A Look at SOS and the ACCT

[El_Diablo]

UNC lost 2 ACC games away from home this year and Duke lost 3 ACC games away from home.

I think you missed the point. UNC had a slimmer efficiency margin on the road. The obvious reasons: they had lots of close wins over inferior opponents, and they lost their two ACC games by more total points than Duke's three ACC losses.

So UNC had a weaker overall road margin...no big deal. It's not an indictment of your program, it's just the math. And some of us actually appreciate the work that went into it.

[gumbomoop]

UNC lost 2 ACC games away from home this year and Duke lost 3 ACC games away from home.

This is an accurate, but somewhere between slightly and significantly misleading, statement. Would you agree - or deny - that the unbalanced schedule always makes simple won-loss comparisons misleading?

IMO, UNC was spared 3 tough away contests: no visit to Md or VaTech, no Singleton at FSU.

Duke avoided 1 additional tough away contest: at Clemson.

It does strike me that the purpose of your particular comment here is to imply that the OP's analysis is neither interesting nor meaningful. But maybe I've misunderstood you, and have inferred a put-down where you intended none.

Having said this, I hasten to add that the Heels played a fine game last Sat eve. I haven't visited IC in quite some time, but assume the early-season angry debate between Drew supporters and detractors has subsided. Nor, I assume, do folks over there pine for either Graves or the Wears. Were I a Heel fan, I'd be feeling confident right now, whistling past the graveyard, perhaps, about close wins, but thinking positively that even unimpressive wins are wins. I'd be saying, "Our guys know how to win the close ones." I'd also be hoping - secretly and with just a small tinge of nervousness - that Kyrie will not reappear this season.

[4decadedukie]

This is a superb, painstaking analysis, for which I am extremely greatful.

With this sincerely said, I have a long-standing problem with such wholly quantitative (completely non-qualitative) evaluations re Duke's strength-of-schedule. Whether at home or on the road, essentially EVERY team the Blue Devils face will play with extra effort, focus and special ferocity, year after year. I believe this is due to the fact that Duke is a -- perhaps "the" -- perennial paradigm for top-tier Division I excellence. Therefore, defeating Duke "makes" any team's season, even if the rest of their campaign is a disaster. We see this in SO MANY games every year, for example our recent loss to VPI and our recent victory against Clemson.

If this foregoing premise is correct, what is a normally mediocre team for any legitimate quantitative rating, is likely to play like an excellent team when it faces Duke. Obviously, this consequentially means that mediocre opponents in the numerical analysis -- based on their aggregate, season-long performance -- should be assessed as much more demanding contestants for Duke.

In essence, our strength-of-schedule should NOT be based on how our opponents generally perform, but rather how they perform against the Blue Devils.

[kong123]

Duke's efficiency margin on the road was significantly better than UNC's margin. Please read the first post.

Road Team	Off	Def	Margin
Duke		109.46	96.95	12.51
North Carolina	104.67	99.81	4.86

thats fine, UNC's conference record was better than Duke's. That is a better indicator to me

[nocilla]

thats fine, UNC's conference record was better than Duke's. That is a better indicator to me

The Cleveland Cavaliers had the best record in the NBA last year.

I don't mean to discredit the regular season. It is a nice achievement. Congratulations to UNC for winning the regular season. You earned a #1 seed for the Conference Championship.

[ns7]

thats fine, UNC's conference record was better than Duke's. That is a better indicator to me

Then I have no idea why you're reading a thread about advanced metrics unless you're trolling.

Here's one for you: Green Bay was the #6 seed in the NFC but when you looked at advanced metrics, they were the best team in the NFC entering the playoffs.

[dukeballboy88]

To me, and ive said this for awhile, when the acc added more teams and everyone didnt play each other twice, it ruined the acc regular season title. I dont know how you can crown a champ when you dont play everyone at home and away.

The conference tourney has to mean more these days. The NCAA should implement a rule that says you cant get a #1 seed unless you win your conference tourney or something. From what Im reading, this ACC tourney is going to give the winner a #1 seed if either Duke or UNC win it. I guess lil roy cant claim to take this weekend off this year!

My question about sos is, if Kansas St and Marquette, 2 teams Duke beat, win both of their conference tourneys will that help Duke's sos or rpi and getting a #1 seed?

[flyingdutchdevil]

Good job to the OP for the heavily quantitative analysis. Quite interesting.

