[Cracked Sidewalks] When the Luck Runs Out...

[brewcity77]

Earlier this year, I wrote an article about Marquette's seeding being propped up by unexpectedly good results in close games. Through the end of January, Marquette was 6-0 in two-possession games. Since then, they are just 1-3 in such games. As a result, their luck rating in Pomeroy has dropped from #26 as of the earlier article to #56 now. They still rate as a fairly fortunate team thanks to their 7-3 record in close games, but the last 7 games seem to indicate that Marquette was & is the team Pomeroy's system thought they were.

Continued: http://www.crackedsidewalks.com/2019/03/when-luck-runs-out.html

[GooooMarquette]

Earlier this year, I wrote an article about Marquette's seeding being propped up by unexpectedly good results in close games. Through the end of January, Marquette was 6-0 in two-possession games. Since then, they are just 1-3 in such games. As a result, their luck rating in Pomeroy has dropped from #26 as of the earlier article to #56 now. They still rate as a fairly fortunate team thanks to their 7-3 record in close games, but the last 7 games seem to indicate that Marquette was & is the team Pomeroy's system thought they were.

Continued: http://www.crackedsidewalks.com/2019/03/when-luck-runs-out.html

Actually, Pomeroy's preseason predictions had MU at 19-11 overall and 10-8 in the BE. So regardless of our recent losses, we will end up considerably better than he predicted...not "the team Pomeroy's system thought they were."

https://www.anonymouseagle.com/2018/10/20/18003164/marquette-golden-eagles-basketball-big-east-kenpom-preseason-rankings-projections

[MU82]

In other words, brewski, it's all your fault!

Thanks for owning up to it. Accountability mattas!

[brewcity77]

Actually, Pomeroy's preseason predictions had MU at 19-11 overall and 10-8 in the BE. So regardless of our recent losses, we will end up considerably better than he predicted...not "the team Pomeroy's system thought they were."

https://www.anonymouseagle.com/2018/10/20/18003164/marquette-golden-eagles-basketball-big-east-kenpom-preseason-rankings-projections

It's worth reading the whole article, as it points out that on game-by-game results, Pomeroy had us 22-8 (12-6). With the Louisville game added, it would be 23-8 and alone in second in the Big East, which is pretty darn close to where we are now projected to end up. Certainly closer than what 19-3 (8-1) at the end of January indicated.

Also worth remembering that in preseason, the way the predictive algorithm works trends everyone towards the middle.

I liked the "I swear to Al McGuire, I will sign a contract written in blood for #66 (defensive ranking) in March right now." If there's one thing almost everyone got wrong, it's our defensive capabilities.

[GooooMarquette]

It's worth reading the whole article, as it points out that on game-by-game results, Pomeroy had us 22-8 (12-6). With the Louisville game added, it would be 23-8 and alone in second in the Big East, which is pretty darn close to where we are now projected to end up. Certainly closer than what 19-3 (8-1) at the end of January indicated.

So he predicts overall results and game by game results...so he can take credit for being "right" (or pretty darn close to right) over a wide range of outcomes?

Nostradamus would not be impressed....

[Dr. Blackheart]

Turnovers make teams unlucky.

[muwarrior69]

So teams with bad luck do better in the tournament? Who woulda thunk it.

[SaveOD238]

So he predicts overall results and game by game results...so he can take credit for being "right" (or pretty darn close to right) over a wide range of outcomes?

Nostradamus would not be impressed....

A good prediction system DOES give lots of room for a wide range of outcomes.  The only reason you know Nostradamus is because his predictions were wide-ranging and vague.

A good example is how 538 predicted the 2016 election.  Every other prediction website said Clinton had like a 99% chance of winning because they focused on basically one outcome.  Nate Silver was more open to alternatives (like losing the popular vote but winning the election) and had Trump around 25%.  We all know how that turned out.

[Lennys Tap]

So teams with bad luck do better in the tournament? Who woulda thunk it.

