UConn Now Almost Even Money to Win The Tournament!!

[Fairfield Fan]

Nate Silver and the 538 Blog now has UConn as only a slight favorite to win the national championship, with a 43% chance to take the title. He has Baylor with with a 41% chance of winning.

Given the way Baylor has dissected all comers in this tournament, I don't think Silver's being unfair.

2017 March Madness Predictions

He has Notre Dame and South Carolina with single-digit odds of taking it all.

 

[bags27]

Nate Silver and the 538 Blog now has UConn as only a slight favorite to win the national championship, with a 43% chance to take the title. He has Baylor with with a 41% chance of winning.
Given the way Baylor has dissected all comers in this tournament, I don't think Silver's being unfair.
2017 March Madness Predictions
He has Notre Dame and South Carolina with single-digit odds of taking it all.

That's DOWN from 48% before yesterday's game and down from 52% from the start of the tournament. So UConn's play the past 3 games has actually lost them nearly 10% chance of winning the title. I think so little of 538 that I like to track it out of a sense of irony.

Currently, 538 gives UConn a greater than 99% chance of beating Oregon and USC an almost 90% chance of beating Miss State. Not a betting person, because it conditions my enjoyment of the actual event, but otherwise I'd happily put a few bucks on the upsets here.

 

[hoophuskee]

That's DOWN from 48% before yesterday's game and down from 52% from the start of the tournament. So UConn's play the past 3 games has actually lost them nearly 10% chance of winning the title. I think so little of 538 that I like to track it out of a sense of irony.

Currently, 538 gives UConn a greater than 99% chance of beating Oregon and USC an almost 90% chance of beating Miss State. Not a betting person, because it conditions my enjoyment of the actual event, but otherwise I'd happily put a few bucks on the upsets here.

538 has been dropping UCONN because of Baylor. And you meant Baylor beating Miss State, right?

I really can't see UCONN losing to Oregon whether it be 1 point or 50 points -- can't see it. Maryland was overrated and Duke had their pg missing and the Duke coach isn't so hot.

What do you do if you have statistical data and it shows Baylor is really good too? And how much money would you put on UCONN vs Baylor? What odds are you going to give the ones who take Baylor?


UConn Won By 61 But Their Chances Of A Five-Peat Dropped

 

[bags27]

538 has been dropping UCONN because of Baylor. And you meant Baylor beating Miss State, right?
I really can't see UCONN losing to Oregon whether it be 1 point or 50 points -- can't see it. Maryland was overrated and Duke had their pg missing and the Duke coach isn't so hot.
What do you do if you have statistical data and it shows Baylor is really good too? And how much money would you put on UCONN vs Baylor? What odds are you going to give the ones who take Baylor?
UConn Won By 61 But Their Chances Of A Five-Peat Dropped

Thanks. Yes slipped: Baylor
I take you point, but wonder about your last statement about "odds". That sounds like the way Vegas figures things. I was under the impression that 538 doesn't consider the question of stimulating betting or balancing the flow of bets, but rather was based purely on some probability algorithm.

 

[hoophuskee]

Nate Silver and the 538 Blog now has UConn as only a slight favorite to win the national championship, with a 43% chance to take the title. He has Baylor with with a 41% chance of winning.

Given the way Baylor has dissected all comers in this tournament, I don't think Silver's being unfair.

2017 March Madness Predictions

He has Notre Dame and South Carolina with single-digit odds of taking it all.

I wonder if Massey might be better though. Yesterday they predicted 80-66 UCONN. Almost nailed it. OFC they didn't predict Oregon winning. Massey has today ND vs Stanford a 1 point victory for ND with ND at 52% overall chance. And 538 has a 75% chance that ND wins. I think 75% is way too high for ND. As bags says - give me those odds for me for ND vs Stanford that 538 is giving and I'd take Stanford. I'm not saying Stanford will win -- but 75% ND wins without Turner?

Massey Ratings - CBW

 

[hoophuskee]

Thanks. Yes slipped: Baylor
I take you point, but wonder about your last statement about "odds". That sounds like the way Vegas figures things. I was under the impression that 538 doesn't consider the question of stimulating betting or balancing the flow of bets, but rather was based purely on some probability algorithm.

Okay -- I didn't know that.

Thanks!

