Remember this? #odds

[huskies20]

http://fivethirtyeight.com/interactives/march-madness-predictions/

 

[UConnTrumbull]

0.664% to win it all...Always fun to make the "experts" look stupid

 

[Durks07]

DOOOOOOOOOOOOOOOOOOOOOMED!!!

 

[Inyatkin]

Louisville the favorite. Apparently the experts watched at least some UConn games

 

[Leebo]

0.664% to win it all...Always fun to make the "experts" look stupid

How was that wrong? I love the chest pounding revisionist history.

 

[CallMeBruce]

How was that wrong? I love the chest pounding revisionist history.

Because UConn won the championship, ergo Silver's objective statistical model should have given them a 100% chance to win, as it was apparently the only possible outcome. Because intangibles, heart, courage, veteranness, clutchitude.

 

[UConnTrumbull]

How was that wrong? I love the chest pounding revisionist history.

I'm just criticizing the so-called experts. Like "oh no way, all the 1/2 seeds and the defending champions have the best chances to win it all, who would've guessed?!" How many commentators and analysts had the cojones to pick a 7/8 seed title game.

 

[Leebo]

I'm just criticizing the so-called experts. Like "oh no way, all the 1/2 seeds and the defending champions have the best chances to win it all, who would've guessed?!" How many commentators and analysts had the cojones to pick a 7/8 seed title game.

You didn't criticize something nebulous. You criticized the very specific calculation given by a specific statistician.

My guess is you had a higher number in mind. But you didn't tell us what you thought it should be.

Let's do just a little bit of really rough math, and I'll explain why I think nitpicking about the number is stupid.

In the St. Joe's game, we were favorites, I'd say 75% chance of winning that game going into it. It ended up being pretty scary, but that doesn't factor into the initial odds.

Now, for the other 5 games, let's be generous and say UConn had a 50/50 chance in each of them. Remember, when you are saying this, you don't know who those games are against. I think only rabid UConn fans would have agreed to such generous odds for the remaining games.

Do all the multiplication and you get... 2.34%.

And that's being super generous. So what was the number you wanted him to say?

 

[CallMeBruce]

I'm just criticizing the so-called experts. Like "oh no way, all the 1/2 seeds and the defending champions have the best chances to win it all, who would've guessed?!" How many commentators and analysts had the cojones to pick a 7/8 seed title game.

He's not a commentator, and the analysis isn't subjective. No one at 538 "decided" that UConn should have a .664% chance to win the title. It's based on a statistical model using things like the team's season-long performance, which to that point, didn't convince me or any other rational observer they were a title contender.

 

[UConnTrumbull]

You didn't criticize something nebulous. You criticized the very specific calculation given by a specific statistician.

My guess is you had a higher number in mind. But you didn't tell us what you thought it should be.

Let's do just a little bit of really rough math, and I'll explain why I think nitpicking about the number is stupid.

In the St. Joe's game, we were favorites, I'd say 75% chance of winning that game going into it. It ended up being pretty scary, but that doesn't factor into the initial odds.

Now, for the other 5 games, let's be generous and say UConn had a 50/50 chance in each of them. Remember, when you are saying this, you don't know who those games are against. I think only rabid UConn fans would have agreed to such generous odds for the remaining games.

Do all the multiplication and you get... 2.34%.

And that's being super generous. So what was the number you wanted him to say?

Again, I'm not criticizing Silver or whoever did the calculations, I don't really care about math. I'm just saying overall no one gave us a chance of winning and it was fun making them look stupid. That's it.

 

[Leebo]

Did you forget what you wrote? It's still up there. You specifically pointed out Silver's calculation.

You'll have to explain to me how it looks stupid if you want me to agree.

 

[UConnTrumbull]

I didn't say he was wrong. I said 'Always fun to make the "experts" look stupid.' Basically my point is who cares about the statistics? No one picked us even when we kept winning. I'm not looking for you to agree with me, I couldn't really care less.

 

[donyell34]

Needed a crystal ball. Could have made a killing betting U conn.

 

[boba]

Did you forget what you wrote? It's still up there. You specifically pointed out Silver's calculation.

