Bracketology 1/28

[Fightin Choke]

Good catch @triaddukefan and @Plebe.

For fun I looked back at some of their schedules from years past. They first hosted Notre Dame in Dec. of 2013. That was the first year Rueck started to get some momentum going. I'm not quite sure how they booked that game. That was year 4 for him. They had losing records in 2 of his first 3 years and were definitely not a name program. Kudos to McGraw for giving them a chance to host a big time opponent. ND returned to Corvallis again in Nov. 2017. Does that mean there is a return game in South Bend during the '19-'20 season? I would guess they didn't schedule each other this year because of the possibility of meeting in Vancouver.

They seem to be alternating years.

Dec 2013 ND @ Oregon St.

Dec 2015 Oregon St @ ND

Nov 2017 ND @ Oregon St.

So maybe next year is Dec 2019 Oregon St @ ND

 

[Fightin Choke]

It's never about raw number of losses. It's all about quality wins, and how they weigh against the bad losses.

Seven losses isn't too many for a 2 seed if most of those losses are to top teams and if they have better wins than the other teams in contention for a 2 seed. That's why I think it might take SC upsetting a Mississippi State or (gulp) UConn to make it happen.

@vowelguy
I looked back over the past 10 years to see if the data support your claim and the data do.

Here are the pre-NCAA tournament losses for 2-seeds over the past 10 years (40 seeds total from 2009-2018 with 4 teams in each of 10 years).

1-loss: 2018
2-loss: 2011, 2013
3-loss: 2009, 2011, 2013, 2014, 2015, 2016
4-loss: 2009, 2010, 2012, 2014, 2014, 2015, 2016, 2016, 2017, 2017, 2018
5-loss: 2009, 2010, 2100, 2011, 2012, 2013, 2015, 2017, 2017
6-loss: 2012, 2014, 2016, 2018, 2018
7-loss: 2009, 2010, 2011, 2013
8-loss: 2012
9-loss: 2015

(Sorry, I was having trouble generating a table.)

Under the scenario I mentioned (9 losses for South Carolina), it seems unlikely but certainly possible for them to make the tournament (as you both suggested). It happened once in 40 seeds (10% of the years).

The mean number of losses for a 2-seed is 4.7 with a standard deviation of 1.7. So I roughly considered anything below 3 losses or above 6 losses to be an outlier, giving us 9/40 instances, which seems fairly high (22.5% total).

Low loss outliers
Baylor 2018: Only had 1 loss but lost in the Sweet 16 (underperformed)
Xavier 2011: Only 2 losses but lost in the 2nd round (vastly underperformed)
Duke 2013: Only 2 losses but lost in the Elite 8 (performed at expectation)

High loss outliers
Kentucky 2015: Had 9 losses and lost in 2nd round (vastly underperformed)
Tennessee 2012: Had 8 losses and lost in the Elite 8 (performed at expectation)
Texas A&M 2009: Had 7 losses and lost in Sweet 16 (underperformed)
Texas A&M 2010: Had 7 losses and lost in 2nd round (vastly underperformed)
Notre Dame 2011: Had 7 losses and lost in NC game (greatly overperformed)
Kentucky 2013: Had 7 losses and lost in the Elite 8 (performed at expectation)

So of the 9 outliers:

1 overperformed
3 performed as expected
5 underperformed

Caveat: I didn't check to see how the middle-of-the-pack 2-seeds performed (i.e., teams with 3-6 losses during the regular season) so I cannot yet say whether the outliers underperformed compared to 2-seeds in general.

 

[Gamecockfan77]

To me elite means you have the talent, the coach, the fan base, and the history, USC fills all those squares and IMHO qualifies as ELITE. That is coming from someone who has not love for USC because they have been a real nemesis for my beloved DAWGS.

Ahh. I'd probably argue that USC's program has entered into the elite programs, if it wasn't already there, with this #1 class full of out of state players being something of a confirmation. I'm sure that would get pushback and it would reduce to what "elite"meant pretty quick.