 

[hoophuskee]

Nate Silver and the 538 Blog now has UConn as only a slight favorite to win the national championship, with a 43% chance to take the title. He has Baylor with with a 41% chance of winning.

Given the way Baylor has dissected all comers in this tournament, I don't think Silver's being unfair.

2017 March Madness Predictions

He has Notre Dame and South Carolina with single-digit odds of taking it all.

Just wanted to reiterate -- Massey predicted 1 point win for ND over Stanford (71-70) with a 52% chance of ND winning vs 48%.

538 has ND winning 75% of the time.

 

[tomcat]

Let's see how the Stanford win changes the odds.

 

[laika]

I wonder if Massey might be better though. Yesterday they predicted 80-66 UCONN. Almost nailed it. OFC they didn't predict Oregon winning. Massey has today ND vs Stanford a 1 point victory for ND with ND at 52% overall chance. And 538 has a 75% chance that ND wins. I think 75% is way too high for ND. As bags says - give me those odds for me for ND vs Stanford that 538 is giving and I'd take Stanford. I'm not saying Stanford will win -- but 75% ND wins without Turner?

Massey Ratings - CBW

It's more relevant that Massey has been "massey"vely underrating Baylor. Baylor might not be an actual co-favorite, but I think Silver is assessing the title odds much more accurately than Massey.

 

[Fairfield Fan]

I wonder if Massey might be better though. Yesterday they predicted 80-66 UCONN. Almost nailed it. OFC they didn't predict Oregon winning. Massey has today ND vs Stanford a 1 point victory for ND with ND at 52% overall chance. And 538 has a 75% chance that ND wins. I think 75% is way too high for ND. As bags says - give me those odds for me for ND vs Stanford that 538 is giving and I'd take Stanford. I'm not saying Stanford will win -- but 75% ND wins without Turner?

Massey Ratings - CBW

Does he have a score prediction for UConn/Oregon?

 

[OldAlum]

Has UConn at about 78-62 (You can adjust for more chance or more "chalk")

 

[hoophuskee]

It's more relevant that Massey has been "massey"vely underrating Baylor. Baylor might not be an actual co-favorite, but I think Silver is assessing the title odds much more accurately than Massey.

Why do you say that?

I've given an example what I felt was an obvious point that with Stanford vs ND - with Turner hurt - 538 had ND winning 75% of the time while Massey had ND winning just 52% of the time. They predicted a 1 point game in the 70's. SO Massey was far superior in this instance, correct?

What can you give me as an example why you think Massey is underrating Baylor? I'm not arguing with you. Just looking for an example. I started a thread a while ago on this subject and openly questioned how Silver/538 could be saying Notre Dame without turner would beat Stanford 75% of the time.

 

[hoophuskee]

Does he have a score prediction for UConn/Oregon?

I'm not sure if the below is accurate. Or if new numbers will derived tomorrow. I'm new to all of this. I hardly look at this. But looks like 82-63.


Massey Ratings

 

[laika]

Why do you say that?

I've given an example what I felt was an obvious point that with Stanford vs ND - with Turner hurt - 538 had ND winning 75% of the time while Massey had ND winning just 52% of the time. They predicted a 1 point game in the 70's. SO Massey was far superior in this instance, correct?

What can you give me as an example why you think Massey is underrating Baylor? I'm not arguing with you. Just looking for an example. I started a thread a while ago on this subject and openly questioned how Silver/538 could be saying Notre Dame without turner would beat Stanford 75% of the time.

Massey shows UConn as extremely far ahead of Baylor in 2nd place, with Baylor being ranked close to the same as the other top 10 teams.
He had predicted them to win their last two games by 19.5 and 8.5- Baylor won by 40 and 34.
Also, I think Baylor would have had a few points in their favor because it would have been basically a home game.

Maybe UConn really is massively better than Baylor or any other team. Or maybe Baylor was underrated but Miss St played the game of their lives. I agree with Massey now that UConn should steamroll to the title against anyone else left.

 

[hoophuskee]

(1)Massey shows UConn as extremely far ahead of Baylor in 2nd place, with Baylor being ranked close to the same as the other top 10 teams.
(2) He had predicted them to win their last two games by 19.5 and 8.5- Baylor won by 40 and 34.
Also, I think Baylor would have had a few points in their favor because it would have been basically a home game.

