This is just really awful. Read his book because you do not have the faintest grasp on what his schtick is.

Fwiw, 538 outsourced the 2015 UK general election model.

This thread was about France. Presumably any outsourcing that went on in the UK will happen there also.

I think you are just making excuses. He was on television explaining his predictions-if he didn't work on them personally that's not the responsibility of someone measuring his performance.

I wasn't going to mention this but he didn't exactly astonish any one with his 2010 performance in the UK election either. He predicted the Liberal Democrats would gain twice as many seats as they actually did and badly undestimated the Labour vote.

I get that a lot of people like Silver because he brings a veneer of intelligence to American political analysis, which badly needs it, but his actual predictions are only marginally better in aggregate than taking a simple weighted average of all the available polls.

If you want a *win* out of this thread then accept this: he may well be right about the overestimation of Le Pen's chances being down to Trump/Brexit, which is how this discussion started. I just don't want people to bet the farm on the basis of his recommendation.

From recent polls it looks like the Melenchonmentum has stalled and Macron is winning late deciders. A systematic error is still possible but not a single poll had Macron losing and the dynamic is good. I'm slightly unshook.

Lol at some of the last posts about Nate Silver. I dont have a strong opinion of the guy but it looks like his haters are pure Dunning-Kruger.

I'm getting tired of you fan boys. Have you got some kind of nerd crush on him or something?

Silver got the UK election badly wrong in 2010.
Silver got the UK election badly wrong in 2015.

(I'm giving him the benefit of the doubt on the basis of statistical error about the 2016 us election but he did only give Trump a one in three chance).

Why exactly should any one trust a pollster that continually gets thing wrong? It is a simple point.

He's not a pollster.

Not a fanboy but I recognize he's more competent than me in that area and I'm interested in his takes. If you're way above his level that's great, we're lucky to have you here and I'm looking forward to more of your analysis.

It isn't a "way above" thing. Maths and stats geniuses, which Silver certainly is, are very good at certain types of mathematically solveable problems. Much better than me.

Unfortunately politics isn't mathematically solveable. It was pretty close to it in past elections but the statistical models have fallen apart after the financial crisis.

Seems really obvious at this point Macron is gonna win. He's leading the first round and he's leading h2h against everyone by a good margin.

What about the fact that the typical margin of error in French polls could put him easily in fourth?

Yeah I guess, I mean he'd have to lose 3 points and Melenchon or Fillon would have to gain 3 points compared to the polls. I still think they're drawing pretty thin, but I do agree it's a possibility.

the markets (les marchés?) seem hesitant to back him. on betfair he's only implied to be about 72% to win the second and roughly the same odds to actually get there. it all seems very conservative so maybe nobody really trusts the polls.

So there was actually good discussion of this on the 538 podcast on Tuesday. Iirc, French pollsters remove undecided voters which is not the best practice as it doesn't display the level of uncertainty.

Also, it seems like bad reasoning to suggest that he needs to lose 3 AND someone else needs to gain 3. As it's a zero-sum game, if he loses 3, that needs to go somewhere. Obviously it may not all go in the same place, but they aren't two independent events.

Basically, don't get bullish on a 3 point swing given that it's well within the margin of error.

A 3 point swing doesn't appear to be 'well within the margin of error', at least not how you seem to be using that. The 95% confidence rating on most of the polls is around 2%. On that 538 podcast they seemed to be saying it was, historically for French presidential elections, 3% (at least I presume that's what they meant).

If you plug those numbers in naive simulations then someone 4-5% back wins a tiny % of the time. For 4% back and a 3 point 95% interval a quick excel simulation suggested Macron would lose to one of Mélenchon or Fillon about 6% of the time*. In the real world then the effects you mention render such simulations suspect, but how far can you really push those effects? You would need to increase the 6% by an order of magnitude to start making Macron less of a favourite than even his biggest supporter believes he actually is.

Seems like the French polls just need to be total crap to not think he's a decent favourite. Maybe they are, but it won't be because of historical MoE I wouldn't think.

* I'm no stats guru, it's possible I messed up here

People were saying this at the US election too. "She's 10 points ahead" "She has her firewall that can't possibly be broken" etc. The analysis was all total rubbish. Trump virtually clean sweeped the swing states.

You're right, no one should listen to him if he gets things continually wrong.

But unfortunately in this world, it doesn't matter if you're an incompetent fool. All that matters is that you can convince people you know what you're doing. Even if you have zero clue. It seems like Nate Silver has done this perfectly.

This is not trivial. One of Silver's contributions is popularizing the idea of aggregating polls rather than trying to decide which poll is the most reliable.

This point would be more convincing if he wasn't one of the only forecasters who repeatedly gave Trump a significant chance when people were saying that.

Silver definitely knows his stuff. He does get things wrong though, I mean all throughout the primaries he got Trump wrong, and even when Trump secured the nom, he said that he had no chance of winning the general (this was in May). But it is true that closer to the election his analysis was insightful and a lot of his theories came true, like saying there's a good chance Trump might lose the popular vote and win the electoral, whereas the opposite was almost impossible to happen. I remember a lot of the pundits saying the opposite, that Hillary might lose popular but win electoral. Pundits I would say were mostly FOS.

Asselineau going with the "I have an american friend" when asked why he doesn't like the US. Solid.

Another shooting in Paris. Everybody assumes it's ISIS and Marine Le Pen probably gains from it by saying that brown people should go back to brown countries.

Gonna be a real **** sandwich if she wins.

Previous attacks have not benefited Le Pen in the polls. Which isn't to say she won't win.

How many people in France are genuinely undecided as of right now?

I'm not sure if it should be a lot because there are so many candidates to compare before choosing or if it's almost no one because there's a candidate for practically every viable political ideology.

It's very difficult to answer that question. Abstention rates seem to be about 25-30%, which is maybe up to 10% higher than normal - maybe some of those people can be convinced. Maybe up to 30 % of those who pick someone in polls say they could change their mind, with people being more sure of their votes for Le Pen and Fillon than others. Given what's happened in the past after attacks I don't think this makes it more likely late breakers will go to Le Pen than would otherwise have been the case, but I could be wrong.

Whatever it means it's certainly not bad news for Le Pen, which makes the election an even bigger concern.


She can hardly say triumphantly "Look, I told you this is what our immigrants always do so let's deport them all", so has said "I feel for and stand by our security forces, who have been targeted again." (my emphasis).


Both Ms Le Pen and Mr Fillon later announced they were cancelling campaign events scheduled for Friday, the last day of canvassing for the vote.

3 of the top 4 candidates all pro putin. The fourth wants to work with him. Guess the EU is doomed, that's a different argument for good/bad overall (certainly with short term pain + Putin certainly wants that)

http://www.ouest-france.fr/elections...region-4940297



Bretagne is solid, as usual. LOL PACA.