[Cracked Sidewalks] The High Risk Game of Luck

[brewcity77]

http://www.crackedsidewalks.com/2019/01/the-high-risk-game-of-luck.html

Watching Marquette win another nail-biter over Georgetown coupled with Andrei Greska's article at Paint Touches about why the computer metrics aren't high on Marquette got me thinking about this team's future. After reading the article, I was thinking a lot about last year's Xavier team, which had some striking similarities to this year's Marquette. Both teams were ranked highly in the AP & Coaches' polls but weren't ranked as highly by computer metrics such as Pomeroy, Sagarin, and Torvik.

These teams shared some hallmarks. Both had a number of close wins. During the regular season, Xavier was 8-0 in games decided by 5 or fewer points while Marquette is currently 5-0 in such games. While neither team lost often, their losses were by big margins. Xavier had an average losing margin of 16.3 ppg with a minimum loss differential of -9 while Marquette has an average losing margin of 17.3 ppg with an identical minimum loss differential of -9.

The third thing that stood out was both teams were ranked high in kenpom's "Luck" category. Xavier was the #1 luckiest team in the country while Marquette currently checks in at #26. According to Pomeroy, Luck "is the deviation in winning percentage between a team’s actual record and their expected record using the correlated gaussian method. The luck factor has nothing to do with the rating calculation, but a team that is very lucky (positive numbers) will tend to be rated lower by my system than their record would suggest."

In my most recent Cracked Sidewalks S-Curve, Marquette checks in as the #12 team in the curve and the last 3-seed. Based on Marquette's current Pomeroy rating of 36, they are overslotted by 24 positions. To figure that out, I simply multiplied the seed by 4 (4 teams per line) & subtracted the seed value from the Pomeroy rank. I looked at similarly overseeded teams over the past 5 years. My criteria were teams with top-50 Luck ratings, seeding disparities of 10+ positions in the S-Curve, & using only teams of at-large quality because sometimes auto-bid winners have to be overseeded just to fill the field.

Code: [Select]

Year   Team       Luck   Kenpom Rank   Seed   Overseed Value     Result
2018 Xavier         1        14          1          10        2nd Round Loss
2017 Maryland      39        43          6          19        1st Round Loss
2017 Seton Hall    43        51          9          15        1st Round Loss
2016 Utah           9        28          3          16        2nd Round Loss
2016 Temple        17        90         10          50        1st Round Loss
2015 Maryland       2        32          4          16        2nd Round Loss
2015 Oregon         5        46          8          14        2nd Round Loss
2014 Colorado       3        68          8          36        1st Round Loss
2014 Massachusetts 40        50          8          18        1st Round Loss
2014 Iowa State    43        24          3          12         Sweet 16 Loss
2014 NC State      39        66         12          18        2nd Round Loss
A few things stand out. First, you have to go back 5 years to find any of these teams that overperformed their seed (2014 NC State). Second, only one of these teams made it out of the first weekend, though the team played to seed (2014 Iowa State). For the most part, teams that get into the tournament by winning close games and losing in blowout fashion tend to have their luck catch up with them on the first weekend of the NCAA Tournament. I can see that being something to make Marquette fans nervous as we get closer to Selection Sunday.

However, if we go back one year further, there is a reason for hope. In 2013, there was a team that ranked 11th in luck, was ranked 25th going into Selection Sunday, and was overseeded by 13 when they were placed as a 3-seed. Their luck held, winning two games on the final possession before making it to the Elite Eight and being the last team to overperform their seed under these criteria. So who was that lucky team? None other than Marquette.

[tower912]

The nature of the losses gives me hope.    Losing to IU at Assembly?    The only real surprise was the margin    Losing to occasional #1 Kansas on a neutral floor when they were healthy.    After leading at half time.    Really only playing poorly for one abysmal 10 minute stretch?    Irksome, but in the end, a single digit loss on a neutral floor to a good team.     Losing a to SJU by a large number was the most disturbing.    However, since then, MU has won two road games.    Both in dramatic fashion, one a miracle (ok, lucky).     Here is my take.    MU IS a team that is going to be hard to project come March.    If Markus is hot and the rest of the team is playing confidently, I can easily see a sweet 16.    If, in one of the first two rounds, MU faces a team with a bunch of tall, physical,  versatile guards who can switch onto Markus and contest everything while simultaneously big enough to be able to disrupt Sam and Joey putting the ball on the floor, I can easily see MU losing to that team.   

