Oops, 538

[schmendrick]

from today at 2:51 p.m.

Our model, however, now pegs UConn’s chances as essentially the lowest they’ve been in the tournament. UConn finds itself in a virtual dead heat with Baylor (who UConn beat by 11 points in November).

Is UConn Still A Big Favorite?

 

[Bigboote]

Well that was someone's calculated probability. Isn't any more.

These are the same folks who had the presidential election going the wrong way by 2 to 1.

 

[tomcat]

As someone else mentioned, a political poll is quite different from a sports probability scheme. Let's see how the odds are adjusted now that ND and Baylor are outta there.

 

[Topdawg]

Baylor was playing very well till they got beat so I am not surprised by their prediction. Spastically models do not do well with intangibles like a player having the game of her life.

 

[SVCBeercats]

As someone else mentioned, a political poll is quite different from a sports probability scheme. Let's see how the odds are adjusted now that ND and Baylor are outta there.

So they are bad at mathematics two ways.

 

[Kaizen]

Now Benjamin Morris may have difficult time to eat his words ....lol

 

[Zorro]

now 74%. I think they forgot to factor in Kim's penchant for blowing big games.

 

[Kaizen]

FivethityEight's new prediction has UConn win by 74%, USCr 12%, Miss Sate 8% and Stanford 5%.

2017 March Madness Predictions

 

[hoophuskee]

FivethityEight's new prediction has UConn win by 74%, USCr 12%, Miss Sate 8% and Stanford 5%.

2017 March Madness Predictions

Again the disrespect for Stanford. I respect what 538 is doing and it's all in fun anyways. But 5%?

**They did say they think their calculations could be off. I find it interesting.

 

[mntncougar]

I'd say our chances of losing to Baylor have lessened considerably.

 

[Buzzyboy]

I'd say our chances of losing to Baylor have lessened considerably.

Can someone send a sympathy card to Mike?

 

[Wbbfan1]

Totally unrealistic to say UConn is 74% favorite to win it all.

FivethityEight's new prediction has UConn win by 74%, USCr 12%, Miss Sate 8% and Stanford 5%.

2017 March Madness Predictions

 

[oldude]

If SC beats FSU today, I would actually make them a slight underdog vs Stanford. Tara will simply outcoach Dawn. I think 538 has it backwards between the 2 teams.

I do like 538 having the Huskies at 99%+ tonight vs the Ducks.

 

[Glenn]

Something was definitely wrong with 538's model, from a pure statistical sense. Baylor was only 11-3 against top 25 teams, using Sagarin computer rankings. Thus Baylor should have had no better than 78% odds of winning against a typical top 25 team and MissState was better than average. So why 538 had 90% odds for Baylor yesterday was a mystery. With regards to UConn, if you think that Oregon is truly not a top 25 type team, then the odds are in the 99% range. But if you think that UO is in the 10-25 range and that this is a typical UConn team, then the odds should be about 95%, we only lose about 5% of the time to top 25 teams that are not in the top ten.

 

[Goodgood]

Can someone send a sympathy card to Mike?

lol! he is the same guy/girl who posted tons of nonsense about Megan Walker going to ND, blah, blah, blah...

 

[Topdawg]

Baylor was playing very well till they got beat so I am not surprised by their prediction. Spastically models do not do well with intangibles like a player having the game of her life.

Correction..should watch autocorrect.. Replace "spastically" with "statistically" !!

 

[DefenseBB]

If SC beats FSU today, I would actually make them a slight underdog vs Stanford. Tara will simply outcoach Dawn. I think 538 has it backwards between the 2 teams.

I do like 538 having the Huskies at 99%+ tonight vs the Ducks.

I prefer Massey myself as theirs meets the eye test as well as usually getting both the outcome and scores fairly right. If I am handicapping the field, UConn, then MSU as they actually beat two good teams, Stanford, then SC, FSU and Oregon.

UConn has to be in 75-80% vs. Oregon and then 70-75% vs. MSU, SC has to be 55-60% vs. FSU and then 50-55 vs. Stanford.

 

[UConnNick]

I'd say our chances of losing to Baylor have lessened considerably.

I'm still worried about them. Kim is busy this a.m. sticking pins in her Geno voodoo doll.

 

[LasVegasYank]

I'm still worried about them. Kim is busy this a.m. sticking pins in her Geno voodoo doll.

Wouldn't slapping it upside the head be more her style?

