MN Sen Recount (UPDATE: Stuart Smalley certified winner, lawsuit forthcoming)

[Keystone Phil]

What was Nate Silver's prediction, again?

It depends when you asked him.

[Torie]

What was Nate Silver's prediction, again?

It depends when you asked him.

He's projecting Franken wins by 40 votes now, excluding the absentee ballots that may have been illegally rejected.

[Alcon]

What was Nate Silver's prediction, again?

It depends when you asked him.

I still don't understand the idea of objecting to dynamic predictions.  I think that ignoring incoming evidence in favor of maintaining a pre-existing decision is pretty much responsible for half the bad in politics these days.

[Keystone Phil]

What was Nate Silver's prediction, again?

It depends when you asked him.

I still don't understand the idea of objecting to dynamic predictions.  I think that ignoring incoming evidence in favor of maintaining a pre-existing decision is pretty much responsible for half the bad in politics these days.

Ok, I'll hold off until the final count is certified and when it's not the exact number that His Highness predicted, we'll hear what excuses His followers have for me.

[Torie]

It is an estimate Phil, and subject to being higher or lower. It is just the top of the bell curve of possibilities. Bell curves have tails. The number is also based on certain assumptions which are explicated, and if those assumptions are wrong, then the bell curve numbers may be wrong. The exercise is a statistical tool for predicting the odds of a result occurring; it is not a crystal ball. Statistics are a beautiful thing Phil, elegant and sometimes illuminating and useful.

[emailking]

What was Nate Silver's prediction, again?

It depends when you asked him.

I still don't understand the idea of objecting to dynamic predictions.  I think that ignoring incoming evidence in favor of maintaining a pre-existing decision is pretty much responsible for half the bad in politics these days.

Ok, I'll hold off until the final count is certified and when it's not the exact number that His Highness predicted, we'll hear what excuses His followers have for me.

The excuse is that you are misrepresenting his prediction. He put large error bars on the 27 votes. Go read that post. It's there. You are acting like the man on the street who sees a poll result +2 and thinks the pollster got it wrong when it ends up -1.

[Alcon]

What was Nate Silver's prediction, again?

It depends when you asked him.

I still don't understand the idea of objecting to dynamic predictions.  I think that ignoring incoming evidence in favor of maintaining a pre-existing decision is pretty much responsible for half the bad in politics these days.

Ok, I'll hold off until the final count is certified and when it's not the exact number that His Highness predicted, we'll hear what excuses His followers have for me.

Is anyone other than BRTD really taking everything he says at face value?    Lunar and I have whined in the past about a number of things, including his fundamentally ridiculous undecided breakdown allocation mechanism.  The fact that it didn't screw up his final prediction still doesn't mean it didn't suck.  Anyway, I can't think of anyone who is both paying attention to him enough to understand/read his methodology and worships him unconditionally.

I'm not sure if I'm one of his "followers," but I'd posit that:

1. His methodology was flawed, which is somewhat understandable considering there are a huge number of variables here; or,

2. He may have introduced his own insidious bias, as he tends to, although much more in non-quantified analysis; or,

3. There could have been a totally unpredictable, un-modelable event that we didn't think of either; or,

4. Some combination of the three.

Either way, I like his system.  Why?  Because he is transparent and dynamic.  I have two stock brokers.  One makes a prediction six months beforehand and tells me to "hang in there," and may also be an angry Italian.  The other has a model, perhaps flawed, but dynamically updated to reflect new conditions.  Who am I going to go to for my financial advice?  Who am I going to for my political advice?

I think any unconditional worship based on misunderstandings of his methodology is about six-of-one with unconditional distrust based on the same.

[justfollowingtheelections]

Franken up by 285 and projected to win by 74

[Sam Spade]

I have no problem with Nate's statistical model.  There a lot of room for error in my calculations and there's a lot of room for error in his. 

My big issue with his model "at this point" is that we're getting to the point where the only challenges left will be "withdrawn challenges", which means, according to my review, that pretty much all the challenges left will be going to the non-challenger candidate.  That means that any analysis that extrapolates based on "present challenges - nonwithdrawn" will be faulty.

Therefore, I'm choosing not to base anything on statistical extrapolation, but rather on two assumptions made looking at the "withdrawn" challenges.

1) The number of challenges left for each candidate are SOS minus Strib.
2) These challenges will all go towards the non-challenged candidate or equal in numbers towards Other/No One.

The rate of error will be the variables on these numbers - administrative on the first, normal statistical error on the second.