Until MU has a 'bad' loss, I will remain cautiously optimistic. 

[brewcity77]

Matchups definitely dictate a ton, though seeing first weekend exit after first weekend exit, it did feel a bit like "is this our future?"

This is also predicated on where I have Marquette seeded on the 3-line (though 13 other prognosticators joined me in that prediction the past two days) and the idea of them ending up overseeded.

Being lucky doesn't necessarily stop a team from March success. 2014 UConn was ranked #23 in luck and all they did was win a National Championship from the 7 line. 2015 Duke was #38 in luck and I'm sure we all remember how their season (and Bucky's in the process) ended. The 2016 Oklahoma team went to the Final Four with #29 luck, though they were roughly accurate in kenpom rank to NCAA seed. Also, UMBC last year was #2 in the country in luck and knocked off 1-seed Virginia.

[mu03eng]

Matchups definitely dictate a ton, though seeing first weekend exit after first weekend exit, it did feel a bit like "is this our future?"

This is also predicated on where I have Marquette seeded on the 3-line (though 13 other prognosticators joined me in that prediction the past two days) and the idea of them ending up overseeded.

Being lucky doesn't necessarily stop a team from March success. 2014 UConn was ranked #23 in luck and all they did was win a National Championship from the 7 line. 2015 Duke was #38 in luck and I'm sure we all remember how their season (and Bucky's in the process) ended. The 2016 Oklahoma team went to the Final Four with #29 luck, though they were roughly accurate in kenpom rank to NCAA seed. Also, UMBC last year was #2 in the country in luck and knocked off 1-seed Virginia.

I get the statistics of it all, but I wonder if the Creighton and Georgetown wins don't cancel themselves out. What I mean is that there is no doubt that Creighton was a very lucky win, however Georgetown was almost a bad luck win in that both Morrow and Markus we out so is that really representative of a close win (assuming both come back relatively soon). I have no way of proving it, but I doubt the margin was that close if Markus was 100% let alone if Heldt's minutes go to Morrow instead.

Certainly think this is something to consider though. Curious as to the match-ups in the tournament for the teams you listed....were they overseeded, underseeded, or properly seeded when you compare their KenPom rank to the seed line.

[geps]

Yeah not sure about the Xavier comparison. X had a RPI of 3 or 4 last year whatever that's worth so what about that metric? This year, the NET has us at around 20 which seems right. Also, don't recall X winning a Big East game on the road last year without Bluiett.

[MuMark]

RPI isn't worth much......which is why they got rid of it.

[brewcity77]

I get the statistics of it all, but I wonder if the Creighton and Georgetown wins don't cancel themselves out. What I mean is that there is no doubt that Creighton was a very lucky win, however Georgetown was almost a bad luck win in that both Morrow and Markus we out so is that really representative of a close win (assuming both come back relatively soon). I have no way of proving it, but I doubt the margin was that close if Markus was 100% let alone if Heldt's minutes go to Morrow instead.

That's fair, though I'd point out that MU was 33rd going into Creighton per Pomeroy, so while their luck rank has gone up of late, they were already showing the signs of winning close and losing big.

Certainly think this is something to consider though. Curious as to the match-ups in the tournament for the teams you listed....were they overseeded, underseeded, or properly seeded when you compare their KenPom rank to the seed line.

I didn't look, but based on where they were at going into Selection Sunday...

• 2018 Xavier lost to #36 9-seed Florida State, below average in luck but per Pomeroy accurately seeded.
• 2017 Maryland, coincidentally, lost to #39 11-seed Xavier, who was actually favored slightly per Pomeroy. That Xavier team was #311 in luck and was the Elite 8 squad that we swept the same season.
• 2017 Seton Hall lost to #40 8-seed Arkansas who was even luckier (#9) and almost as badly overseeded (but not quite). Someone had to win.
• 2016 Utah lost to #27 11-seed Gonzaga, who was unlucky (#273) and underseeded. Pomeroy would've had the Zags as the favorite.
• 2016 Temple lost to #19 7-seed Iowa, who was very unlucky (#327) and underseeded. Not sure what's more shocking, that Temple made the field in the first place or that the Owls took Iowa to overtime.
• 2015 Maryland lost to #27 5-seed West Virginia, who was very lucky (#15) and pretty correctly seeded (likely favorites over the Terps).
• 2015 Oregon lost to #1 1-seed Wisconsin at the BC, who was right in the middle on luck (#146).
• 2014 Colorado lost to #17 9-seed Pittsburgh, who was #264 in luck, undoubtedly favored, and insanely underseeded based on Pomeroy.
• 2014 UMass lost to 11th ranked 11-seed Tennessee, who was #341 in luck, certainly favored, and even more underseeded than that Pitt team above.
• 2014 Iowa State lost to 25th ranked 7-seed UConn, who was also lucky at #23, correctly seeded, and went on to win the NC.
• 2014 NC State lost to 35th ranked 5-seed St. Louis, who wasn't quite as lucky (#88) but was also overseeded.