 

[bags27]

Something was definitely wrong with 538's model, from a pure statistical sense. Baylor was only 11-3 against top 25 teams, using Sagarin computer rankings. Thus Baylor should have had no better than 78% odds of winning against a typical top 25 team and MissState was better than average. So why 538 had 90% odds for Baylor yesterday was a mystery. With regards to UConn, if you think that Oregon is truly not a top 25 type team, then the odds are in the 99% range. But if you think that UO is in the 10-25 range and that this is a typical UConn team, then the odds should be about 95%, we only lose about 5% of the time to top 25 teams that are not in the top ten.

Interesting analysis: thanks!
Let's remember that in the one modelling game where people pour unlimited resources and intelligence and for which there is now nearly 200 years of extremely rich data--the stock market--it is still not a reliable predictor, certainly not in the short term. And in macro economics generally, the Fed's predictors of long term economic trends are constantly shifting and remain at best an approximate game.

Compare all that to predicting basketball, with relatively very little data, but infinite variables. For example, in what you cite above about records against ranked teams, you don't factor in won-lost on home, away, and neutral courts. And yet, what exactly is a neutral court? But the sampling available is suspect because we don't know how to weigh games early in the season vs. late in the season, or against teams that are improving or regressing, etc.

Modelling sports outcomes over a single season is like modelling the weather: we know with absolute certainty that somewhere in New England it's going to rain today...most likely....probably....possibly....but where and when it's going to rain exactly, well, everyone better bring an umbrella from CT to Maine.

 

[Kaizen]

I'm still worried about them. Kim is busy this a.m. sticking pins in her Geno voodoo doll.

Kim: It is all UConn's fault caused our loss

 

[Phil]

So they are bad at mathematics two ways.

Actually they are quite good at mathematics.

And while it is easy to throw brickbats and note they got the presidential election wrong, virtually ALL the major pundits got it wrong, and 538 had it closer than almost anyone.

 

[SVCBeercats]

Actually they are quite good at mathematics.
And while it is easy to throw brickbats and note they got the presidential election wrong, virtually ALL the major pundits got it wrong, and 538 had it closer than almost anyone.

OH! I didn't know they were playing horseshoes. As professor Dudsinsky told me with his wonderful Polish accent, when I complained I lost points on a calculus test due to faulty arithmetic even though the calculus was perfect; "My good young man, the airplane, it did not fly!"

 

[Phil]

Something was definitely wrong with 538's model, from a pure statistical sense. Baylor was only 11-3 against top 25 teams, using Sagarin computer rankings. Thus Baylor should have had no better than 78% odds of winning against a typical top 25 team and MissState was better than average. So why 538 had 90% odds for Baylor yesterday was a mystery. With regards to UConn, if you think that Oregon is truly not a top 25 type team, then the odds are in the 99% range. But if you think that UO is in the 10-25 range and that this is a typical UConn team, then the odds should be about 95%, we only lose about 5% of the time to top 25 teams that are not in the top ten.

538 might have it wrong, but you claimed it is "definitely wrong". I don't see your analysis as persuasive. It's a simple heuristic, but far too simple to draw such a conclusion. Baylor was second only to UConn in Massey strength ratings. Their site is down at the moment, but I believe they had a probability of winning in the 70's, which is lower that 538, but you are suggesting that even 70's is too high.

On a positive note, I do like the approach of using a simplified model to test reasonableness.

 

[Phil]

Compare all that to predicting basketball, with relatively very little data, but infinite variables. For example, in what you cite above about records against ranked teams, you don't factor in won-lost on home, away, and neutral courts. And yet, what exactly is a neutral court? But the sampling available is suspect because we don't know how to weigh games early in the season vs. late in the season, or against teams that are improving or regressing, etc.

.

This is a perfect example of why modeling is difficult. And Oregon is a case in point. They have some losses to some weak teams. So how much do we weight that in? I think it is relevant that the Oregon team is so very young, starting three freshmen and relying heavily on one freshmen. You always expect the team to get better over the course of the year but all of the things being equal, a team heavily dependent on seniors and juniors is less likely to improve over the course of the season the team heavily dependent on freshmen. Thus, I think it makes sense to wait the early-season losses against Oregon less than you might for someone else.

Not that we need any more nails in the coffin but this is one more flaw in the RPI which places exactly the same weight on the opening game result as it does against the most recent result. I am sure the 538 model reflects recency. I suspect Massey does as well, but it is virtually certain that their weighting schemes are different.