That said, there is an argument going on between pro-UNCers (ie Kong) and everyone else. Hate to say it, but I agree with Kong. I may be completely off base, but I feel like a lot of people are using the OP's analysis to say that even though UNC won the reg. season ACC, Duke 'deserved' to win the reg. season ACC because of a more difficult SOS. Really? Kong is right - UNC won the ACC fair and square. That does not mean that UNC is the better or more talent team. It means that they pulled out the most wins against a similar set of opponents. They have a better record and so they deserve the credit.

That said, I hope (and know) that Duke will absolutely destroy their opponents during the ACC tournament. If that means facing UNC during the ACC final, so be it. I'm looking forward to the challenge.

[mus074]

Awesome, awesome work. Thanks for sharing it. I enjoyed reading about your process at least as much as the results of the analysis. It is worth remembering that in the end, we are taking about statistical probabilities. If each match up were a best of seven series, the probabilities would be a much better indicator of who will win, and we could only prove the accuracy of the probabilities if we could somehow get the teams to play the same tournament over and over and over again to compare the distribution of results to the forecasted probabilities.

I'm not using his numbers, I'm using mine. They differ in three respects - first, I don't adjust for opponent; he does. That is, he uses an "adjusted" offensive and defensive efficiency that takes into account strength of schedule, I don't (because I'm not mathematically savvy enough to do it). Second, as I described in my initial post, I split the conference into 24 teams, rather than 12 - a home and road version of each team. So if you only play FSU in Tallahassee, for example, only FSU's performance as a home team is factored in to your strength of schedule. Third, I suspect (although am not certain) that his strength of schedule uses season long performance by the teams, including non-conference games, while mine uses in-conference performance only.

I'm sure my method is subject to criticism. Pomeroy's is, too - frankly, the fact that he has UNC with only the 6th toughest schedule in conference just seems wrong when, as you and I have both pointed out, they had to face Duke, FSU, and Clemson twice.

Just for the sake of further comparison, Pomeroy adjusts for home court advantage by giving a small adjustment (I think maybe 4%) to the home team and the same amount taken from the away team. I have a fundamental problem with this format. It works well when looking at all teams in the aggregate and for seeking to compare apples to apples. However, I would love to see him use all of his raw data and derive two different factors for each team.

One would be how far from the projected efficiencies (not taking into account his home/road adjustment) each team is on the road and at home to see how much of an individualized factor it is for each team. I think your idea of splitting teams into two alter egos captures this concept much better. This would be very helpful in predicting future performance, as some teams (like UK) have different "road" and "home" factors than other teams.

The second would be to compare how teams fare against the predicted adjusted efficiencies based upon the level of competition. The adjusted efficiency seeks to take this into account, but I would guarantee some teams are more influenced by the opponent's ability than others, but the pythag theory gives equal "gravity," if you will, to all the teams relative to their performance. Looking for a statistical correlation of opponent level to distance from the predicted value would tease out what I think of as the team's "will." Do they trounce weak teams and play mediocre against better ones? Do they only show up against great teams and then sleep walk through lesser competition? This would also be valuable predictive information come tourney time, I think.

Thanks for indulging my inner geek. Your work is awesome and I thank you again for sharing it with us.

[dukeballboy88]

By the way, awesome work on the analysis, thanks for that!

[kong123]

Good job to the OP for the heavily quantitative analysis. Quite interesting.

That said, there is an argument going on between pro-UNCers (ie Kong) and everyone else. Hate to say it, but I agree with Kong. I may be completely off base, but I feel like a lot of people are using the OP's analysis to say that even though UNC won the reg. season ACC, Duke 'deserved' to win the reg. season ACC because of a more difficult SOS. Really? Kong is right - UNC won the ACC fair and square. That does not mean that UNC is the better or more talent team. It means that they pulled out the most wins against a similar set of opponents. They have a better record and so they deserve the credit.

That said, I hope (and know) that Duke will absolutely destroy their opponents during the ACC tournament. If that means facing UNC during the ACC final, so be it. I'm looking forward to the challenge.

I understand that the uneven schedule does create a debate over the regular season SOS and even the regular season champion. I understand that the numbers provided in this graph tell a different story than what the actual outcome was. In the end, you guys think the tournament champion trumps the regular season championship, but in another thread, many have pointed out that they think UNC has a tougher road to the final game than Duke does. If it were the best of 3 or 5 games, then the tournament would be a better indication of a true championship, but since the teams are seeded due to their conference record, not statistics like the ones provided in this thread, then can't someone make the same argument against the tournament? Isn't the longer ACC regular season a better indicator to the strength of a team than a 3 day tournament where the seeding of the teams doesn't reflect the strength of a team, therefore the tournament is even less balanced and fair than the regular season?