You'll have to explain to me how it looks stupid if you want me to agree.

It looks stupid because it makes specific determinations based on Frequentist assumptions when a Bayesian approach is more appropriate. The very same factors he used to give Villanova a 21% chance (plush travel schedule, close to home) were apparently ignored for UConn. I am not saying it's "stupid" per se, I am saying attempting to justify those conclusions based on those data is unwise. Silver gets away with some really sloppy work that would never get published in any reputable journal because most of us don't understand the mechanisms at work. We remember when he right, but the reality is he's wrong more often than not, and well, when you consider the frequentist formula of 50/50 for being right or wrong, well, he should be right at least 50% of time just on chance alone.

 

[Leebo]

I didn't say he was wrong. I said 'Always fun to make the "experts" look stupid.' Basically my point is who cares about the statistics? No one picked us even when we kept winning. I'm not looking for you to agree with me, I couldn't really care less.

Okay, apparently you are determined to ignore what the topic is about and what you specifically said.

It looks stupid because it makes specific determinations based on Frequentist assumptions when a Bayesian approach is more appropriate. The very same factors he used to give Villanova a 21% chance (plush travel schedule, close to home) were apparently ignored for UConn. I am not saying it's "stupid" per se, I am saying attempting to justify those conclusions based on those data is unwise. Silver gets away with some really sloppy work that would never get published in any reputable journal because most of us don't understand the mechanisms at work. We remember when he right, but the reality is he's wrong more often than not, and well, when you consider the frequentist formula of 50/50 for being right or wrong, well, he should be right at least 50% of time just on chance alone.

Thank you for discussing the actual topic.

 

[CallMeBruce]

It looks stupid because it makes specific determinations based on Frequentist assumptions when a Bayesian approach is more appropriate. The very same factors he used to give Villanova a 21% chance (plush travel schedule, close to home) were apparently ignored for UConn. I am not saying it's "stupid" per se, I am saying attempting to justify those conclusions based on those data is unwise. Silver gets away with some really sloppy work that would never get published in any reputable journal because most of us don't understand the mechanisms at work. We remember when he right, but the reality is he's wrong more often than not, and well, when you consider the frequentist formula of 50/50 for being right or wrong, well, he should be right at least 50% of time just on chance alone.

This is some pretty revisionist history. When he created PECOTA, it immediately became the most accurate tool for projecting baseball performance. The model he used on 538 originally was incredibly accurate at projecting political races. When he scored nearly perfectly in 2012, he even said he got lucky. Read the guy's book; he knows what he's doing. UConn won the title last year, which is great. But there was no logical reason to predict that they would going into the NCAAs.

 

[Marty Jackson]

I'm just criticizing the so-called experts. Like "oh no way, all the 1/2 seeds and the defending champions have the best chances to win it all, who would've guessed?!" How many commentators and analysts had the cojones to pick a 7/8 seed title game.

it's like the pundits telling us 'the only schools worthy of a NC in football are from the P5'... if they keep saying it, it must be true.

 

[UCweCONN]

0.664% to win it all...Always fun to make the "experts" look stupid

Good thing Senhor doesn't listen to the 'experts'.

 

[whaler11]

it's like the pundits telling us 'the only schools worthy of a NC in football are from the P5'... if they keep saying it, it must be true.

Lots of crazy in this thread.

There are 68 teams in the NCAA tourney. In modern college basketball any team at best has a 10-15% of winning. It's 6 single elimination games with an incredible amount of variance.

That UConn has gone on two incredible runs in recent years doesn't invalidate statistical models. UConn was one trip away from being eliminated by St Joe's. Would someone be posting about how bad the model was if Brimah doesn't make that shot?

Silver's NCAA model might be good, and it might not be good. That UConn won the 2014 tournament doesn't validate or invalidate the model in any way.

Yes the model gave Villanova a 21% chance to make the Final Four. So only 4 times out of
5 they don't? Why would the fact they lost in the second round be notable?

The model really liked Kentucky relative to their seed even being bracketed with Louisville... does that mean it's really good?

