Thursday, November 5, 2020

Oh, that's where that came from

Be honest. How many did you get?







Growing up in Arkansas, I remember KARK, the NBC affiliate in Little Rock, used an excerpt from "Karn Evil 9" as their opening theme.




Someone once defined an intellectual as anyone who could hear the the William Tell Overture without thinking of the Lone Ranger. I wonder how many people would make that connection today.




Posted by at No comments:

Wednesday, November 4, 2020

I need to take a break and this one's worth repeating

Working from home I've been burning through a lot of instrumental music. (You come to miss the background noise.) I've been digging into Classical, Jazz and scores (mainly Goldsmith, Morricone, Thomas Newman), even some new stuff.

Around a hundred days, though, you start getting restless so I reached out to my friend Brian Phillips, a sometimes DJ whose knowledge of popular music is beyond encyclopedic, and asked for some recommendations that were eclectic and off the beaten path.

He delivered. Enjoy.

The Sons Of Moses - Soul Symphony






The T-Bones





Shocking Blue - Acka raga




The Black Exotics


The Tornados - Telstar




The Jumping Jewels Africa





Mr Bloe - Groovin` With Mr.Bloe





Santo & Johnny - Sleep walk


Posted by at No comments:

Tuesday, November 3, 2020

The greatest comedy ever put on film


I thought you could use a break.



Posted by at No comments:

Monday, November 2, 2020

Schrödinger's election

This isn’t really the best of analogies, but bear with me I am working under the clock here.

While everyone is aware of the basic facts, I get the feeling that most analysts have not yet really internalized the new cadence of the election and its implications. They are still tending to  think of this as something that will basically happen next Tuesday instead of something that is 2/3 over. At this point, like Schrodinger‘s hypothetical cat, the election is largely locked in a box, its fate decided but unknown (though unlike with the cat, we do have some idea what’s going on in the box – like I said, this is not a great analogy).

(This, of course, does assume that we actually count all of the ballots that have been cast in person or received and certified by mail. Those not yet received may be in a kind of limbo.)

The first and most obvious aspect is that anything that happens, any political event, any news story will have no impact on the votes that have already been cast and certified. They are a fixed but unknown quantity.

The second point is a bit more subtle.

We are talking about a mixture of things that have happened and things that may happen. Since it appears that the early voting was heavily Democratic and it is assumed that the in-person voting will be mainly or at least disproportionately Republican, this basically puts a floor under Democratic support and a ceiling on Republican.

When you put these two ideas together it leads to some interesting conclusions.

Events and breaking news that can radically change results on election day are almost always on the downside. It is very difficult to imagine something breaking at the last minute that would double the turn out in a certain district. By comparison, it is not difficult at all to imagine events that would greatly suppress it. Given the extremely chaotic times we are living in, it would be foolish to rule out the possibility of some otherwise unimaginable black swan event, but it seems unlikely we see one that will cause a huge surge in voter turn out on November 3.
Posted by at 2 comments:

Rethinking Likely Voter Models

At this point just throwing out a lot of post as quickly as I can and relying too much on my phone’s not so reliable dictation app. Apologies in advance.


1. Most journalists are horribly confused by the differences between gathering and aggregating polling data and building likely voter models. Two completely different processes using entirely different tools prone to entirely different types of failure.

2. Of the two parts of the process, the modeling component will tend to be less stable and more vulnerable to the kind of changing conditions that can undercut basic assumptions.

3. It is impossible to believe that there is not a strong relationship between likely voter models and voter suppression. This is something we need to think more about and discuss more openly.

4. One of the implications of 3. is that changes in the effectiveness of voter suppression techniques can completely upend likely voter models. Remember the J shaped curve. There is considerable evidence showing that in many states the effect of voter suppression has reversed directions in a big way.

5. Likely voter models were also based on historical data from a period where early voting was much less accessible and played a much smaller role.

6. The models were also based on (at the risk of being obvious) non-pandemic data. This also has the potential to greatly undermine their predictive power this year.

7. Finally, returning to the subject of early voting. I wonder if the modelers working on these problems have been overly static in their thinking. If a month in advance of the election, you see a large number of people whom you have pegged as unlikely to vote having already voted, shouldn’t there be a way of using this information to adjust your models?

Posted by at No comments:

Russian Roulette Odds

I hate assigning subjective probability’s, particularly in chaotic times, but if I had to I would say that when you add in all the possibilities of variation in polls and plans to steal the election, we are looking at something on the low end of Russian roulette odds. I’ve found this is a useful framing when talking with nervous friends about the election.  

On one hand it captures that a Trump victory, even with extensive cheating, appears unlikely. On the other hand, it also reinforces the fact that we are generally more concerned with expected value than we are with probabilities.

