> In any race where Silver didn't make a pick — again, he isn't playing the pick-every-contest game — Diggler's numbers treat the non-prediction as a wrong prediction. So Silver's accuracy rate through May 7, according to Diggler, was just 55 percent, while his own was 89 percent.

https://www.washingtonpost.com/news/the-fix/wp/2016/05/12/so...

So the comparison counts non-predictions by Silver makes your statement correct.

> even without a fancy statistical model

Remarkably anti-intellectual statement.

I think that you would agree that null is not a correct prediction, so clearly the Silver model has failed for those votes.

Null is neither correct nor incorrect. If I don't play a game of chess today and you do, I haven't lost anything. I simply haven't played. The world is not binary.

This claim is tantamount to a schoolyard taunt like, "If you don't fight me, you're a coward." And claiming victory when your opponent wasn't even at school that day to hear the taunt.

Right, but if you play chess against a lot of children and crush them 100% of the time and I play against moderately experienced players and win 90% of the time, I'm probably better at chess

Fine. Fuck the analogies because yours isn't even close to the actual situation.

At the time of the writing on the article I posted as a rebuttal: Each time both Silver and Diggler predicted results they agreed except for 8 states. When they didn't agree, they had an equal number of correct results (and incorrect results), 4-4. They're effectively tied.

The ones that aren't predicted indicate nothing about Silver's methods other than a lack of sufficient data to put through his models. Which says something to his ethos. He doesn't just make wild stabs in the dark like a typical pundit when he doesn't have the data to back up his prediction.

EDIT: Cleaned up the second paragraph.

EDIT FURTHER: I suppose it's my fault. Chess was the wrong analogy if someone were to pick an analogy. It's played head-to-head, and comparison to people that aren't playing against each other requires more complex analysis. The situation here is two parties participating in an open contest, where everyone can win, everyone can lose, or something in between can occur.

The point still stands. Someone not participating in a given contest doesn't equate to losing. They just aren't competing.

There are lots of pundits. It's quite likely that one of them would get lots of correct guesses by chance.

Step 1: set up a whole bunch of pundit blogs.

Step 2: for each election, have 50% predict one winner, have 50% predict the other winner.

Step 3: after the election, close all of the blogs that predicted incorrectly.

Step 4: repeat with a dwindling pool of more and more "accurate" pundit blogs.