I would not say this particular team is elite, but it does seem to have improved tremendously and is now fun to watch. I'm interested to see what we do against Kentucky who used to be something of a thorn in the side before A'Ja.

 

[willtalk]

I believe a major problem with the RPI could be corrected with an end of the season adjustment. The problem with the RPI is that for a top team beating a 150 rated team is not much harder than beating one rated 250, but the difference it makes to the RPI is considerable. Neither of those teams have a realistic chance of coming away with a win. So if scores do not matter what actual difference does it make if a team you beat is so far below you to have no impact.

I would suggest that a realistic differential be established for each team at the end of the season. It would then eliminate any team that did not have reasonable chance to beat your team from both the wins and RPI . This would give a more realistic measure of a teams actual applicable strength of schedule. That way any team that was listed as more than say 150 points lower than you, on the RPI scale, would not be counted, unless it results in a loss. The 150 is just a example and would be adjusted based on the statistical history of a lower teams probability of victory against a higher rated team.

This not only would eliminate cupcakes from counting as victories in ratings, but also eliminated them from counting against your RPI score. I believe that was the original purpose of RPI to eliminate the value of cupcake victories.

 

[Gamecockfan77]

I believe a major problem with the RPI could be corrected with an end of the season adjustment. The problem with the RPI is that for a top team beating a 150 rated team is not much harder than beating one rated 250, but the difference it makes to the RPI is considerable. Neither of those teams have a realistic chance of coming away with a win. So if scores do not matter what actual difference does it make if a team you beat is so far below you to have no impact.

I would suggest that a realistic differential be established for each team at the end of the season. It would then eliminate any team that did not have reasonable chance to beat your team from both the wins and RPI . This would give a more realistic measure of a teams actual applicable strength of schedule. That way any team that was listed as more than say 150 points lower than you, on the RPI scale, would not be counted, unless it results in a loss. The 150 is just a example and would be adjusted based on the statistical history of a lower teams probability of victory against a higher rated team.

This not only would eliminate cupcakes from counting as victories in ratings, but also eliminated them from counting against your RPI score. I believe that was the original purpose of RPI to eliminate the value of cupcake victories.

Sure, if the Golden State Warriors played UCONN's schedule they'd get no more credit for their record than UCONN would. But Golden State would (oh no, hope this goes over) certainly be a better basketball team.

That's ultimately the drawback to any rating and the problem is compounded by the fact that the RPI does not take margin of victory into consideration, because the NCAA does not want to incentivize running up the score.

When you remove margin of victory, you pull out an important tool for trying to compare performance versus different strengths of schedule.

But ultimately, there isn't really an objective way of determining who the best team is. You can only use tools that can help you be somewhat consistent in seeding and let the results speak.

 

[cockhrnleghrn]

Bracketology with Charlie Creme

Some highlights and changes..... he has ND moving to the #1 seed in Greensboro. Stanford as the #2 in Albany. Lower Carolina moving to the #4 in Albany.

Creme isn't as well versed as he thinks, since the Gamecocks can't host in Columbia this year - the Men's NCAAT 1st and 2nd rounds will be here. The Gamecocks may be able to host at another arena in South Carolina, but there isn't another one big enough in Columbia.

 

[cockhrnleghrn]

I can hear Dawn saying "Albany again, what do I have to do to get away from them".

I think the Gamecocks' style matches up better against UCONN this year. We'll find out on the 11th.

 

[Gamecockfan77]

I think the Gamecocks' style matches up better against UCONN this year. We'll find out on the 11th.

Eck.

 

[Plebe]

I believe a major problem with the RPI could be corrected with an end of the season adjustment. The problem with the RPI is that for a top team beating a 150 rated team is not much harder than beating one rated 250, but the difference it makes to the RPI is considerable. Neither of those teams have a realistic chance of coming away with a win. So if scores do not matter what actual difference does it make if a team you beat is so far below you to have no impact.

I would suggest that a realistic differential be established for each team at the end of the season. It would then eliminate any team that did not have reasonable chance to beat your team from both the wins and RPI . This would give a more realistic measure of a teams actual applicable strength of schedule. That way any team that was listed as more than say 150 points lower than you, on the RPI scale, would not be counted, unless it results in a loss. The 150 is just a example and would be adjusted based on the statistical history of a lower teams probability of victory against a higher rated team.