No, but teams who are "over ranked" and "over seeded" because of a great record in close games (i.e., luck) tend to underperform their ranking/seeding in the tournament. Xavier was the poster child last year but several other teams also fit that pattern.

[GooooMarquette]

A good prediction system DOES give lots of room for a wide range of outcomes.  The only reason you know Nostradamus is because his predictions were wide-ranging and vague.

A good example is how 538 predicted the 2016 election.  Every other prediction website said Clinton had like a 99% chance of winning because they focused on basically one outcome.  Nate Silver was more open to alternatives (like losing the popular vote but winning the election) and had Trump around 25%.  We all know how that turned out.

That doesn't change the fact that it's hardly impressive to be "right" with such a vague prediction. Many people treat KenPom like some kind of college hoops sage, when in fact his prediction on MU's season was kinda close because it was vague.

[Silkk the Shaka]

That doesn't change the fact that it's hardly impressive to be "right" with such a vague prediction. Many people treat KenPom like some kind of college hoops sage, when in fact his prediction on MU's season was kinda close because it was vague.

It's literally just an algorithm that takes data inputs and spits out results. It's a nice tool for evaluation of team strengths and weaknesses, and is predictively useful. No predictor will be right 100% of the time, otherwise they wouldn't play the games. There is always uncertainty. But some models will be better than others, his being one of them.

[Marcus92]

KenPom is like weather forecasting. Nobody can predict what's going to happen even 24 hours in advance with perfect accuracy. There are simply too many variables: topography, temperature, atmospheric pressure, moisture levels, cloud cover, wind speed/direction, etc.

But that doesn't make weather forecasting worthless. It's still a valuable tool for calculating likely probabilities. Say meteorologists call for a 90% chance of snow, with an accumulation of 1-3 inches. They don't pull these numbers out of nowhere. It's all based on what actually resulted in the past under similar conditions. If it doesn't snow, that doesn't make the forecast "wrong." It just means the least likely probability occurred.

The same goes for predicting basketball. Think of all the individual factors that can influence the outcome of a single game:

   
• Hundreds of dribbles and passes and dozens of shot attempts, all presenting opportunities for deflections, steals, blocks and rebounds
• Substitutions and matchups
• Whether fouls are seen by the refs, and how the game is called
• Coaching strategy, player execution and adjustments on both sides

And that doesn't take into account things like pre-game preparation, travel, team dynamics, player fatigue/health/injuries, individual motivation or psychology, etc.

So try not to look at KenPom making "correct" or "incorrect" predictions. Probabilities aren't perfect. There are no 100% guarantees. Instead, the statistical model tries to assess the most likely future outcomes (whether it's for the next game or the upcoming season) based on past data (efficiency data for each player and team).

Exhibit A: Virginia vs. UMBC, March 16, 2018. While no 16-seed had ever beaten a 1-seed before in the history of the NCAA tournament, that doesn't mean there was a 0% probability of it happening. Most models give a 16-seed a less than 2% chance to win, or roughly 2 out of 100 games. With only 4 matchups between a 1-seed and a 16-seed each year, we might not see it happen for another 10 or 20 years, maybe even longer. But eventually some team was going to pull off the "impossible." That much is certain.

[GooooMarquette]

KenPom is like weather forecasting. Nobody can predict what's going to happen even 24 hours in advance with perfect accuracy. There are simply too many variables: topography, temperature, atmospheric pressure, moisture levels, cloud cover, wind speed/direction, etc.

But that doesn't make weather forecasting worthless. It's still a valuable tool for calculating likely probabilities. Say meteorologists call for a 90% chance of snow, with an accumulation of 1-3 inches. They don't pull these numbers out of nowhere. It's all based on what actually resulted in the past under similar conditions. If it doesn't snow, that doesn't make the forecast "wrong." It just means the least likely probability occurred.

