PDA

View Full Version : College Basketball Analytics Conference in Indy



whatazag
01-13-2017, 11:58 AM
http://www.espn.com/mens-college-basketball/story/_/id/18468008/ncaa-tournament-officials-meet-analytics-experts-consider-creating-new-metric

I know a lot of people on the board will be happy to reduce importance of the RPI and go to the advanced analytics, but I am not sure its such a great idea either. RPI is not a good system, and I don't have a better solution, but I fear this could take college basketball down the path of college football where teams will be incentivized to blow out opponents by as many points as possible.

In theory we could be cost ourselves a seed line putting in our less efficient bench players in a blowouts throughout the season.

tinfoilzag
01-13-2017, 12:16 PM
If they really want to bring margin of victory into it, they should base it on ranges of scores. For example, group score differential by range:

1-3 points - within one possession of a tie or win/loss
4-9 points - close game at the end. The winning team probably had it's starters in the whole time
10+ points - game was well in hand at the end

I still don't like it much because the only stat of a game that matters is W/L but if they need something, this system wouldn't encourage destroying lower quality teams.

CDC84
01-13-2017, 01:00 PM
In addition to concerns about running up the score, an over-emphasis on margin of victory also keeps coaches from being able to use bench guys who really need playing time. It also doesn't allow coaches to experiment with running plays, new defenses, etc., that you can afford to experiment with if the game is a blow out.

Another issue......when Gonzaga plays a 200 RPI team in league, it's not the same as when Duke plays such a team out of conference. For one thing, 50% of the time the Zags are facing that 200 RPI team on the road in a hostile arena. Duke gets them at home as a buy game. Also, Duke might face that 200 RPI team once every thirty years. Gonzaga has to face that team (and that coach) two or three times a year. What I am trying to say is that conference familiarity and road environments are going to play a part in margin of victory. Duke might beat, say, LMU by 50 points because the game is at Cameron Indoor Stadium, and the Lion coaches have zero experience of scouting Duke and its players. Gonzaga might beat the same LMU team by 26 points because Coach Dunlap and company play the Zags 2 to 3 times per season. They know their players inside and out, etc.

My view has always been that it takes six one point wins to win the NCAA tournament.

The object of the sport is to win basketball games. Period. I don't like there being an over-emphasis on margin of victory.

RenoZag
01-13-2017, 01:17 PM
Thank you for the article link, whatazag. Thought this passage was the most telling:


"You need to stay relevant in the age that you're operating in," Gavitt said on NCAA.com. "Certainly relevant today is embracing analytics and technology to the appropriate level. In an imperfect process, I think what the committee strives to get as perfect as possible is to have justification and rationale for their decisions. And the more that can be rooted in fact and in data, the more comfortable they can be with those decisions and the more justifiable they can be in explaining them."

Gavitt said the National Association of Basketball Coaches wants advanced metrics -- an "even more powerful microscope" -- to play a bigger role in the selection process. That prompted next week's meeting.

Jeff Sagarin (Sagarin), Kevin Pauga (KPI), Ken Pomeroy (KenPom.com) and Ben Alamar (ESPN's BPI) are quite the line-up.

I think come Selection Sunday 2018, we'll still hear phrases like : " Who did you beat and when did you beat them? "; " [team] passes the eye test"; and " In the view of the committee, [Power Conference team]'s 18 win season is better than [mid-major team]'s 25 win season.

CDC84
01-13-2017, 01:27 PM
Many analysts pointed out that they felt that seeding in last year's tournament was highly influenced by sites like kenpom.com. More than ever before.

webspinnre
01-13-2017, 01:51 PM
Keep focusing on things like efficiency stats!

MBAGael
01-13-2017, 02:01 PM
Whatever they decide, I hope they are transparent about it to remove any guesswork.

MickMick
01-13-2017, 02:35 PM
*The "running up the score" concerns are relative to bubble teams fighting for an "at large" bid. How many is that, 4-8 teams in any given year? There are more "locks" than bubble teams. This is how basketball is different than football.

*Lopsided victory is nothing new in college basketball. GU wins by 20 points or more quite often.

