable similar to the RPI table page? Thanks Again for the hard work trying to improve things.

That DOES change on a daily basis.

Posted by cornfused on 12/5/2011 12:23:00 PM (view original):

Posted by spottratz on 12/5/2011 11:52:00 AM (view original):
I have asked Seble a dozen times over the past  2 years to work in a "nitty gritty" report that can be seen throughout the season. It would be similar to the one you see on ESPN's posts in the RPI Bracketology area of their website. The report would be fluid or dynamic changing on a daily basis based on where your opponents RPI currently is unlike the vs Top 25 that shows in the standings( which is pretty much worthless for 95% of teh teams  and shouldn't be used for any seeding or projection logic). Here is a sample of how the report could look with my team King's that made it into the NT on the new system vs 2 teams that just missed the NT. 

TEAM	Record	RPI	RPI Rank	SOS	SOS Rank	Vs Top 25 RPI	Vs 26-50 RPI	Vs  51-100 RPI	VS 101-200 RPI	Vs 201-300	Vs 301+	Away
King's	18-11	0.5827	68	0.5714	30	0-8	1-1	6-1	5-1	5-0	1-0	8-6
Shenandoah	23-5	0.5967	50	0.5127	140	0-2	0-2	2-1	10-0	4-0	7-0	11-3
Hobart WSm	24-4	0.5919	57	0.4989	200	0-0	1-0	3-3	9-1	6-0	5-0	11-2

Margin of Victory should be included to some degree but it shouldn't be a major factor. The percentage of games vs top 100 and winning % vs Top 100( or 200 ) should be a factor considered by the engine when deciding on tournament selections. Teams Playing 40-50% of their games vs teams 200-384 in the RPI  should have very little margin for error and a "bad losses" by any bubble team vs teams with RPI's >100( or 150,200 etc) should be factored in more than margin of victory. If a report like this one was used or factored into the selection process you could make a logical argument King's deserved a spot in the NT over the other 2 teams. 60% of their games were against top 100, with a record of 7-2 vs teams rated 26-100. BTW I don't know if my team deserved to be in but I shouldn't have been a #12 seed I thought was too high. 
Seble - I like the new projection breakdown page. Is their anyway to get the table tweaked to included a dynamic/fluid table that changes on a daily basis with some of the W-L breakdown like in the table above? I think it would help when coaches are trying to compare how teams stack up vs each other. Also, can you make the columns sortable similar to the RPI table page? Thanks Again for the hard work trying to improve things.

That DOES change on a daily basis.

Cornfused- I know those stats change on a daily basis. I was refering to if  RPI Record VS TOP____ was added I wouldn't want it like the The RECORD VS TOP 25 that is listing in HD standings and a small part of the formula for seeding or top 25 rankings. That is a static number, depends on when you play the team whether they were ranked or not. Record vs TOP 25 Ranked can be a very poor measure of judging some teams. I would rather have Record vs Top 50 or Top 100 RPI listed in the standings. That is a more important measure of a team IMO.  Here is an very simple example of why the VS Top 25 Ranked Teams is not a good measure. 

In 1st game of season Team A plays Team B, an unranked team,  and wins. Team B goes 25-0 rest of the regular season, is ranked #5, with #1 RPI but TEAM A doesn't receive credit for a win vs TOP 25 in the standings page. On the flip side Team A beats Team B 1st game of the season, Team B is ranked #10 at the time, but then ends up tanking going 10-15 for the season should Team A really get credit for beating a Top 25 ranked team?

The main point is that any type of Bracketology Report with Record vs TOP ___ RPI shouldn't be based on the  RPI was when you played them, but what the RPI's are currently. All of a teams opponents RPI rankings at the end of the season would make up vs 1-25, 26-50, 51-100 and so on. If you beat the #1 RPI team in Game #10 in the regular season, and at theend of the season the that team is now #60 RPI, your win vs that team is counted in the report vs #51-100 RPI teams not Top 25 RPI.     

Sean, Margin of victory shouldn't be included at all. Your team, King's, plays in a very tough Freedom Conference with 12 human coaches.  Coaches in Sim-dominated conferences shouldnt' be given extra credit for a large MOV

Posted by alblack56 on 12/5/2011 3:09:00 PM (view original):
Sean, Margin of victory shouldn't be included at all. Your team, King's, plays in a very tough Freedom Conference with 12 human coaches.  Coaches in Sim-dominated conferences shouldnt' be given extra credit for a large MOV

Exactly.  Especially now that seble has increased the value of wins and decreased the value of SOS.

Yeah, in real life you can tell if a game was 50-25 at half and ends up 85-72 due to a buzzer three and a ton of scrubs or if a tight game that is 4 points with 2 minutes to go baloons to 12 due to some good FT shooting and a dunk at the buzzer to cap it off.

In both cases a team lost by 12, and the human viewer knows that in Game #1, there is a WIDE margin between the 2 teams.  In Game #2, it could have gone either way, but the final score doesn't always reflect that in basketball, where the last minute can take 30.

