About people who have supposedly great track records of predicting elections

[Silent3]

Aside from the very low odds of an independent candidate winning in a US presidential election, even someone who knows nothing at all about politics, or even who's running for election, can take a random Democrat or Republican stab at guessing a winner, and guess correctly 50% of the time.

If someone does that for 10 elections in a row, their odds drop to 1 in 1024 of keeping a perfect record — a little less than 0.01% odds. Not particularly impressive.

But think of how many people are out there playing the presidential election prognostication game! If you have 10,000 would-be pundits doing no more than flipping a coin, you'd still end up with around 10 people with perfect records, people who are nothing more than lucky. Yet they'd be still get press coverage for their "success". They'd be the ones people looked to for confidence (or worry) about who is going to win next.

Yet they'd know nothing at all worth the investment of your hopes or fears.

And of course, it doesn't take a huge amount of political savvy to be better at picking winners than someone who's just flipping a coin.

For pundits who merely know enough about politics to be raise their odds to two thirds of picking a winner each time, a crowd of such pundits would generate on average about 173 apparent political geniuses with perfect records for the last ten elections.

So here's the important point:

There will ALWAYS be a whole bunch of people out their in the world with amazing track records for picking presidential winners. ALWAYS. There will also be a whole lot more failures, but you just won't hear much about them.

It might seem counter-intuitive, but because of the winners bias for news coverage about successful prognosticators, their tracks records mean close to nothing.

[edhopper]

1. Plus there were some elections easier to call

And if they predicted GW Bush and Trump, did they predicted they would lose the popular vote but squeeze out an electoral win by the slimmest of margins through a series of very unique circumstances?
Because if they predicted outright wins by those two, they were wrong.

[KPN]

2. Yep. Count on nothing.

Don’t get complacent. Winning is and will be the result of hard work by many and a dogged determination to not only vote but make sure our individual votes are counted.

And it’s not just a matter of winning. They are going to do everything they can to steal the election. We must win by a landslide to ensure victory.

[Silent3]

5. I'm not quite saying "count on nothing".

Nor of course, do I ever advocate complacency.

If it weren't for fears of a deliberately sabotaged postal system and whatever other shit Republicans and their Russian cohorts might play, I think I could feel very confident of a Biden win right now, and rightfully so.

But my main point is specifically about the way people react the track records of other people who have done well with picking presidential winners. Nothing more or less.

I'll leave the "Don't get complacent!" worrying to others (because not every sign of hope is actually a risk of complacency).

[Cicada]

3. Warren Buffett has had average returns for the past several years

I stock picker out of 100,000 will become a billionaire by chance alone?

A thousand drugs are tested for efficacy against covid. Maybe a drug looks good in a solid double blind trial. But then is disappointing.

I make ten baskets in a row. Do I really have a hot hand?

Yes. You are right on the money my random walking colleague.

[mr_lebowski]

4. I'm not sure that's entirely true ...

This theory applies probability concepts that would be totally valid if elections were entirely random events, but they're not. It would also be true if people were entirely guessing as to the outcome. But are they?

If someone has built a system that takes in a bunch of statistics and produces an expected outcome, and that system consistently predicts the winner going back for many elections ... one should logically give that system consideration as being, at minimum ... useful for predicting future elections.

If you don't believe this sort of thing is possible, then you also logically don't believe we can know what's going to happen in the future as we pump CO2 into the atmosphere. Same principle ... known as modeling.

[Silent3]

7. I'm not dismissing the value of modeling at all.

If I were doing that, then I wouldn't have posited the idea of pundits with a 2/3 chance vs. a 50% chance.

And a good model might well be better than 2/3. Perhaps much better.

But because there are so many models to compare, just like their are so many pundits to compare, you'd need hundreds of elections to pin down the best models. Without that data, there's a lot of variance in the way modelers pick their inputs, and weigh their inputs against each other. There's nothing purely empirical or logical about those choices, so until centuries have passed to provide empirical comparisons, it's reasonable to view the outputs of these models as having random variation.

Plus, given the chaotic nature of elections, there is almost certainly an upper limits on possible accuracy, just like long-term weather forecasts are limited by chaos in their ultimate achievable accuracy.

[wryter2000]

6. Thanks

Great analysis. Lichtman has been bugging me for years.

[zipplewrath]

8. 538

It's what makes Nate Silver so interesting. He doesn't pick winners, he assesses likelihoods. When the race is over, one can assess how well his assessments were, even for "obvious" races.

[Buckeye_Democrat]

9. I'm more surprised by people with bad track records...

... who still get paid well for their supposed expertise.

People like NFL "scouts" who have a long history of just being dead-wrong, but always act confident about themselves.

Fox News idiots are another example. No matter how much they're wrong -- e.g., reporting that WMD's were found in Iraq -- they just ignore it and move onto their next load of BS.

[Brainfodder]

10. Are you trying to tell us a failed pandemic responding President can still win legitimately?

I didn't think so!

[Silent3]

12. My OP wasn't about any specific pro-Biden or pro-Trump prediction.

Just the prediction game in general.

Personally, I don't think Trump can win legitimately. I don't think that requires amazing skills to see, either.

What I certainly wish I knew better was the odds of illegitimate tactics being successfully. That's going to keep me very worried.

[Turbineguy]

11. It's a little like my method for winning at roulette.

It works until it doesn't.