With a six shot revolver, your odds of a bad outcome are in the mid teens. Unfortunately, a bad outcome in that game is really, really bad.

Posted by at No comments:

Today’s post is brought to you by the letter J



I’ve made this observation before but I thought it was worth calling out in a freestanding post.

We have lots of reason to believe that the relationship between voter suppression efforts or perhaps more precisely awareness of those efforts and actual turnout of the targeted groups is J-shaped. Making voting more inconvenient for certain people up to a point will have the effect of making them vote less, but eventually they get so angry at being disenfranchised that they actually become more likely to vote.

We have data, anecdote, and common sense telling us that we have reached the point where these efforts appear to have become massively counterproductive for the Republican party. The problem is the GOP is addicted to these tactics. If voter suppression has lost its effectiveness for sometime to come and the party doesn’t come up with something to replace it, they are absolutely screwed.



Posted by at No comments:

Friday, October 30, 2020

A few quick lunch time notes on making sense of the early and “late” votes -- UPDATED

Here’s how I’m framing this. Very much looking for feedback

Think of the following groups:

1. Vote on election day

2. Vote early in person

3. Vote by mail (drop box)

4. Vote by mail
    a. Before Oct 25
    b. After Oct 25

2., 3., and 4a. appear to favor Democrats.

 

We assume 1. will favor Republicans.

I’m under the impression that the GOP’s attempts to throw out votes is focused on group 4b.
[Is this correct?]

Because Trump et al telegraphed their strategy months in advance, the Democrats have had time to push back hard with the message vote as early as possible and under no circumstances rely on the mail in the final week.

In order for the GOP tactic to work:

1. The election in important swing states must be fairly close.

2. A large number of voters need to have mailed their ballots in that state less than a week before the election.

3. These must be heavily Democratic.

I am dubious of 2 and 3. At this point only the lowest of low information Democrats would think that mailing a ballot at the last minute is a good idea, but I don’t have any data to back that up.

Can anyone shed some analytic light on this?

 

______________________________________

UPDATE:

 A bit of context. There were 137 million votes cast in 2016. 

 

With what will very likely be more than half of this election's votes already cast, the order in which they are counted will probably determine the story Trump's advocates will be able to tell.


I also wonder if  this massive and successful push for early voting, especially among Democrats, and the related telegraphing of the Trump/USPS/SCOTUS plan to steal the election means that there won't be that many votes left to worry about arriving after Nov. 3.


Finally, not as a prediction but just as an observation. Early votes are in the bank; election day votes are accounts receivable. One of the problems with encouraging your followers to wait till the last day is that you don't have a floor if that day turns ugly.

Posted by at No comments:

Are Nate and Elliot and Andrew klever enough to kope with a K

Many years ago in Paris, Arkansas, there was a small shop owned by a local family named Kafka. I never met them and have no idea what relation they had to the writer.

The shop was of a very common type in the Ozarks, folksy with more often than not a made-in-Taiwan hillbilly décor. I don’t recall ever going inside but I do remember the sign which read something like this:

Krafts
Antiques
Fun Stuff
Kwilts
Art Supplies

This memory is not all that relevant but then neither is this post. It would have been had I written it when I first intended to a month or two ago. Back then, economic indicators and forecasts probably probably had a bigger roles in the models of 538 and the Economist. I’d imagine now it’s all about weighing polls and estimating turnout. But even if I missed the timeliness window for this post, there are some less ephemeral issues I still want to hit.

On some level, all predictive modeling relies on the assumption that the important relationships and trends we’ve observed in data in the past will continue to hold in the future. We don’t talk about it all that much but this is one of those things that makes all competent statisticians at least a little worried. This is especially true when we go out of the range of our data, when the variables we put into our model start having values we’ve never seen before.

Pretty much serious election models factor in the economy and where it’s going. The actual relationship may be complicated but, at the risk of oversimplifying, an economy that’s good or trending up favors the party in power and vice versa.

But what happens when the economy is good for half the people and terrible for the rest? Many economists have described our current situation as a K-shaped recovery with white-collar knowledge workers doing fairly well while those in other sectors such as the service industry suffering horribly.

As far as I know, we haven’t had a presidential election during a K-shaped recovery, at least not since we starting scientific polling. This is outside the range of data (as is the pandemic, as is having a president openly undermining the election, as is…).  This is where the art of modeling kicks in. The statisticians at 538 are smart and experienced and I have faith in their judgement.

But when you read credulous story about model confidently predicting some wildly counter-intuitive development, it is also good to remember that modeling is a mixture of science and art and some people aren’t very good at the latter.

Posted by at 2 comments:

Thursday, October 29, 2020

Rethinking modeling assumptions in 2020

 

 

The idea that you should incorporate a correlation matrix into an electoral model is one of those things that just makes sense across the board. It is supported by the data, it is intuitively obvious, and it is easy to justify from first principles.