*The quality of RPI is relative to sample size. The more games played, the more reliably representative the results. RPI shouldn't even be referenced until the selection committee meets. Indeed, I believe the biggest "flaw" with RPI is that it is commonly referenced before the season is over and at a time when it isn't reliably representative. People that claim RPI is flawed or often the same people that are flawed in it's usage.

seacatfan
01-13-2017, 02:44 PM
Whatever they decide, I hope they are transparent about it to remove any guesswork.

Don't count on it. They've paid lip service to transparency in recent years but I'm not buying it. What goes on behind closed doors they don't want us to know about.

MDABE80
01-13-2017, 02:54 PM
Any time we let the software run the show and generate conclusions, we must remember the "validation" process. Most retrofit the data and then weight in. In the end, when experts quote such things, they leave much of the "sausage making " out. Why? Because it's just simpler for them.

It's like a prolonged discussion we had last year when the team was flailing. Some here relied on numbers....the data points...when it was obvious the data points simply didn't concur with the actually playing of the games. I guess I'm a bit old fashioned. I love the data based stuff but when it betrays your own eyes, go with your eyes. As CDC said above. 6 "1 point wins" matters as much as a 20 point win when it comes right down to it. "It" being the NCAA Tournament. Problem is that use of the data as a defining resource may generate a good team getting a hard pathwhen its obvious the team wins, just not by blowouts.

Birddog
01-13-2017, 03:06 PM
Title of thread could have been "RIP RPI"

CDC84
01-13-2017, 03:31 PM
The quality of RPI is relative to sample size. The more games played, the more reliably representative the results. RPI shouldn't even be referenced until selection the committee meets. Indeed, I believe the biggest "flaw" with RPI is that it is commonly referenced before the season is over and at a time when it isn't reliably representative. People that claim RPI is flawed or often the same people that are flawed in it's usage.

+1. Also, many media types like Jim Nantz, who doesn't cover a single college basketball game until late February, misunderstands how the committee uses the RPI.

CanadianZagsFan
01-13-2017, 03:32 PM
If they really want to bring margin of victory into it, they should base it on ranges of scores. For example, group score differential by range:

1-3 points - within one possession of a tie or win/loss
4-9 points - close game at the end. The winning team probably had it's starters in the whole time
10+ points - game was well in hand at the end

I still don't like it much because the only stat of a game that matters is W/L but if they need something, this system wouldn't encourage destroying lower quality teams.

I like that. It seems too simple to accept if that makes sense (HS sports need that rule in the States ,but without no shot clocks is more the issue there).

I'd go

1-6 PTS
7-14 PTS
15 + PTS

10 + PTS are not well in hand games here in Canada for higher level HS/College levels. But , I do like where you are going with this at very least at grassroots levels for reg season seeding and so forth.

Birddog
01-13-2017, 03:39 PM
I can see 6, 9, and 12pts. 1 to 2 possession, 3 possession, 4 possession win margins. How much weight to add for those margins is the problem. I've felt for a long time that no team that finishes in the bottom half of their conference should get in, that would open it up a lot.

ZagsObserver
01-13-2017, 03:51 PM
This would benefit GU. GU typically has a much larger margin of victory than most power 5 conference teams but gets no credit for margin of victory. Rather, we get penalized for playing weak conference foes with the rpi calculation. While I cannot say whether is a good thing or bad thing, it would likely help gonzaga in the metrics department.

soccerdud
01-13-2017, 04:25 PM
I've felt for a long time that no team that finishes in the bottom half of their conference should get in, that would open it up a lot.

completely disagree with this. the goal should be to get the best teams in the tournament. as much as i disagree with penalizing a team simply because they play in a weak conference, the argument to penalize a team because they play in a really good conference is even dumber and weaker.

we should be looking at ways to eliminate conference-based prejudice in the selection process, not invert it.

seacatfan
01-13-2017, 04:42 PM
I've felt for a long time that no team that finishes in the bottom half of their conference should get in, that would open it up a lot.