[Keystone Phil]

You are acting like the man on the street who sees a poll result +2 and thinks the pollster got it wrong when it ends up -1.

No, I'm the man on the street tired of the poll showing us eighteen different results and then hailing the poll as genius at the end of the day even if the number isn't even that close to the prediction.

[Sam Spade]

You are acting like the man on the street who sees a poll result +2 and thinks the pollster got it wrong when it ends up -1.

No, I'm the man on the street tired of the poll showing us eighteen different results and then hailing the poll as genius at the end of the day even if the number isn't even that close to the prediction.

You mean a Zogby poll, right?

[Torie]

I have no problem with Nate's statistical model.  There a lot of room for error in my calculations and there's a lot of room for error in his. 

My big issue with his model "at this point" is that we're getting to the point where the only challenges left will be "withdrawn challenges", which means, according to my review, that pretty much all the challenges left will be going to the non-challenger candidate.  That means that any analysis that extrapolates based on "present challenges - nonwithdrawn" will be faulty.

Therefore, I'm choosing not to base anything on statistical extrapolation, but rather on two assumptions made looking at the "withdrawn" challenges.

1) The number of challenges left for each candidate are SOS minus Strib.
2) These challenges will all go towards the non-challenged candidate or equal in numbers towards Other/No One.

The rate of error will be the variables on these numbers - administrative on the first, normal statistical error on the second.

Nate takes the withdrawn challenges into account, and awards Coleman a net of 385 when those ballots are included in the totals.

[Alcon]

You are acting like the man on the street who sees a poll result +2 and thinks the pollster got it wrong when it ends up -1.

No, I'm the man on the street tired of the poll showing us eighteen different results and then hailing the poll as genius at the end of the day even if the number isn't even that close to the prediction.

Why not pick the model you think is most accurate and rely on that...?

He's a modeler, not a predictor/bettor.  But the only thing he did a single model for (the Presidential election) sure turned out rather well (NE-2 excepted)

You are too results-focused.  Nailing three elections may be impressive, but with a little common sense, it is not difficult.  Producing a representative model of conditions is a lot more impressive to me, and that requires dynamic input, full stop

[Keystone Phil]

 Nailing three elections may be impressive, but with a little common sense, it is not difficult.

The personal attacks continue.

Sorry we can't all be as smart as you and The Great Nate.

[Alcon]

 Nailing three elections may be impressive, but with a little common sense, it is not difficult.

The personal attacks continue.

That wasn't a personal attack, and I can't imagine how it could be interpreted as one.

If someone nails three elections, that's great, but if you just echo conventional wisdom on three elections, chances are good that you'll nail them.  That is "you" in the "one" sense, not "you," because I wasn't taking about Keystone Phil there.

CW is usually based on solid ground, of course, but if someone can reproduce the roots of CW in a statistical model, I find it a lot more impressive than "sticking to one's gut" or whatever.  That includes accepting dynamic information.  It's what I do in my own predictions.  Unless it is for a "foresight contest," ignoring new data seems fundamentally pointless to me.  Political prognostication is a wet finger to the wind, not a foresight contest, at least to me.

On the other hand, if someone can predict something but not explain their prediction, I essentially consider it trash.  Not quite trash, maybe, but gut instinct.  Gut instinct has its value, because our minds do pick up on things, but it only goes so far.  Basically, Silver provides more transparency, elaboration, mechanization and currency than virtually anyone out there -- and they're all traits I admire in prognostication.

[Keystone Phil]

If the comments about "nailing" three elections wasn't about me, I don't know why it was mentioned.

Whatever the case, kindly blow it out your ass.

[Alcon]

If the comments about "nailing" three elections wasn't about me, I don't know why it was mentioned.

As I said, getting three elections in a row right on is nice and nothing to sneeze at, but not difficult if you follow conventional wisdom and pay some attention.  Trying to understand what that conventional wisdom coalesces around is more risky and interesting to me, and that's what Silver does.  Quantifying CW, and poking at it, is interesting to me, and that's what Silver does.  That involves accepting information on-the-fly, and I'm saying I don't understand why you see that as such an evil.

I don't know what three elections you're talking about.  I wasn't talking about any one person, or any specific three.  I definitely wasn't referring to any specific incident relating to you.

[Alcon]

If the comments about "nailing" three elections wasn't about me, I don't know why it was mentioned.

Whatever the case, kindly blow it out your ass.