But here’s the part that bothers me just a little bit. Historically, these models predicted the outcome of a collection of events that happened in different states but mostly at a single point in time simultaneously, election day.

That’s not how elections work in 2020. Different states now have wildly different cadences and rules. Most of the votes might be cast in one state before another even starts the process. Is the likelihood of a candidate outperforming the polls in the first two weeks of October in one state still strongly correlated to the probability in another state on election day? Do certain aspects of the model fare better than others under these new conditions? Would 538’s model handle these changes differently than the Economist model would?

I have absolutely no idea whether or not these are important issues. I am woefully ignorant on the subject, but it seems like an interesting topic for discussion so if anyone better informed than I (which is to say pretty much anyone reading this blog) would care to join in, I would love to hear some opinions.

Posted by at 4 comments:

Wednesday, October 28, 2020

Wednesday Tweets -- these anti-mask activists really get in your face



Lots of topics here I've been meaning to blog on.

The implications if we start thinking of the effect of voter suppression as a J-shaped curve.



The culpability of the Hoover Institute for the current crisis.

Long time readers will know where these are going.

I've been thinking a lot about the "surge ahead on election day and cut off the count strategy." Among other things, I've been thinking how stupid it was to telegraph that plan months in advance. If Trump plays Goldfinger in the remake, he'll tell Bond about his Fort Knox plan during the gin game.



The approach of quietly chipping away a reproductive rights on a state level has given moderate Republicans a lot of breathing room. That may be going away.


[Search for the "great unwinding."] In 2020 the GOP will still have to appease the cult members. Though unlikely, I would not rule out Don Jr. or Ivanka or even the return of Trump himself. Whoever gets it will find that holding onto the base while building a broad coalition will be... challenging.


We all need snow days (even if some of us have to drive to them).

Posted by at No comments:

Tuesday, October 27, 2020

The good news about Tesla's Full Self Driving technology is that it's not advanced enough to lull you into a false sense of security

Interesting companion piece to yesterday's post on Tesla's disturbing beta rollout of FSD. It promises that the driver has to do almost nothing. The claim turns out to be false but perhaps that's a feature, not a bug. It turns out that doing almost nothing is really hard.

John Pavlus writing for Scientific American.

People often use the phrase “in the loop” to describe how connected someone is (or is not) to a decision-making process. Fewer people know that this “control loop” has a specific name: Observe, Orient, Decide, Act (OODA). The framework was originally devised by a U.S. Air Force colonel, and being “in” and “out” of the OODA loop have straightforward meanings. But as automation becomes more prevalent in everyday life, an understanding of how humans behave in an in-between state—known as “on the loop”—will become more important.

Missy Cummings, a former Navy fighter pilot and director of Duke University’s Humans and Autonomy Laboratory, defines “on the loop” as human supervisory control: "intermittent human operator interaction with a remote, automated system in order to manage a controlled process or task environment.” Air traffic controllers, for example, are on the loop of the commercial planes flying in their airspace. And thanks to increasingly sophisticated cockpit automation, most of the pilots are, too.

Tesla compares Autopilot with this kind of on-the-loop aviation, saying it “functions like the systems that airplane pilots use when conditions are clear.” But there’s a problem with that comparison, Casner says: “An airplane is eight miles high in the sky.” If anything goes wrong, a pilot usually has multiple minutes—not to mention emergency checklists, precharted hazards and the help of the crew—in which to transition back in the loop of control...

Automobile drivers, for obvious reasons, often have much less time to react. “When something pops up in front of your car, you have one second,” Casner says. “You think of a Top Gun pilot needing to have lightning-fast reflexes? Well, an ordinary driver needs to be even faster.”

 ...

But NASA has been down this road before, too. In studies of highly automated cockpits, NASA researchers documented a peculiar psychological pattern: The more foolproof the automation’s performance becomes, the harder it is for an on-the-loop supervisor to monitor it. “What we heard from pilots is that they had trouble following along [with the automation],” Casner says. “If you’re sitting there watching the system and it’s doing great, it’s very tiring.” In fact, it’s extremely difficult for humans to accurately monitor a repetitive process for long periods of time. This so-called “vigilance decrement” was first identified and measured in 1948 by psychologist Robert Mackworth, who asked British radar operators to spend two hours watching for errors in the sweep of a rigged analog clock. Mackworth found that the radar operators’ accuracy plummeted after 30 minutes; more recent versions of the experiment have documented similar vigilance decrements after just 15 minutes.


Posted by at No comments:

Monday, October 26, 2020

The autonomous driving equivalent to Elon's bolting wheels on the side of the car

One of the foundational beliefs of 21st century tech narratives is that the biggest obstacles to innovation are legal. If not for lawyers and regulators, we'd living in a techno-paradise. 