(3) Maybe UConn really is massively better than Baylor or any other team. (4) Or maybe Baylor was underrated but Miss St played the game of their lives. I agree with Massey now that UConn should steamroll to the title against anyone else left.

(1) -- And how was this proven to be wrong?

(2)-- Okay -- but you said 538 was better than massey. Even 538 is suggesting their model is off per the below link and the quotes I've provided. But I see what you're saying /you've used the 34 and 40 as a better tool that Baylor should have been closer to UCONN. I seriously questioned it because of the discrepancy between Notre Dame and Stanford when 538 had them as 75%. Take a look at the thread created by poster wbbfan asking predictions in the region and nearly everyone is saying it will be tight. I realize all 75% of ND's wins could have been tight-- but I still think it way way way off. Even now I'm telling you they are seriously underestimating Stanford. Five percent?

(3) We'll never know.

(4)- Could be. Tulane this year might have played "the game of their lives too." But I'm skeptical of "game of their lives." UCONN was awful, an injury they weren't yet prepared for, probably overlooked the game a bit and the refs made some horrific calls that impacted the flow. So in general I'm skeptical of the game of their lives excuse. We've heard this before with Baylor, haven't we?
I recall when the UCONN Men won their 1st title, in the final four they played Ohio State led by sharpshooters Scoonie Penn and Michael Redd. Later UCONN beat Duke. Kind of funny that when Rickey Moore and Kevin Freeman were guarding the opposition - the other team never seemed to have the game of their lives.
Therefore I dont buy the game of their lives excuse. And I think it more likely that Baylor wasn't that good, and their conference wasn't that good. And imo the comments from 538 themselves seems to confirm that there should be some skepticism of what 538 was showing with Baylor vs UCONN.

Is UConn Still A Big Favorite?

"Interestingly, however, the markets don’t seem to share our model’s enthusiasm for Baylor. The latest prices that I could find still give UConn a greater than 75 percent chance (for example, here, where the odds of -420/+300 imply fair odds of 76 percent), which is similar to what was out there before the tournament.

The huge disparity between our model and the markets suggests that it’s worth having some skepticism toward our model results. In general, models are at their worst when it comes to extreme outliers, which UConn certainly has been. For one thing, it can be hard to tell their true strength, because so many of their games are fairly noncompetitive."

 

[UcMiami]

I think 538 model has used point differential in the first two rounds as to big an indicator of current strength - an 89 point win was deemed stronger than a 60 point win, a 40 point win in round 2 as significantly different from a 30 point win. In those two rounds anything over 30 points in the first round or 25 in the second round for a 1 or 2 seed is more about bench strength and coaching/play philosophy than it is about comparative strength of top teams (and perhaps vagaries of seeding choices.) But the 538 model looked at those two games and saw a swing of almost 10% in tightening the race. Whether they were correct in their original percentages or not, the changes after the first two rounds were an over reaction for the women's tournament (this is an off-shoot of the men's model where it might not be quite so much of an over reaction.)

The reaction after the third round further tightened the percentage almost to 'pick'em' - that they moved based on the MOV disparity (34 v 15) is not wrong, but that change on top of the previous changes resulted in a tighter race than it probably should have been.

It also seems to me that 538 over weights chalk in fairly even match-ups, so that they see a Stanford/ND match-up that chalk says favors ND by maybe 5 points as being more of a certainty, discounting the teams 'betas' for performance. That was probably compounded by a crucial injury not being factored into the model that likely increased ND's beta.

In watching the in game percentage I have decided it really isn't dynamic in any real sense - Baylor tied with MSU late in the fourth had Baylor's win probability at basically the exact same value as at the beginning of the game, whereas by that time it was pretty clear this was much closer to a pure toss-up because Baylor was not playing up to 'expectation' and MSU was surpassing their 'expectation'. MSU with a two point lead late was still seen as the underdog. The same thing was true in the ND/Stanford game.

 

[hoophuskee]

I think 538 model has used point differential in the first two rounds as to big an indicator of current strength - an 89 point win was deemed stronger than a 60 point win, a 40 point win in round 2 as significantly different from a 30 point win. In those two rounds anything over 30 points in the first round or 25 in the second round for a 1 or 2 seed is more about bench strength and coaching/play philosophy than it is about comparative strength of top teams (and perhaps vagaries of seeding choices.) But the 538 model looked at those two games and saw a swing of almost 10% in tightening the race. Whether they were correct in their original percentages or not, the changes after the first two rounds were an over reaction for the women's tournament (this is an off-shoot of the men's model where it might not be quite so much of an over reaction.)