This not only would eliminate cupcakes from counting as victories in ratings, but also eliminated them from counting against your RPI score. I believe that was the original purpose of RPI to eliminate the value of cupcake victories.

The idea is a good one, but the devil's in the details. I don't think it'd be tenable to have a ranking system in which some teams' games only count vs. certain opponents and not others.

For example, just throwing some numbers out there, if the #1 team plays #150, you're saying that game won't count. But if #50 plays #150, it does? So where do you arbitrarily draw the line? At the #10 team? #25?

The validity of any computer ranking is partly based on a level-playing-field treatment of all teams in its "universe," without the natural human biases toward a team's reputation, history or conference affiliation. Hard for me to imagine this adjustment not undermining that level field of play. I hear you saying it'd be based on "statistical history of ... probability of victory," but even if that "statistical history" is at or near zero, that guarantees nothing. Stanford was a #1 seed in 1998 and lost to a #16 seed in the first round of the NCAA tournament. "Statistical history" would suggest that such a victory was impossible, yet it happened.

 

[bennyskid]

To be fair to the committee, supposedly they look at "W-L v. Top 100", "W-L v. Top 50", etc.. as much as they look at the raw RPI. The flaws in the RPI are real, but the committee is not totally oblivious to that fact.

 

[willtalk]

The idea is a good one, but the devil's in the details. I don't think it'd be tenable to have a ranking system in which some teams' games only count vs. certain opponents and not others.

For example, just throwing some numbers out there, if the #1 team plays #150, you're saying that game won't count. But if #50 plays #150, it does? So where do you arbitrarily draw the line? At the #10 team? #25?

The validity of any computer ranking is partly based on a level-playing-field treatment of all teams in its "universe," without the natural human biases toward a team's reputation, history or conference affiliation. Hard for me to imagine this adjustment not undermining that level field of play. I hear you saying it'd be based on "statistical history of ... probability of victory," but even if that "statistical history" is at or near zero, that guarantees nothing. Stanford was a #1 seed in 1998 and lost to a #16 seed in the first round of the NCAA tournament. "Statistical history" would suggest that such a victory was impossible, yet it happened.

What I am saying is that the RPI stays as is except the adjustment is used by the selection committee for the purposes of rectification in order to eliminate games that give a false impression of the strength of the team via the won loss record of the teams they played. What it will do is eliminate cup cakes. You know those teams that have no chance in hell of winning. It also will eliminate then as counting as a win while also eliminating the negative impact on the final RPI. Therefor a team could play a tune up or do a favor for a coach without it negatively affecting their end RPI. Many of those really bad D1 teams are no better than D2 teams anyway. Those are not counted as wins by selection committee's right now anyway.

The whole point of the RPI is to give a more rational picture to a teams won/lost record. They have always eliminated non D1 schools from their won lost record in respect to bracketing any way. All this does is to make the same sort of adjustment in respect to lower D1 teams that are also not competitive enough to deserve to be counted as a win. Any team in the top ten should have any team rated lower than 200 even count as a win. The bracketing committee does this anyway by punishing teams for weak schedules. What this will do is create a standard system for do so. It will eliminate those games both in wins and RPI handicapping. It does not eliminate RPI . it is still used but this would be used to create a more consistent measure of strength of schedule. You obviously have to have the RPI to begin with.

The whole point is to eliminate cupcakes from a teams won lost record while also adjusting the negative impact those cupcakes would have on a strength of schedule. The do this anyway, but only in their heads, This would provide them a formula for doing so and create more consistency to how cup cake schedules are dealt with. What is considered a cup cake would of course vary depending on where they ended up on the RPI scale. This would have no real effect since the original RPI would still act as a starting point and the adjustment would only serve to eliminate a few non relevant games. It's not much different than eliminating the top and low scores in the Olympics. That is also nothing but an adjustment to create more consistency in scoring.