Per Pomeroy, 9 of the 11 teams would have been underdogs in the games they lost. Seems like the opponent luck spread is pretty all over the place, as is the seeding accuracy based on Pomeroy.

[Marcus92]

I've been thinking about this lately, as well. There's another very prominent team that's been very lucky in recent years according to KenPom: Kansas.

This season, Kansas is 15-2 and currently ranked #7 in the AP Top 25 poll. They started the season at #1. KenPom ranks them 9th overall and 21st in Luck. Their record includes a lot of close finishes -- including 2 overtime games and 5 others within 5 points.

So what about past seasons?

In 2017-18, KenPom ranked the Jayhawks #9 after winning the Big 12 conference tournament -- which would be worthy of a 3 seed. But the NCAA committee gave them a 1 seed. Kansas won their first four games of the tournament (including an overtime win and two other games decided by just 4 points), losing to eventual champion Villanova in the Final Four. At season's end, they ranked 23rd in Luck.

In 2016-17, KenPom ranked the Jayhawks #10 after losing in the first round of the Big 12 conference tournament -- which would be worthy of a 3 seed. But the NCAA committee gave them a 1 seed. Kansas won three games before losing to Oregon in the Elite Eight. At season's end, they ranked 32nd in Luck.

In 2015-16, KenPom ranked the Jayhawks #1 after winning the Big 12 conference tournament. The committee gave them a 1 seed. Kansas won three games before losing to eventual champion Villanova in the Elite Eight. At season's end, they ranked 75th in Luck.

In 2014-15, KenPom ranked the Jayhawks #9 after losing in the championship game of the Big 12 conference tournament. The committee gave them a 2 seed. Kansas lost in the second round and finished the season ranked 33rd in Luck.

So is it just luck that Kansas has finished in the 75th percentile of luckiest teams each of the past four seasons -- and in the 90th percentile of luckiest teams 3 out of the past 4 seasons? If Kansas is so lucky, wouldn't you think they'd perform significantly worse than making three Elite Eight appearances and a Final Four?

I'm not quite sure what to make of KenPom's Luck statistic. Seems like it's a catch-all meant to cover everything the system can't explain or doesn't predict very well.

[brewcity77]

Kansas was interesting and I did look at their history, but when you're never outside the top-10, you can't be seeded 10 seed positions lower than your actual position. Ultimately, they generally end up seeded within a line or two of where they should be.

As far as the definition, I think it just shows consistency. Close wins and blowout losses, you're lucky. Close losses and blowout wins, you're unlucky. Play consistently all the time (good or bad) and you'll be in the middle.

I'd definitely rather be lucky than unlucky, but I will admit I'll probably look at seeding with a little more critical eye once that bracket comes out.

[Goose]

There is still plenty of room for improvement in the program. This team has done extremely well in close games and hat’s off to them, but some not so great teams have held close to the boys.I still believe they need to put together a solid 40 minute performance. At this point, I would not be booking hotel rooms for the second weekend of the tourney.

[Marcus92]

As far as the definition, I think it just shows consistency. Close wins and blowout losses, you're lucky. Close losses and blowout wins, you're unlucky. Play consistently all the time (good or bad) and you'll be in the middle.

Maybe it's still too early in the season for the KenPom rankings to be truly predictive. Consider Wisconsin. They were ranked as high as #10 after the road win against Iowa. After that, they snuck by #121 Rutgers 69-64 at home, then lost 5 of their next 8 -- all by 7 points or less. Their wins were blowouts against two sub-300 teams and #64 Penn State.

That's dropped them to #16, with a Luck ranking in the bottom 30 of Division I. Are they really a Top 20 team that's just exceptionally unlucky? The next 5 games or so should be telling.