[mus074]

Good job to the OP for the heavily quantitative analysis. Quite interesting.

That said, there is an argument going on between pro-UNCers (ie Kong) and everyone else. Hate to say it, but I agree with Kong. I may be completely off base, but I feel like a lot of people are using the OP's analysis to say that even though UNC won the reg. season ACC, Duke 'deserved' to win the reg. season ACC because of a more difficult SOS. Really? Kong is right - UNC won the ACC fair and square. That does not mean that UNC is the better or more talent team. It means that they pulled out the most wins against a similar set of opponents. They have a better record and so they deserve the credit.

That said, I hope (and know) that Duke will absolutely destroy their opponents during the ACC tournament. If that means facing UNC during the ACC final, so be it. I'm looking forward to the challenge.

I think you may be misstating the value or meaning of pfr's awesome analysis. It looks at the approximately 1100 possessions per team or 13,200 total data points to see how well each team did on each possession. Pure win/loss records gives you a separate and different analysis of only 192 data points. They tell different stories, but don't trump the other. A better per possession efficiency doesn't mean you deserved to win more games. Nor does winning more games mean you will necessarily win any one future game.

[mus074]

Isn't the longer ACC regular season a better indicator to the strength of a team than a 3 day tournament where the seeding of the teams doesn't reflect the strength of a team, therefore the tournament is even less balanced and fair than the regular season?

In the end, the team that wins 6 in a row in single elimination wins the crown in college basketball. In football, its the season that matters. Not trying to say one is better - that's just how its done.

[kong123]

They tell different stories, but don't trump the other. A better per possession efficiency doesn't mean you deserved to win more games. Nor does winning more games mean you will necessarily win any one future game.

I completely agree with this.

in the end, you have to play the opponent in front of you. after that, you are judged by whether or not you won the game. the other statistics provided, like the ones in this thread, can provide a different perspective, but ultimately, they are only for entertainment and can be used to confirm the results or to spin the results in a different direction. bottom line, you still have to win the games to make it count.

[davekay1971]

same could be said for UNC.

There's a very simple explanation: Duke's strength of schedule was hurt by having to play UNC twice, while UNC's was helped by having the opportunity to play Duke twice. UNC's status as a basketball school has been bouyed up by their proximity to Duke for the last 25 years...why should this year be any different?

(The above written with my tongue planted firmly in cheek, I assure you)...

Joking aside, the OP, like KenPom's analysis technique, is a marvellous way to get past the more simple win-loss record to predict future events in sports. The limitation, of course, is that a single game is, to some degree, unpredictable, which is why college basketball's post-season single game formats have a nasty way of refusing to follow predictions (KenPom's well documented success in predicting NCAAT contenders being a notable exception). To the point of this discussion, Kong may consider the OP's analysis to be data spin, but it's actually a very nice statistical look at the overall ACC conference performance of each team. It MAY be helpful in predicting outcomes of the ACCT...as long as groups of 18-22 year olds perform, during the individual conference games, to the level they performed over the course of the season.

With regards to which is a better measure of who's the best team in the conference - regular season record vs conference tournament, I would submit that both are flawed. If the conference season was balanced, I would say it was the better measure. With the conference season being markedly unbalanced, conference season record becomes a deeply flawed way to determine who was the best team in conference. Thus the decision in the days of the old Southern Conference to make the tournament winner the conference champion. During the days of the 8 team ACC, the regular season could have (perhaps should have) fairly determined the conference champion. Now that we're back to unbalanced schedules, however, the tournament can be argued to be a more equitable way to determine a champion.

If, however, we think that comparing regular season records of teams that play different schedules is the most equitable way to determine a champion, perhaps the NCAA should use the regular season performance to determine the national champion? A nice computerized amalgamation of national polls and computer polls would leave everyone feeling satisfied, I'm sure...

[DevilBen02]

I may be completely off base, but I feel like a lot of people are using the OP's analysis to say that even though UNC won the reg. season ACC, Duke 'deserved' to win the reg. season ACC because of a more difficult SOS. Really? Kong is right - UNC won the ACC fair and square. That does not mean that UNC is the better or more talent team. It means that they pulled out the most wins against a similar set of opponents. They have a better record and so they deserve the credit.

You are correct that the information presented here should not be used to argue that Duke "deserved" to win the regular season. With regard to the regular season, playing with these numbers is merely an academic exercise because the games have been played and the outcomes are known. No matter how you look at the numbers, UNC got the wins and should be congratulated for earning the top seed.