It also seems to me that 538 over weights chalk in fairly even match-ups, so that they see a Stanford/ND match-up that chalk says favors ND by maybe 5 points as being more of a certainty, discounting the teams 'betas' for performance. That was probably compounded by a crucial injury not being factored into the model that likely increased ND's beta.

QUOTE]


Thanks UC. I expected the 1st paragraph. But what do you mean by chalk regarding ND and Stanford?

 

[UcMiami]

By chalk - I mean the projected score kind of thing from Massey - They decided ND was the likely winner by a 1.5 points (chalk) and Massey then assigns a percentage to that at say 52% which is based on their models beta value for ND and beta value for Stanford which are classic bell curves of possible point totals by each team against the others 'average' defense. How those two bell curves overlap determines the number of times ND will in fact win over an infinite series of contests. So because they think 52 times out of 100 ND will score more points they assign 52%.

Massey gives both the projected point spread, and then a percentage. 538 only gives us the percentage from their calculations so we don't know what their 'chalk' is but I doubt it is too far out of line with Massey (which is one of their sources I believe.) But it appears that 538 assigns a significantly lower beta of possible scores to determine that a fairly tight contest points wise is still much more likely to be the result. Massey doesn't get up to 75% on possible outcomes for most matches unless their point spread is around 8 (73% for Baylor with a 7.5 point spread prediction.) I don't believe 538 would have thought ND was +7 against Stanford and yet they project a 75% win probability to ND.

Actually just looked up the definition of 'Chalk' in betting - it is simply the favorite. Which works in my long winded post above: 538 determines who the favorite is and assigns a greater weight to that favorite, than most systems that look at the team betas and the point spread normally do. (Much shorter and more succinct than the above! )

 

[Jimbo]

I think 538 model has used point differential in the first two rounds as to big an indicator of current strength - an 89 point win was deemed stronger than a 60 point win, a 40 point win in round 2 as significantly different from a 30 point win. In those two rounds anything over 30 points in the first round or 25 in the second round for a 1 or 2 seed is more about bench strength and coaching/play philosophy than it is about comparative strength of top teams (and perhaps vagaries of seeding choices.) But the 538 model looked at those two games and saw a swing of almost 10% in tightening the race. Whether they were correct in their original percentages or not, the changes after the first two rounds were an over reaction for the women's tournament (this is an off-shoot of the men's model where it might not be quite so much of an over reaction.)

The reaction after the third round further tightened the percentage almost to 'pick'em' - that they moved based on the MOV disparity (34 v 15) is not wrong, but that change on top of the previous changes resulted in a tighter race than it probably should have been.

It also seems to me that 538 over weights chalk in fairly even match-ups, so that they see a Stanford/ND match-up that chalk says favors ND by maybe 5 points as being more of a certainty, discounting the teams 'betas' for performance. That was probably compounded by a crucial injury not being factored into the model that likely increased ND's beta.

In watching the in game percentage I have decided it really isn't dynamic in any real sense - Baylor tied with MSU late in the fourth had Baylor's win probability at basically the exact same value as at the beginning of the game, whereas by that time it was pretty clear this was much closer to a pure toss-up because Baylor was not playing up to 'expectation' and MSU was surpassing their 'expectation'. MSU with a two point lead late was still seen as the underdog. The same thing was true in the ND/Stanford game.

Nailed it, particularly about 538's overreaction to Baylor's early-round blowouts. After the first round alone, Baylor's chances of winning the tournament, according to 538, spiked from 23% to 32%. Because of a game they played against Texas Southern. Which is preposterous.

UConn Won By 61 But Their Chances Of A Five-Peat Dropped

I appreciate your comments about the in-game percentages, as I'd never looked at them in that level of detail before. I had noticed earlier that it takes a while before an underdog with a lead cracks the 50% probability threshold (the "chalk effect"). I think that's reasonable, at least to a point; certainly a heavy favorite who's trailing 2-0 in the first minute doesn't have a <50% chance of winning at that time. But it may be that 538's model is too slow to react to in-game developments. Its underlying assumption seems to be that each team's performance will regress to the mean, so it probably won't recognize in time that a team is having an especially good or bad night, and I'd expect it to have a very hard time making sense of the performance of an unpredictable, high-beta team like Tennessee.