 

[selles]

Lots of crazy in this thread.

There are 68 teams in the NCAA tourney. In modern college basketball any team at best has a 10-15% of winning. It's 6 single elimination games with an incredible amount of variance.

That UConn has gone on two incredible runs in recent years doesn't invalidate statistical models. UConn was one trip away from being eliminated by St Joe's. Would someone be posting about how bad the model was if Brimah doesn't make that shot?

Silver's NCAA model might be good, and it might not be good. That UConn won the 2014 tournament doesn't validate or invalidate the model in any way.

Yes the model gave Villanova a 21% chance to make the Final Four. So only 4 times out of
5 they don't? Why would the fact they lost in the second round be notable?

The model really liked Kentucky relative to their seed even being bracketed with Louisville... does that mean it's really good?

In fairness to the OP, he was just trying to remind us that we came out of nowhere to win the National Championship 3 months ago. I like being reminded of that. In fairness to Trumbull, he was just chest-thumping because we won the National Championship 3 months ago. I like chest-thumping about that, I'm not sure why it's caused such consternation.

The thread had a chance to turn very interesting once it got into Silver and the value/lack thereof of his NCAA analysis but so far it hasn't. I'm not a statistician and obviously some here know a lot more about the field than I do, so I'd be interested in reading boba/Bruce/leebo/whoever going back and forth about it a bit. The initial back and forth got me to go back and read some of the 538 articles on this year's tourney. I get that Silver is not 'predicting' results per se, that he's trying to analyze probabilities but...it strikes me as kind of a useless exercise. He seems to use most of the same criteria as the seeding committee and he seems to end up with most all of the higher seeds advancing. If you based your pool on his analysis you'd be that guy in the office pool who picks almost no upsets, does well for the first weekend, and flames out badly thereafter.

So...what is the value of what Silver's doing? Is there any? I'm not trying to argue here, I'm interested in reading the thoughts of people who know more about where Silver's coming from than I do.

 

[whaler11]

In fairness to the OP, he was just trying to remind us that we came out of nowhere to win the National Championship 3 months ago. I like being reminded of that. In fairness to Trumbull, he was just chest-thumping because we won the National Championship 3 months ago. I like chest-thumping about that, I'm not sure why it's caused such consternation.

The thread had a chance to turn very interesting once it got into Silver and the value/lack thereof of his NCAA analysis but so far it hasn't. I'm not a statistician and obviously some here know a lot more about the field than I do, so I'd be interested in reading boba/Bruce/leebo/whoever going back and forth about it a bit. The initial back and forth got me to go back and read some of the 538 articles on this year's tourney. I get that Silver is not 'predicting' results per se, that he's trying to analyze probabilities but...it strikes me as kind of a useless exercise. He seems to use most of the same criteria as the seeding committee and he seems to end up with most all of the higher seeds advancing. If you based your pool on his analysis you'd be that guy in the office pool who picks almost no upsets, does well for the first weekend, and flames out badly thereafter.

So...what is the value of what Silver's doing? Is there any? I'm not trying to argue here, I'm interested in reading the thoughts of people who know more about where Silver's coming from than I do.

Is there any real value in trying to predict the outcome of games?

Unless you are gambling on the results... no.

If you are a network that wants people to discuss and watch there certainly seems to be. Specific to the NCAA tourney it wouldn't be the event it became without people trying to predict the results.

 

[huskiesouth]

In fairness to the OP, he was just trying to remind us that we came out of nowhere to win the National Championship 3 months ago. I like being reminded of that. In fairness to Trumbull, he was just chest-thumping because we won the National Championship 3 months ago. I like chest-thumping about that, I'm not sure why it's caused such consternation.

The thread had a chance to turn very interesting once it got into Silver and the value/lack thereof of his NCAA analysis but so far it hasn't. I'm not a statistician and obviously some here know a lot more about the field than I do, so I'd be interested in reading boba/Bruce/leebo/whoever going back and forth about it a bit. The initial back and forth got me to go back and read some of the 538 articles on this year's tourney. I get that Silver is not 'predicting' results per se, that he's trying to analyze probabilities but...it strikes me as kind of a useless exercise. He seems to use most of the same criteria as the seeding committee and he seems to end up with most all of the higher seeds advancing. If you based your pool on his analysis you'd be that guy in the office pool who picks almost no upsets, does well for the first weekend, and flames out badly thereafter.

