Nate Silver's projections
- Thread starter ChiTownLion
- Start date
The ELO ranking was interesting--a form of that's been used in chess for decades (I was once a tournament player--though not at that high a level). That puts Wisconsin at No. 3
So Nate thinks we have a 22% chance of making the CFPs. That is a better chance than Nate gave to Donald Trump to beat Hillary Clinton. Nate is having an off year.
Nate had Hillary at a 71.4 chance to win on election eve.
Why? That 22% was higher than the market or pretty much any other prognosticator. If you followed him, you won $.
Yes, Silver's web page is nicely constructed. You can see the odds change as you pick winners.
That's all great, but he's still mind reading. If this is all based on the "eye test" as a final determinant, can someone explain to me how this is different than relying on polls?
If your point is that he was way off, 28.6% for Trump to win is still pretty high. I think in most people's minds, the chances were less than 5%. Trump just hit his 1 in 4 chance.
I'm kicking myself for not betting 1x on Trump when the odds hit 8 to 1 and then betting 4x on Hillary at 1 to 3.
I prefer these projections:
http://www.cbssports.com/college-fo...errated-underrated-predicted-order-of-finish/
http://www.cbssports.com/college-fo...errated-underrated-predicted-order-of-finish/
My point was he, like most other pollsters, was way off (better, but still off). Hopefully it will be so here.
At least the Presidential polls were supposedly random samples of about 1000 likely voters taken from the US population. Thus, there was a statistical basis for the results. It appears that this is just Silver's guesses. He assigned win percentages to each game and then guessed at how the selection committee will respond on Sunday.
Last edited:
How do you know he was way off? He gave Trump a 28% chance to win. That's like a solid MLB batter getting a hit.
If that is standard of predictive accuracy we expect from our pollsters today, I am in the wrong profession.
What predictive accuracy? You can't actually measure predictive accuracy based on a single event. If you want to evaluate his accuracy, you have to look at all of his predictions and see how the results compare to what was predicted.
Ok. If that is the type of predictive accuracy we expected from our pollsters in the 2016 Presidential Election I am in the wrong profession. I told my wife on election eve that based upon all I knew, I gave Trump a 45% chance of winning. I like Nate but he, like most, missed the mark. If you want to give him a pass because he did better than others, fine. I will not.
Again. How did he miss the mark? You are looking at his prediction and interpreting it as "Hillary is almost definitely going to win". That's not the case. He gave Trump better than 1 in 4 chance to win the election. When I play poker I gamble on 1 in 4 odds all the time (depending on the pot). It's not really that much of a long shot.
What's really interesting are the stats if Clemson or Washington loses and Penn State wins. If Clemson loses, Penn State has a 77% chance, and if Washington loses a 66% chance (due to the increased chance of Colorado making it). In either scenario, Michigan only has a 10% chance. That's very surprising to me, but hope it's right.
Exactly:
50 / 50 to win tomorrow
50 / 50 that either UW or Clemson loses
= 25% chance (roughly) of Final Four
Which is what "we" determined last week
He gives us a better chance than Wisconsin and he gives Michigan virtually no chance.
The ELO ranking is always reliable for great songs...Telephone line, Strange Magic, Turn to Stone, etc...
I am not interpreting it that way at all. Let me put it this way. I have been involved in politics for nearly 50 years and have helped raise money for candidates. Usually a potential donor will ask what chance the candidate has to win. If I tell them "about 25%" I usually won't get a dime. If I tell them 75%, the money will flow. Nate, like most, blew it. You feel otherwise. We just don't agree.
Exactly. As soon as they introduce the "eye test" and talk about OSU, for instance, as definitely being a top 4 team (without the games even being played) they eliminate the logic of the process and are left with something no better than the polls.
Ok. I'm going to make this simpler. You can't be wrong when you make a prediction based on percent probability and there is only one event. The only way you can be wrong is if that event is run over and over again showing that your percentages were off. You could be right and his percentages may have been off, but it is not possible to declare that based on one single event.
Therefore coming to conclusions like "Silver was wrong" or "Silver was way off" is not possible in this situation. There is no way to measure and prove it.
To take it a step further, it's actually much worse than relying on polls because you have a lot fewer people making the decisions.
That's nice, ...
.... even if you don't have a clue about prediction performance assessment.
Let me also make this simpler. If Silver were working for either Clinton or Trump and they relied on his 2016 Presidential Prediction to their detriment, he would be out of a job.
No, but I know a hell of a lot about politics. Time to move on. I heard there is a football game tomorrow.
A dozen folks - or whatever it is - most of whom have some idea of what they are looking at......and, at the least, spend some time, and thought, and deliberation, and sharing of thoughts and ideas, on it (not that they are EVER going to take a subjective process and get it "unequivocally" right)
Versus
50 or 60 or 70 folks - most of whom haven't got a clue .......none of whom actually spend any time evaluating any team except the one they cover (as a media "folk"), or the team they coach (or, in most cases, the team they fold the laundry for - since most coaches don't even fill out their own "ballot")
Uh......given the horrendously idiotic parameters of the system.......I think I'd be pretty comfortable in assuming the "committee" is the MUCH lesser of two evils
That has nothing to do with this discussion. You are trying to assign accuracy to a situation where you can't. You are trying to say Silver was wrong, but there is no wrong. I'm not sure why that is so hard to understand.
Another pet peeve of mine. I love when people complain about weathermen being wrong that one time out of ten and forget the other 9 times that they nailed it. They are predicting the freaking future and people think they are idiots when the miss one prediction.
If I told you there was a 16% chance of rolling dice once and hitting a six, would you tell me I was way off if it hit?
You can't compare an election to dice and weather. Trump vs Clinton 2016 only happens once. I guess they could have had people vote 100 times that day but the results would still have been the same. There's nothing else to test that model against. Yes 25% is still a chance but so is 1%.
Silver's models are very sophisticated and were off this time. By a lot. So were eveybodies.
Similar threads