What I find most interesting is the fact that the pregame percentages for the MSU-Baylor game were virtually the same as when the game was tied late. That strikes me as curious. Even if you disregard what's already happened in the game to that point, and continue to believe Baylor is far superior to MSU, the probability that MSU can outscore Baylor for (let's say) 5 minutes must be higher than the probability that MSU can outscore Baylor for 40 minutes. Fewer possessions means a smaller sample size and thus a lower probability that the chalk wins out.

 

[Ouremblem]

Massey has UConn winning tonight 81-63 with 95% probability. They have the South Carolina game close with only 58% probability, SC winning 71-68.

 

[Fairfield Fan]

Wow!! Talk about being behind the curve. Here's a feature article on UConn's chances from 538:

Is UConn Still A Big Favorite?

"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). Though Baylor has seemed to be UConn’s most formidable competition since before the tournament began, their rise has nonetheless been dramatic. They’ve won their tournament games by 89, 40 and 34, compared with UConn’s margins of 61, 30 and 15."

Yep, got it.

 

[Fairfield Fan]

The tournament bracket on 538 blog, however, now shows UConn with a 74% chance of winning it all:

2017 March Madness Predictions

According to Fivethirtyeight, Oregon has less than a 1% chance of defeating UConn. Not sure I'm putting a lot of faith in that model....

 

[bags27]

Massey has UConn winning tonight 81-63 with 95% probability. They have the South Carolina game close with only 58% probability, SC winning 71-68.

Well, this pretty much sums up the unreliability of these percentages. If a predicted margin of victory is within 3 points, we fans know that the probability of victory is really 50-50. There are just far too many variables in a close game to predict the actual outcome.

Further, what cannot be weighed, because the sampling is close to zero and team composition changes annually, is how teams perform under the pressure of advancing to the next level within the tournament. Some teams play better, some worse, but that is only discovered at that very moment and is essentially unrepeatable data: a team that might play terrifically in the elite 8, which is a regional game with maybe 3500 in attendance, might completely collapse under the lights of a final four game. These things happen both absolutely and relatively all the time, and yet isn't usable date for future calculations.

 

[UcMiami]

Well, this pretty much sums up the unreliability of these percentages. If a predicted margin of victory is within 3 points, we fans know that the probability of victory is really 50-50. There are just far too many variables in a close game to predict the actual outcome.

Further, what cannot be weighed, because the sampling is close to zero and team composition changes annually, is how teams perform under the pressure of advancing to the next level within the tournament. Some teams play better, some worse, but that is only discovered at that very moment and is essentially unrepeatable data: a team that might play terrifically in the elite 8, which is a regional game with maybe 3500 in attendance, might completely collapse under the lights of a final four game. These things happen both absolutely and relatively all the time, and yet isn't usable date for future calculations.

But a +2 point spread isn't really 50% and good models recognize that - Massey looks at that and comes in at 54/46 or something like that, and uses team specific beta values so Uconn at +3 might be 60%, while TN +3 might be 53%. Fans and a lot of betters look at 60% and go 'we got this in the bag' while realists look at 60% and realize they are only 1 game in five above 50%. And every better who bets the favorite in a ten horse race every time is walking home shirtless!

 

[bschwartz]

Another factor you have to account for is how familiar a team is with its competitor. MD and Baylor would have had a benefit against UConn in the tourney because they played the Huskies this season (though that would be counterbalanced by the fact that UConn has played them too). Oregon is a young team who hasn't seen anything like the Huskies. IMO, that matters and would impact my percentages.

If UConn does get past the Ducks, Baylor would have been a real bear to deal with. I think they would have focused on getting Gabby and Pheesa in foul trouble based on what they learned in November. MS State players also have some experience with last year's Huskies having been butt whupped by them in the tourney. So not as much benefit as having played UConn this year but some.

As an aside, you really need to like State's Moriah Jefferson Morgan William. What an amazing performance. If it happens, can wait to see who Geno puts on her defensively. As I watched her performance, I envisioned Crystal playing the same way some day. And yeah, my eyes got a little sweaty when she was talking about her dad.