UConn finished 9th out of 16 teams in the Big East in 2010-11...and won the National Championship that year. Only went 9-9 in conference but went 23-0 in OOC and post season. Might be an anomaly but using your guideline a team that proved themselves to be the best over 6 games in the Tourney would've been left out of the field.


edited--actually they wouldn't have been left out because they got the auto bid for winning the Big East tourney, but it still illustrates the point what can happen if you make a hard and fast rule about where a team finishes in their conference

DixieZag
01-13-2017, 06:59 PM
I can see 6, 9, and 12pts. 1 to 2 possession, 3 possession, 4 possession win margins. How much weight to add for those margins is the problem. I've felt for a long time that no team that finishes in the bottom half of their conference should get in, that would open it up a lot.

This is the rule that I want more than any other.

If a team cannot finish in the top half of its conference, it doesn't deserve to be in the NCAA tournament.

And I don't care that some team with a conference record of 9 wins and 11 loss's, has more "Top 50 wins" or "11th hardest schedule" or whatever. I just. don't. care.

It also makes the regular conference season mean that much more, you could have "mini-tournament" games in the last few weeks of conference season in some big conferences. That's a way to add excitement/eyeballs/$$ to those TV games.

Birddog
01-14-2017, 03:08 AM
edited--actually they wouldn't have been left out because they got the auto bid for winning the Big East tourney, but it still illustrates the point what can happen if you make a hard and fast rule about where a team finishes in their conference
Winning your conference tournament solves the problem. I just don't think you should be rewarded for a bottom half finish in your conference. I could accept a 50% win margin in conference just don't think losing more than 50% should get a reward without winning conference tourney.

Coach Crazy
01-14-2017, 04:47 AM
"Gavitt said the National Association of Basketball Coaches wants advanced metrics -- an "even more powerful microscope" -- to play a bigger role in the selection process. That prompted next week's meeting."

This. This should also be an influence in how people weight a player's status. This is one thing I like about Coach Cal. He gets this, and overall I believe it helps put together a lineup with depth, while also avoiding the trouble of having to share focus on the court. If Few and Co. can successfully be a high efficiency team, recruiting will get better, and you'll see more seasons like this one. It would be great to be the team that gets a pipeline to the euro's, transfers, and a McD's or two every year. They'll be a blue within a decade or before, with enough years of that formula.


Sent from my iPhone using Tapatalk

kitzbuel
01-14-2017, 05:00 AM
Any margin of win calculation is exceptionally problematic because of styles of play. A six point deficit to Bennet's WSU teams seemed like a 20 point deficit. They just would not let opponents have enough possessions to effectively cut into that lead.

Margin of win is either going to be misleading or drive teams to one style of play.

Bogozags
01-14-2017, 05:59 AM
I can see 6, 9, and 12pts. 1 to 2 possession, 3 possession, 4 possession win margins. How much weight to add for those margins is the problem. I've felt for a long time that no team that finishes in the bottom half of their conference should get in, that would open it up a lot.


I agree completely and have said this many time that if a school cannot win at least 50% of its conference games, then that school has no business being in the NCAA Tournament. The lone exception would be, as the current rules allow, that any team winning its conference tournament has an automatic bid.

There are two many BIG FIVE schools that do not win 50% of their league games and then get in the Dance, while a Monmouth doesn't get it. Football use to have a "rule" that said D-1A schools had to win a minimum of six gamers and have at least a .500 record. Now, as long as you have six wins you can play in a bowl game regardless of how many losses a team might have...

Ekrub
01-14-2017, 06:13 AM
Any margin of win calculation is exceptionally problematic because of styles of play. A six point deficit to Bennet's WSU teams seemed like a 20 point deficit. They just would not let opponents have enough possessions to effectively cut into that lead.

Margin of win is either going to be misleading or drive teams to one style of play.

Kenpom combats that with his adj. Offensive efficiency.

TexasZagFan
01-14-2017, 06:21 AM
If I may, allow me to add a wrinkle to the conversation.

IMO, it's all about the money. The more teams you get into the Dance, the more your conference rakes in. So, let's not put a limit on the number of Power 5 schools getting into the tournament, limit the guaranteed money to no more than half the schools from a particular conference. Simply funnel that extra money to conferences that placed only one team from the auto-bid.

So the ACC gets 11-12 of their teams in this year. Limit the conference to first round money on eight teams.