In Naismith D1, 31 of the top 32 seeds are from the Big 6 conferences.   In the just-completed Crum NT, using the old logic, only 25 of the 32 are from the Big 6.  Frankly, I prefer the latter.

This new logic has greatly enhanced the Big 6 at the expense of the mid-majors.

al, you may end up being right, but not so sure that the one-time Naismith sample proves anything. I've seen results like this for awhile in Allen.

Hopefully the tweaks that seble just made to the selection process will even things out (they definitely will, the only question is whether they'll even them out enough).

In last year's NCAA,  26 of the top 32 seeds were from the Big conferences

Posted by alblack56 on 12/5/2011 3:09:00 PM (view original):
Sean, Margin of victory shouldn't be included at all. Your team, King's, plays in a very tough Freedom Conference with 12 human coaches.  Coaches in Sim-dominated conferences shouldnt' be given extra credit for a large MOV

Al, this is just my guess and could be 100% wrong, part of the reasoning for giving margin of victory some weight may keep away from the practice of scheduling 8-10 non-conf road games which in some ways unfairly weights the winning % of that team. If Team A plays 10 road games and wins all of them by 10 points each, their record is considered 14-0 by the road win factor of 1.4%. If Team A Played the same teams at home and went 10-0 winning each game by 20 pts each, their record would be only 6-0, with very little margin for error when it comes to the RPI calculation. In each of those scenarios if the team lost their next 4 games on the road( losing each game by 15 pts ) the 14-0 record would be 14-5.6 or a .7143 winning %. The other scenario with the team playing all home games and a 6-0 record would be 6-5.6 or a .5172 winning % and they lost all 4 games by 1 pt. In scenario 1 the team would be 75-100 spots higher in RPI because of their scheduling of the 10 road games. If there were a way to factor in a team playing close games versus good teams at home vs being blown out by the same team on the road, It is worth a look to see what this does to the selection process.

I like the idea of margin of victory with similar opponents when deciding on bubble teams if it could be incorporated into the selection. I.e. bubble teams Shenandoah lost to Carnegie Mellon by 11, beat Chestnut by 11.  King's beat Carnegie Mellon by 13, beat Chestnut Hill by 25. 

There is very little risk in D3 scheduling 8-10 road games in Non-conf, adding a small factor of margin of victory I think will make a coach think twice before scheduling a powerhouse team to build up SOS and/or Road wins to boost RPI.

I am 100% in agreement with you for margin of victory on teams rated 200+ in RPI shouldn't be factored in at all. Whether a team beats the #350 RPI team by 10 points or 50 points isn't important, it would be important only if they lost to that team. Losing to a team with and RPI >150 or 200 should be a calculation in the projection seeding that counts as almost a double loss for that game.

I don't know how much MOV is being counted by SEBLE but I am reserving the right to change my mind until after I see how it plays out another season or two to see what problems it may cause.
 

 

Posted by girt25 on 12/5/2011 6:08:00 PM (view original):
al, you may end up being right, but not so sure that the one-time Naismith sample proves anything. I've seen results like this for awhile in Allen.

Hopefully the tweaks that seble just made to the selection process will even things out (they definitely will, the only question is whether they'll even them out enough).

For the most recent NTs of the worlds that I play:

Naismith - of the top 32 seeds, only ONE is a mid-major. This is with the new selection logic. Last season, with the old logic, there were EIGHT mid-majors.

Under the old selection logic:

Rupp- 8 mid-majors in the top 32 seeds
Wooden- 3 mid-majors
Smith-  8 mid-majors
Iba-  7 mid-majors
Crum- 7 mid-majors

The new logic is DEFINITELY  geared toward helping the Big 6 conferences and needs to be changed...pronto

Posted by girt25 on 12/5/2011 6:08:00 PM (view original):
al, you may end up being right, but not so sure that the one-time Naismith sample proves anything. I've seen results like this for awhile in Allen.

Hopefully the tweaks that seble just made to the selection process will even things out (they definitely will, the only question is whether they'll even them out enough).

The tweaks that were made allowed mid majors with 40 rpi's to make the PIT. I don't think any change has helped one mid major.

Posted by alblack56 on 12/5/2011 3:09:00 PM (view original):
Sean, Margin of victory shouldn't be included at all. Your team, King's, plays in a very tough Freedom Conference with 12 human coaches.  Coaches in Sim-dominated conferences shouldnt' be given extra credit for a large MOV

Yeah, but it's all relative, and normalized. A huge MOV against a crappy team doesn't count as much as a much smaller MOV against a good team. So they're not getting "extra credit". In a simplified sense, if you're favored by 44, then you've gotta win by 44 just to break even; anything short of that and you're losing ground in the power ratings... which doesn't give you much margin for error to make real headway... I know I rarely cover the spread when I'm a huge favorite like that. If you're favored by 4, then you only need to win by 4 to break even, so to speak. Of if you're a 7-point dog and you play closer than that, you're coming out ahead, win or lose. I think the bigger danger is that teams in good conferences are going to be over-ranked due to playing a bunch of competitive games (and losing) against good teams, rather than teams in weak conferences getting over-ranked due to beating up on crappy teams, and I think the first batch of data bears that out.