The same goes for predicting basketball. Think of all the individual factors that can influence the outcome of a single game:

   
• Hundreds of dribbles and passes and dozens of shot attempts, all presenting opportunities for deflections, steals, blocks and rebounds
• Substitutions and matchups
• Whether fouls are seen by the refs, and how the game is called
• Coaching strategy, player execution and adjustments on both sides

And that doesn't take into account things like pre-game preparation, travel, team dynamics, player fatigue/health/injuries, individual motivation or psychology, etc.

So try not to look at KenPom making "correct" or "incorrect" predictions. Probabilities aren't perfect. There are no 100% guarantees. Instead, the statistical model tries to assess the most likely future outcomes (whether it's for the next game or the upcoming season) based on past data (efficiency data for each player and team).

Exhibit A: Virginia vs. UMBC, March 16, 2018. While no 16-seed had ever beaten a 1-seed before in the history of the NCAA tournament, that doesn't mean there was a 0% probability of it happening. Most models give a 16-seed a less than 2% chance to win, or roughly 2 out of 100 games. With only 4 matchups between a 1-seed and a 16-seed each year, we might not see it happen for another 10 or 20 years, maybe even longer. But eventually some team was going to pull off the "impossible." That much is certain.

I get all that...but at least weather forecasters get it better more often than your average Joe who just looks out his window, sees sunshine and assumes it's warm outside.

KenPom - I'm not so sure he gets it any better than your typical knowledgeable fan. His predictions about MU (between 10-8 and 12-6 for the BE, and between 19-11 and 23-8 overall) really aren't any better than most fans predicted. Most of us around here had MU in the top 2 or 3 and expected us to be a top 25 team, and most preseason publications (even the ones without fancy computer algorithms) said the same.

I get that his fancy advanced stats are all the rage, but at the end of the day, he didn't predict us any closer to where we are than I did using my fancy schmancy gut feel algorithm.

[Frenns Liquor Depot]

I get all that...but at least weather forecasters get it better more often than your average Joe who just looks out his window, sees sunshine and assumes it's warm outside.

KenPom - I'm not so sure he gets it any better than your typical knowledgeable fan. His predictions about MU (between 10-8 and 12-6 for the BE, and between 19-11 and 23-8 overall) really aren't any better than most fans predicted. Most of us around here had MU in the top 2 or 3 and expected us to be a top 25 team, and most preseason publications (even the ones without fancy computer algorithms) said the same.

I get that his fancy advanced stats are all the rage, but at the end of the day, he didn't predict us any closer to where we are than I did using my fancy schmancy gut feel algorithm.

Your obsession with this can acually be proven with (gasp) facts and data.  He publishes all his data and for a small fee even more.  Instead of all this gut feel you can squash him by looking at his track record.

Fancy Vegas and bettors would benefit you disproving them too!

[GooooMarquette]

Your obsession with this can acually be proven with (gasp) facts and data.  He publishes all his data and for a small fee even more.  Instead of all this gut feel you can squash him by looking at his track record.

Fancy Vegas and bettors would benefit you disproving them too!

If I was obsessed, I'd get a subscription.

[Marcus92]

KenPom - I'm not so sure he gets it any better than your typical knowledgeable fan. His predictions about MU (between 10-8 and 12-6 for the BE, and between 19-11 and 23-8 overall) really aren't any better than most fans predicted. Most of us around here had MU in the top 2 or 3 and expected us to be a top 25 team, and most preseason publications (even the ones without fancy computer algorithms) said the same.

Keep in mind that Marquette is 1 team out of 351. Think of a bell curve. The x-axis along the bottom represents possible regular season win totals -- from 0 on the far left, to 31 on the far right. The y-axis represents the likelihood of each victory total.

KenPom's preseason win prediction for MU merely represented the median value, the most likely outcome based on the available data. Marquette has clearly over performed that expectation. They're on the leading edge of the bell curve, but probably close to a within a standard deviation of the median. (I'll leave that to an actual statistician.)