I believe that the point of the numbers, however, is that wins and losses alone don't tell the whole story of what a team has done and what they may be capable of doing in future games. In that regard, although we can't really argue that Duke should be considered the regular season champ, the numbers provide a reasonable argument that Duke is in fact the better team despite the conference record and is more likely to be successful going forward.

[kong123]

You are correct that the information presented here should not be used to argue that Duke "deserved" to win the regular season. With regard to the regular season, playing with these numbers is merely an academic exercise because the games have been played and the outcomes are known. No matter how you look at the numbers, UNC got the wins and should be congratulated for earning the top seed.

I believe that the point of the numbers, however, is that wins and losses alone don't tell the whole story of what a team has done and what they may be capable of doing in future games. In that regard, although we can't really argue that Duke should be considered the regular season champ, the numbers provide a reasonable argument that Duke is in fact the better team despite the conference record and is more likely to be successful going forward.

agreed. I also feel that Duke is the more experienced and therefore the more consistent of the two teams. This is reflected in the data presented in this thread.

[flyingdutchdevil]

You are correct that the information presented here should not be used to argue that Duke "deserved" to win the regular season. With regard to the regular season, playing with these numbers is merely an academic exercise because the games have been played and the outcomes are known. No matter how you look at the numbers, UNC got the wins and should be congratulated for earning the top seed.

I believe that the point of the numbers, however, is that wins and losses alone don't tell the whole story of what a team has done and what they may be capable of doing in future games. In that regard, although we can't really argue that Duke should be considered the regular season champ, the numbers provide a reasonable argument that Duke is in fact the better team despite the conference record and is more likely to be successful going forward.

This is a great post. I agree with basically everything that you have said. However, stats don't tell the whole story. It looks like standard deviation has not been considered a factor. On either a game or a single possession basis, I'd be surprised if Duke has a lower SD than UNC. That is relevant because that is a prime statistic for consistency. While Duke may have overall better efficiency in the long haul, the fact that it is more inconsistent (assumption based on eyeballing numbers, not actually doing the analysis) with a higher SD isn't better than having lower efficiency numbers with a smaller SD (ie UNC). For instance - and this is incredibly simple and not based on any data - but winning 5 games by 5 points is less efficient than winning 4 games by 12 and losing 1 by 1, but the former has a lower SD and higher consistency.

[mus074]

This is a great post. I agree with basically everything that you have said. However, stats don't tell the whole story. It looks like standard deviation has not been considered a factor. On either a game or a single possession basis, I'd be surprised if Duke has a lower SD than UNC. That is relevant because that is a prime statistic for consistency. While Duke may have overall better efficiency in the long haul, the fact that it is more inconsistent (assumption based on eyeballing numbers, not actually doing the analysis) with a higher SD isn't better than having lower efficiency numbers with a smaller SD (ie UNC). For instance - and this is incredibly simple and not based on any data - but winning 5 games by 5 points is less efficient than winning 4 games by 12 and losing 1 by 1, but the former has a lower SD and higher consistency.

I also would love to see the SD on the teams' conference season performances. I would think at least that Duke's offensive performance is less consistent for the simple fact that Duke relied much more heavily in conference play on 3's than did UNC. Duke got 29.5% of its point in conference games from 3's (6th of 12), while UNC got only 18.9% of its points from 3's (11th) and 61.0% of its points from 2's (1st). I am assuming that UNC's distribution of effective shooting percentages was much tighter than Duke's, particularly since the eFG% gives a 50% premium to the 3-point variety.

[pfrduke]

I also would love to see the SD on the teams' conference season performances. I would think at least that Duke's offensive performance is less consistent for the simple fact that Duke relied much more heavily in conference play on 3's than did UNC. Duke got 29.5% of its point in conference games from 3's (6th of 12), while UNC got only 18.9% of its points from 3's (11th) and 61.0% of its points from 2's (1st). I am assuming that UNC's distribution of effective shooting percentages was much tighter than Duke's, particularly since the eFG% gives a 50% premium to the 3-point variety.

Ask and ye shall receive.

Duke's standard deviation for offensive efficiency was 13.5, with performances ranging from a low of 88.2 to a high of 127.6. 7 performances were between 100 and 120, with 5 below and 4 above. Our eFG standard deviation was 6.1% - we were never lower than 40, and only once higher than 60.

Carolina was actually less consistent, in part because of a higher high (139.5) and a lower low (77.4). They did end up with 8 performances between 100 and 120 (including 5 of the last 6), with 6 below and 2 above. Their eFG also varied more - a standard deviation of 9% with 4 games below 40 and 2 above 60.