So...what is the value of what Silver's doing? Is there any? I'm not trying to argue here, I'm interested in reading the thoughts of people who know more about where Silver's coming from than I do.

I am more likely to be selected in the first round of the next NBA draft then acquire any understanding of statisical anaysis - but Silver's models seem to me as a layperson to work well in political races but simply cannot in a sports setting. There are so many variables that affect the outcome of any sporting event that any predictions are wishes at best.

For example we are down 16-4 to Florida - how many of us would have bet the house at that moment that we would prevail. Same against Kensucky - we blew them out of the arena 30-15 but outcome was not decided until a**hole concluded that we would make our free throws in the last 90 seconds, folded his arms and told his players to figure it out on their own.

Taking these two examples and expanding into the entire NCAA tournament I think demonstrates why any level of sport outcome predictions, including Silver's, are great fodder for discussion but in the real world are worth squat.

 

[Scot-e]

This thread has turned into a hot steamy pile.

 

[boba]

I am more likely to be selected in the first round of the next NBA draft then acquire any understanding of statisical anaysis - but Silver's models seem to me as a layperson to work well in political races but simply cannot in a sports setting. There are so many variables that affect the outcome of any sporting event that any predictions are wishes at best.

For example we are down 16-4 to Florida - how many of us would have bet the house at that moment that we would prevail. Same against Kensucky - we blew them out of the arena 30-15 but outcome was not decided until a**hole concluded that we would make our free throws in the last 90 seconds, folded his arms and told his players to figure it out on their own.

Taking these two examples and expanding into the entire NCAA tournament I think demonstrates why any level of sport outcome predictions, including Silver's, are great fodder for discussion but in the real world are worth squat.

IMHO, this is exactly correct. Silver has a hammer, and every problem (question) in a nail to him... Or more accurately, I don't believe Silver is using this tool to predict or reproduce outcomes of experiments, as much as it is to entertain the populace. Much like Dubner and Levitt, Silver has discovered there's gold in providing "answers," even if the answers are dubious and unreliable. We as a species dislike uncertainty, and if you have a believable response to a conundrum, then there's a good chance the majority of the audience will like it.

I'm not a statistician - my job title is Lab Manager/Code Monkey. I use the tools of statistics (R, Matlab) to predict the outcome of experiments, to prove that the outcomes are not the product of chance or other factors, but my real job is in image analysis and acquisition (more tuned to linear algebra and calculus than statistics). It appears to me that Silver is seeing how many times he can use his 'trick' (and trick is a science metaphor for technique or process to minimize confounding data) to to make a believable prediction. You don't see Silver (or his lab mates) running to Las Vegas to place large bets (even hedging them) based on his findings. He makes his money the old fashioned way, he swindles the gullible.

As Selles pointed out, the OP was just commenting how the team "beat the odds," and to that there is no debate. And that guy who let his bet ride in Vegas to the tune of big bucks, he saw the "black swan" on the horizon and followed it to the pot of gold. All science contains just a bit of happenstance (or if you prefer, luck) to succeed. You need the effort, you need the intelligence and knowledge to take advantage of the situation; but you also need just a bit of luck to be at the right place at the right time and be looking in the right direction.

Apologies for adding to the steamy pile.

 

[whaler11]

Well you have no idea if Silver or anyone on his team is making bets based on their models. Tens of thousands certainly used his baseball model to play roto baseball which is really just a form of gambling.

Had you used his model for the World Cup you cleaned up in group play, his model destroyed the betting markets.

99% of people who are gambling seriously and successfully on sports are using statistical models. I don't think he takes it all that seriously anyway that people should be upset. He wrote about the limits of his soccer model (how it dealt with shootouts was a big issue), he does it because people like to look at it - it's not like he writes in a way to jam stuff down your throat.

BTW the new site blows. So shallow.