[wadesworld]

Being good in crunch time is not luck in my opinion. Good teams win close games.

And another issue I have with the KenPoms of the world is beating Southern by 25 instead of 40 says nothing to me personally.

[brewcity77]

So is it just luck that Kansas has finished in the 75th percentile of luckiest teams each of the past four seasons -- and in the 90th percentile of luckiest teams 3 out of the past 4 seasons? If Kansas is so lucky, wouldn't you think they'd perform significantly worse than making three Elite Eight appearances and a Final Four?

I thought a little more about Kansas. It's worth remembering that when you're a high seed, the measure for meeting seed expectations is significantly higher:

• 2015 2-seed, lost in second round, underachieved.
• 2016 1-seed, lost in Elite Eight, underachieved.
• 2017 1-seed, lost in Elite Eight, underachieved.
• 2018 1-seed, lost in Final Four, met expectations.

While Kansas' end season results would be the envy of most programs, they've only met seed expectations once in 4 years and never exceeded expectations. If anything, Kansas reinforces the idea that lucky, overseeded teams tend to underperform their seed come March.

[MU82]

Interesting stuff to talk about.

We'll see if the "pattern" holds over the next 2 months.

Gonna enjoy watching our lads "luck" their way to the Final Four!

[muwarrior69]

I rather be lucky than not. So do teams that are not lucky do better in the NCAA tourney?

[Slim]

Interesting stuff to talk about.

We'll see if the "pattern" holds over the next 2 months.

Gonna enjoy watching our lads "luck" their way to the Final Four!

When you say “lads”, I often wonder if you are wearing a monocle.

[MU82]

When you say “lads”, I often wonder if you are wearing a monocle.

Sometimes I like saying "gallant lads." But those extra syllables make my monocle fall out.

[brewcity77]

Being good in crunch time is not luck in my opinion. Good teams win close games.

And another issue I have with the KenPoms of the world is beating Southern by 25 instead of 40 says nothing to me personally.

But when those teams lose close games in the tournament, does that not make them good teams? Xavier, Seton Hall, Temple, NC State, Iowa State, are they less good because they lost two-possession games in March? There's certainly truth that there is a talent to executing in the closing moments of a tight game, but luck also factors in. Does that contested look rim in or out, does the ref give you the call, is the shot blocker able to get his fingertips on the last ditch attempt or is it just out of reach?

Good teams win close games, but good teams also lose close games. Duke lost two close games this year, they've still a good team. Iowa State lost 3/4 by 5 points or less but has wins over Kansas and Texas Tech, they're a good team. Villanova has 3 losses in games that were decided by 1 possession in regulation, still a good team.

I agree wholeheartedly with the kenpom difference between 25 and 40. The NET did try to fix that with their point cap, but at the same time overlooked the efficiency aspect of it. That needs to be adjusted. I think the best way to handle that is to institute a regressive model. Teams should be rewarded for margin of victory (it's one of the best predictors of future results) but I agree that the difference between a 25 point win and a 40 point win is often just how early you rested your stars. How ugly could Bethune Cookman, North Dakota, and (especially) Southern been if Markus and the starters play 35 minutes? Are we a significantly better team if we win those games by 50 each instead of the 34 average we did win by? Safe to say we both agree the answer is no.

[CTWarrior]

I think this is meaningful and that we haven't been quite as good as our record would indicate.  On the other hand,  I just have a feeling we are better than we've played and that we are going to play better moving forward, as long as Howard and Morrow get healthy and the remainder of our core remains healthy.  We have about 8 weeks of season left.  Lets see if we start winning by more comfortable margins.  I think we will have one more non-competitive loss where we are on the road and just play poorly.

[brewcity77]

I rather be lucky than not. So do teams that are not lucky do better in the NCAA tourney?

I did look at the tealed bit seriously, and it seems the answer is not really. Most of the teams I looked at that were unlucky and incorrectly seeded because of it had greatly varied results, but nothing that seemed to be any clear cut proof. Honestly, if I had any real takeaway, it's that the teams you can most trust are the teams that are in the middle on the luck scale. Teams in the 110-225 range are the ones you know are generally creating their record on the basis of team quality (or lack thereof) rather than the occasional random bounces in tight games. Those are the teams you can most count on when it comes to March.

There are always exceptions...2014 UConn standing out most.