Power 5 conferences won't miss that money at all. Shouldn't affect WCC either, we should normally have two bids in a given year.

amaronizag
01-14-2017, 07:05 AM
Ekrub has it right. Adjusted Offensive Efficiency is a tempo-free statistic. It is the number of points we score per possession. The number of possessions doesn't matter. Same with Adjusted Defensive Efficiency -- it's the number of points we allow our opponents to score against us per possession. The number of possessions (pace of the game) doesn't matter. Teams that play tight defense and limit the number of possessions in a game will (generally) win by smaller margins so any system that considers margin of victory or number of points scored will tend to discriminate against good defense. Something the statisticians will consider when they formulate the new system.

JPtheBeasta
01-14-2017, 08:13 AM
I propose a KenSagarinPomUSATodayESPNCoachesPollbprpi aggregate rankings index that is a weighted averages of all the best ranking indexes. I might even throw in the electoral college and a Nickelodeon Kids' Choice poll for added accuracy.

This tongue-in-cheek idea is based on the wisdom of the crowd principle, and is mostly serious. Some sort of averaging between several ranking indexes that people like should provide better results than one or two. The more we use, the more closely we will get to that Platonic ideal of a the Perfect March Madness Tournament Field. Whether that actually exists I will leave to the philosophers in the crowd.
https://en.wikipedia.org/wiki/Wisdom_of_the_crowd

CDC84
01-14-2017, 09:45 PM
I got into a discussion with a co-worker regarding this whole thing. He made some interesting points. He first started off by saying that the RPI was by no means a perfect tool. It has its flaws. However, when it gets to the end of the season, the RPI and kenpom.com end up being more similar than you might think. What one ranking system likes in a team, the other might not like by a few spots. He also said something that really made me think........KenPom's number is meant to be PREDICTIVE. What does PREDICTIVE mean? It means its worth is to be a reliable tool for gambling. Is the NCAA going to be comfortable with this?? They seem to be so paranoid about gambling as it is.

soccerdud
01-14-2017, 10:03 PM
I got into a discussion with a co-worker regarding this whole thing. He made some interesting points. He first started off by saying that the RPI was by no means a perfect tool. It has its flaws. However, when it gets to the end of the season, the RPI and kenpom.com end up being more similar than you might think. What one ranking system likes in a team, the other might not like by a few spots. He also said something that really made me think........KenPom's number is meant to be PREDICTIVE. What does PREDICTIVE mean? It means its worth is to be a reliable tool for gambling. Is the NCAA going to be comfortable with this?? They seem to be so paranoid about gambling as it is.

predictive is about guessing what will happen. yes, because of the nature of gambling it can be applied there. but the NCAA having an issue with predictive models because they can be used for gambling is like the NCAA having an issue with teams flying to games because airplanes can be used to get to vegas. just a complete non sequitur here.

CDC84
01-14-2017, 10:38 PM
Yeah, but the NCAA doesn't use logic. They are so freaking paranoid about being associated with anything even remotely related to gambling, even though their whole tournament is based on it. And not just in Vegas. 25% of all US office workers participate in an official pool.

JPtheBeasta
01-15-2017, 07:13 AM
predictive is about guessing what will happen. yes, because of the nature of gambling it can be applied there. but the NCAA having an issue with predictive models because they can be used for gambling is like the NCAA having an issue with teams flying to games because airplanes can be used to get to vegas. just a complete non sequitur here.

ESPN seems to have no qualms about posting Team X's % chance of winning. I could be wrong, but I think I have seen graphics like that many times on their broadcasts. Their pundits also try to predict winners all of the time- I do appreciate that they don't handicap games with a spread, though. Colin Cowherd on the radio is an example of guy who focuses a lot on betting, but I don't think he is with ESPN any more (but he did it when he was employed there).

thegloriousgoateeofKP
01-20-2017, 08:43 AM
It's taking place today.

Andy Glockner says the Committee's goal seems to be "to move the team sheet evaluation away from RPI defining accomplishment to a better composite metric.."

https://twitter.com/AndyGlockner/status/822481452625514496

His account will be a good one to follow. Jeff Sagarin and Michael DeCourcy are there as well. Not sure who else is in the room.