With the possible and very qualified exception of pharmaceuticals, this has always been unfounded and in many cases, is the opposite of true. Companies like Tesla have long treated legal checks as figurative and often literal jokes, manipulating markets, misrepresenting products and in some cases, threatening public safety. 









Posted by at No comments:

Friday, October 23, 2020

The standard narrative on the Uri Geller/Amazing Randi conflict comes from the New York Times, which apparently got it from Uri Geller

RIP Randall James Hamilton Zwinge

James Randi, a magician who later challenged spoon benders, mind readers and faith healers with such voracity that he became regarded as the country’s foremost skeptic, has died, his foundation announced. He was 92.

The James Randi Educational Foundation confirmed his death, saying that its founder succumbed to “age-related causes” on Tuesday.

 ...

On a 1972 episode of “The Tonight Show,” he helped Johnny Carson set up Uri Geller, the Israeli performer who claimed to bend spoons with his mind. Randi ensured the spoons and other props were kept from Geller’s hands until showtime to prevent any tampering.

The result was an agonizing 22 minutes in which Geller was unable to perform his tricks.

 

 

 

 

For Randi, those 22 minutes of magic tricks not being done would ironically become the high point of the magician's biography but there was one more twist in the story

Adam Higginbotham writing for the New York Times Magazine in 2014.

“I sat there for 22 minutes, humiliated,” Geller told me, when I spoke to him in September. “I went back to my hotel, devastated. I was about to pack up the next day and go back to Tel Aviv. I thought, That’s it — I’m destroyed.” But to Geller’s astonishment, he was immediately booked on “The Merv Griffin Show.” He was on his way to becoming a paranormal superstar. “That Johnny Carson show made Uri Geller,” Geller said. To an enthusiastically trusting public, his failure only made his gifts seem more real: If he were performing magic tricks, they would surely work every time.

 It's a great tale except that there's little reason to believe it actually happened that way. Start with the fact that Geller seems to be the main source, which should have raised some red flags for Higginbotham.

 How about the appearance on the Merv Griffith Show? Wasn't he invited shortly after the Carson debacle? Not exactly. He was invited back. 


From IMDB:

The Merv Griffin Show (1962–1986)
Alfred Drake, Pamela Mason, Uri Geller, Captain Edgar Mitchell
TV-PG | 1h | Comedy, Family, Music | Episode aired 19 July 1973

The Tonight Show Starring Johnny Carson (1962–1992)
Ricardo Montalban/Eskimo-Indian Olympians/Uri Geller
1h 45min | Comedy, Talk-Show | Episode aired 1 August 1973

The Merv Griffin Show (1962–1986)
Eartha Kitt, Richard Dawson, Michelle Phillips, Uri Geller
TV-PG | 1h | Comedy, Family, Music | Episode aired 15 August 1973

 

 Geller's telling makes it sound like it was the Carson appearance that got him on Griffin, but he was a returning guest and there's no reason to believe he wasn't invited back simply because he had done well a couple of weeks earlier.

[Late Edit: He'd also made appearances on Jack Parr and Mike Douglas before doing Carson -- MP]

Nor is there evidence that Geller's career took off in late 1973.



If anything, it looks like Randi's debunking of Geller starting with the Tonight Show and culminating with 1975's The Magic of Uri Geller was what brought the charlatan down.

Journalists love people-are-stupid narratives, but, while I believe cognitive dissonance is real, I think the lesson here is not "To an enthusiastically trusting public, his failure only made his gifts seem more real" and is instead that we should all be more skeptical of simplistic and overused pop psychology.

 

Posted by at No comments:

Thursday, October 22, 2020

Farewell, Quibi. We hardly had time to mock you

When rich, successful  people raise huge amounts of money for undeniably stupid business ideas, there is a natural tendency to assume that it can't be as bad as it sounds. Fight that impulse and remind yourself that the PR people who fed the journalist the story you just read went to great pains to put the best possible spin on the enterprise. 

Quibi really was worse than it sounded. 


From the LA Times:

Katzenberg and Whitman told investors Wednesday afternoon that Quibi, which raised $1.75 billion to tackle the growing digital video market, will wind down after failing to attract viewers willing to pay to watch its shows, according to people familiar with the matter who were not authorized to comment.

Hollywood-based Quibi, which employed 265 people as of April, plans to use its remaining cash of about $350 million to pay back investors, sources said, a stunning turn of events for a company that promised to transform the entertainment industry.

Quibi said in a statement Wednesday it will close down the business and begin the process of selling off assets over the coming months.

 


Posted by at No comments:
Subscribe to: Posts (Atom)