It was, as any dullard could see (and, yes, you can take that as a personal attack; I'm sick of your victimization and have no problem with attacking you on it), a generalized statement. Anyone can predict three elections correctly, the results of many prognosticators and predictors, solid though they might be, are not important, but it takes a powerful statistical model such 538's to be applicable to multiple situations and continuously able to be updated, which Alcon believes to be more impressive (and I am inclined to agree with him).

Yes, thank you, that's exactly what I mean (and better than I put it)

He takes the exact same input we do -- but he processes it in a model, instead of using "common sense."  It's not that I'm against common sense; it tends to work, but it also is subject to insidious bias.  So is statistical modeling, but there is a degree of separation between the data and the interpretation.  It reduces some biases, for instance, the human tendency to see recent information as having a heavier "weight" than older information.  It's still subjective -- you have to decide the inputs, variables, and model adjustments always lead to potential injection of bias.  But it is less inherently the slave of our intuition.

Is it perfectly objective?  No, although by and large I haven't seen much partisan bias in Silver's numbers analysis.  Do I have problems with some of his models?  You betcha, as I said with his undecideds allocation in the General which was horrible.  But he's the standard-bearer of the field that I think is the best, most interesting part of political analysis today.  Shrug.

[I spent the winter writing songs about getting better]

He's reading the three elections as:

-PA-13 2004
-Santorum 2006
-McCain in Pennsylvania 2008

Basically the three notorious ones that he blew big time.

[Sam Spade]

Settle down, folks.  Or at least take this one outside..

[Franzl]

He's reading the three elections as:

-PA-13 2004
-Santorum 2006
-McCain in Pennsylvania 2008

Basically the three notorious ones that he blew big time.

the problem is, Phil didn't blow PA 2008, he never claimed McCain would win, even if he probably hoped and prayed for it.

[Alcon]

ah.  I registered at the tail end of the PA-13 2004 debate, and definitely never read that thread, so I didn't even know about his prediction there.  If I did, it didn't stick.  I'm just not smart enough for that kind of subtle dig anyway--

It was a generalized statement.  Even if it hadn't been, following the lines of my logic, blowing 3/3 isn't necessarily any more of an indicator of poor methodology than nailing 3/3.  Although (not to over-emphasize conservative analysis) I think it's much easier to nail a race than blow it, but that was not what the comment was about.

It was entirely a comment on how useful I found the process behind Silver-style analysis, vs. the sort of almost completely subjective analysis we do.  Both have their values but I think the former is more meaningful beyond having good intuition, sources, luck, etc.  Both have their values but I think the former is better and I can defend that belief.  That's all I was saying.

[I spent the winter writing songs about getting better]

He's reading the three elections as:

-PA-13 2004
-Santorum 2006
-McCain in Pennsylvania 2008

Basically the three notorious ones that he blew big time.

the problem is, Phil didn't blow PA 2008, he never claimed McCain would win, even if he probably hoped and prayed for it.

He did roughly up until September and kept arguing that the polls were all wrong because of the power of the Almighty Bradley Effect or something. He then admitted Obama would win Pennsylvania but still kept arguing that it would be a lot closer than the polls showed and that he would do worse than Kerry in Philadelphia because of his anecdotal evidence. All of it of course ended up being garbage.

[Franzl]

He's reading the three elections as:

-PA-13 2004
-Santorum 2006
-McCain in Pennsylvania 2008

Basically the three notorious ones that he blew big time.

the problem is, Phil didn't blow PA 2008, he never claimed McCain would win, even if he probably hoped and prayed for it.

He did roughly up until September and kept arguing that the polls were all wrong because of the power of the Almighty Bradley Effect or something. He then admitted Obama would win Pennsylvania but still kept arguing that it would be a lot closer than the polls showed and that he would do worse than Kerry in Philadelphia because of his anecdotal evidence. All of it of course ended up being garbage.

many people thought PA would be closer than the polls, myself included.

[I spent the winter writing songs about getting better]

He's reading the three elections as:

-PA-13 2004
-Santorum 2006
-McCain in Pennsylvania 2008

Basically the three notorious ones that he blew big time.

the problem is, Phil didn't blow PA 2008, he never claimed McCain would win, even if he probably hoped and prayed for it.

He did roughly up until September and kept arguing that the polls were all wrong because of the power of the Almighty Bradley Effect or something. He then admitted Obama would win Pennsylvania but still kept arguing that it would be a lot closer than the polls showed and that he would do worse than Kerry in Philadelphia because of his anecdotal evidence. All of it of course ended up being garbage.

many people thought PA would be closer than the polls, myself included.

So you're yet another believe in that Almighty All Powerful Bradley Effect?