CDC84
01-20-2017, 08:56 AM
Michael DeCourcy ‏@tsnmike · 1h1 hour ago
Jeff Sagarin is talking now. Just started throwing out algebra terms. It's here where I climb under the table and hide. #NCAAanalytics

I would love to be a fly on the wall during this meeting.

CDC84
01-21-2017, 08:02 PM
Good summary of the "conference" that took place with Pomeroy, Sagarin, etc., and several college basketball writers:

http://www.sportingnews.com/ncaa-basketball/news/ncaa-tournament-selection-committee-ratings-system-sagarin-kenpom-rpi-bpi-jerry-palm/1jq5rphzlboje1agjwdwiv0wk1

thegloriousgoateeofKP
01-21-2017, 08:58 PM
Good summary indeed. Here's hoping the Committee actually takes action, instead of just talking about it.

Also, my goodness, I can barely believe this. Unfathomable stupidity.


he told a story of once visiting a gathering of coaches and presenting a hypothetical: There’s a 10-team league in which each team plays the other nine members once each. Team A wins all nine of its games by a point. Team B goes 8-1 with all of its victories coming by 50 points. He asked the coaches: Which is the better team? Their responses were split down the middle.

RenoZag
01-22-2017, 08:36 AM
From the DeCourcy article linked by CDC:


Pauga suggested that one step the committee could take to refine the process, even without conceiving a new ratings system, is to redesign the “team sheets” that are presented in evaluating each tournament candidate. For instance, Clemson’s team sheet would feature its breakdown of record against the RPI top 50, teams ranked 51-100, and so on, with venue noted and non-conference games highlighted. Pauga contends those resumes don’t properly acknowledge the challenge of gaining road victories.

Glockner also stressed that road wins aren’t given proper weight under the current system, using last season’s snub of Monmouth as an example.

Pauga presented the idea of breaking down a team’s record into four categories. The most prominent category would include, hypothetically, a team’s results at home against top 25 teams, on neutral courts against top 50 teams, and on the road against those in the top 75.

“If I look at that as a coach or as a fan,” Pauga said, “I trust the process more.”


NCAA vice president Danny Gavitt acknowledged that the selection of at-large teams “based on results” is spelled out in the selection procedures. So while the ratings system will change, it seems unlikely that core principle will.

“The coaches have recommended a composite,” he said. “They may not totally understand what they’re recommending.

kitzbuel
01-30-2017, 11:03 AM
The expectation here is that changes will actually be implemented in next year's selection.


Changes will not affect how teams are selected and seeded for the 2017 NCAA Tournament. The change would likely not be instituted until 2018.

Read more here: http://www.kansas.com/sports/college/wichita-state/article127523749.html#storylink=cpy

http://www.kansas.com/sports/college/wichita-state/article127523749.html

RenoZag
02-05-2017, 09:52 AM
http://kenpom.com/blog/that-meeting-at-the-ncaa-hq/#more-1417


That the meeting happened at all is a big step. The NCAA is a large organization with diverse membership and the fact that it is even this far down the path to changing the data used in the selection process should be applauded. Especially considering that basis for the existing process has been locked in for over three decades.

Change isn’t worth it when the cost of change outweighs the benefits of it. But in this case, basketball people at the administrative level and on the sport’s front lines increasingly believe that the benefits of modernizing the tools of the selection process will offset the cost of redesigning the technical workings that support it. And we shouldn’t take that for granted.


. . . the outcome of the game has to matter. This is why we watch the game. Make the selection process, and thus the games, purely about points scored and allowed and the games become less entertaining.4 There is no special purpose to having one more point than your opponent. No point in managing foul trouble. No point in hoisting threes in the final minute to catch up. The contest becomes one of points accumulation. . .

However, the reason margin-of-victory works in identifying the best teams is because the goal is to win the game. Teams are happy with a one-point win after the fact, but they’d rather not have the game come down to the final possession while its occurring. There’s already incentive to run up the score (to an extent) in order to avoid the game coming down to a team’s players not being able to count off 5.6 seconds in their heads so they know when to shoot.

Interesting read