Saturday, September 24, 2016

College Football Week 4

For most of the week my model has been producing gobs of #N/A errors and until yesterday I couldn't figure out why. Then yesterday I realized (as is usually the case) that it's one of the dumbest possible things and a bunch of scores are just attached to the wrong games for weeks 2 and 3. Luckily the only impact is the "Last Week's Ranking," and the Team Dashboards, so I've taken down the former and have updated the sidebar to reflect this week's ranking only.












  • Watchability combines how good the two teams are and how likely the game is to be close
  • A team is shaded green according to their chance to win (darker = better chance)
  • This post has more detail on the math behind Watchability

Wednesday, September 14, 2016

College Football Week 3

The week before the season started, I tweeted out one of these for every FBS team. Now, you can make your own for any FBS team at any time. Just click here or follow the link on the top right side of the page.

I've updated the right sidebar to reflect data through week 2, and week 3 games. I'm still toying with how I want that side to look and am not fully happy about it. So there could be more changes coming For now I changed the Top10/Bottom10 chart to just show the Top 20. It was a fun gimmick to show the bottom 10 preseason, but I find looking at #s 11-20 way more interesting.

Week 3 looks spectacular. I think I'm most curious to see if Louisville and their ridiculous quarterback are the real deal. 







  • Watchability combines how good the two teams are and how likely the game is to be close
  • A team is shaded green according to their chance to win (darker = better chance)
  • This post has more detail on the math behind Watchability

Tuesday, September 13, 2016

9/13 Election Update and Head to Head vs. 4 way polls

New polls:

  • Good for Trump
    • Trump +3 in the LA Times tracking poll
    • Clinton (only) +3 in ME
    • Clinton (only) +4 in the NBC News/SM national poll. This is a big national poll that typically causes jumps in the model in one direction or another, and she's been holding at +6-7 in this poll for a while. This one probably moved the needle. 
  • Good for Clinton
    • Clinton +8 in VA
  • Neutral
    • Trump +12 in KS
    • Trump +6 in TX
    • Clinton +25 in CA
      • None of these states will be decisive in the electoral college

One key way my model differs from 538 is in the choice of poll used. Not a choice based on which pollsters or which states (we both include them all) but on how many candidates the poll asked about. Specifically, when given the choice I opt for head to head polls, and 538 opts for 4 way (Clinton/Trump/Johnson/Stein) or 3 way (Clinton/Trump/Johnson) polls when available. 538 then subtracts totals from third (and fourth) party vote share to reflect how they think votes will move from those 3rd/4th party candidates by election day. I find this needlessly complex.

Voters who say they support 3rd party candidates early in the cycle tend to coalesce around either the D or R candidate as election day approaches (although for Gary Johnson that hasn't happened yet). This is a natural and expected process. Were I so inclined, I might indulge in the luxury of supporting whomever I Truly Like, but then as election day approaches I decide I can't abide Trump, and I vote for Clinton.

Of course, there will still be some people who vote for Johnson/Stein/Misc on Election Day. I've made the modeling decision (actuarial judgement) that those people would be reasonably approximated as the people who respond 'neither' when presented with a head to head poll. For one, people who answer "none of those" when only given two options are going to be the diehard 3rd party voters who likely will persist until election day. And for another, this is a functionally a head to head election and should be thought of as such. I'm sure Johnson and Stein voters are bristling, but the simple reality is that the winner of this election will be Clinton or Trump and it's silly to pretend otherwise.

A good model should reflect a logical construction of what's being modeled. This is a head to head race, we should use head to head polling.

Monday, September 12, 2016

9/12 Election Update and Health Care Economics

Not too many new polls today. Clinton ticked down slightly on that FL poll.

If her current bout with walking pneumonia impacts her polling, that'll take a while to come through in new polls, and it'll take a while longer to reflect in the model. That's not because the model is "too slow" to move but more because most new cycles come and go and turn out to be a much smaller deal than we think when we're in the moment. If they stick though, we'll learn that slowly as the new cycle around pneumonia drags on and her polling continues to reflect that.
  • Really Good for Trump
    • Trump +4 in FL
  • Neutral
    • LA Times/USC tracking tied
    • Clinton +5 in MI
    • Trump +15 in UT 

If you've ever thought "getting rid of employer-based health insurance is the silver bullet" then read on.

De-coupling employment from insurance might be a good idea, and it might be a bad idea, but it's not a silver bullet. The reason it's not a silver bullet is that employer-provided health insurance solves a critical problem in health insurance: antiselection.

In order for insurance (of any kind) to work you need to have some people for whom premiums exceed claims, have some people for whom claims exceed premiums, then hope you wind up roughly even. The problem is if people are fairly confident insurance is going to be a losing proposition for them, they won't buy it, this is the problem of antiselection. For some kinds of insurance (e.g. earthquake insurance) that's not a problem, because no one has a great idea what their expected earthquake related claims will be. For health insurance though, it is a problem. 

I bet you have a pretty good idea what your health care costs will be this year (barring some catastrophic event) right? So do most people. If you sell insurance to everyone individually, the people for whom insurance is a good bet will likely buy it, and the people for whom it's a bad bet will likely not. If that happens insurance doesn't work for what should be obvious reasons (claims exceed premiums and insurance companies all go bankrupt).

There are solutions to this problem. Here are some examples:

-Sell insurance to a whole group of people at once, getting both the low and high risk customers
-Screening on the part of the insurance company (e.g. excluding pre-existing conditions, refusing coverage)
-Sell individually with a strong mandate (everyone has to buy it)
-Refusing to cover more predictable costs (e.g. well visits, babies)
-Single payer

So while de-coupling health insurance from employment might be a bad idea, doing so un-solves a huge problem that we would need to re-solve, or health insurance wouldn't work.

Sunday, September 11, 2016

Election Update: 9/11

Many new polls today, on balance they've had a neutral impact and Clinton is holding steady in the low 90s.
  • Good for Trump
    • Clinton +1 in NV
    • Clinton +2 in NH
    • Trump +3 in GA
  • Good for Clinton
    • Clinton +8 nationally
    • Clinton +7 in Ohio
  • Neutral
    • LA Times/USC tracking still at Clinton +1
    • Clinton 2 in FL
    • Trump +1 in AZ
It's 9/11 and just like most other Americans I've done some thinking about where I was and what I was feeling on that horrible day. Shortly after 9/11, President Bush said "they hate our freedoms, our freedom of religion, our freedom of speech, our freedom to vote and assemble and disagree with each other."

I think that's a pretty gross oversimplification of why "they" hate us, but I agree that those freedoms are a fundamental part of America or what it means to be American, and I would add an openness to anyone who want to come here and try to make their own american dream. Donald Trump's candidacy is a betrayal of those freedoms. He's already made this country much worse. Every time he says something racist, he normalizes racism. Every time he says something sexist, he normalizes sexism. Every time he boils down a complex and changing world into a fortune cookie he furthers the idea that complexity is bad, or a vice.

I don't want to write only bad things about Trump. I like Clinton and think she'd be at worst an above average President, and I plan to write more things that constitute a positive defense of her. But Trump's downside far outweighs her upside. In the days after 9/11 President Bush visited a mosque and said Islam is peace. In doing so he set an example for the country. How would Trump have reacted to 9/11 if he were President?

Friday, September 9, 2016

Election Update: 9/9

Today was a slow polling day. I've noticed Fridays-Sundays usually are, apparently pollsters need weekends too. We're holding steady at 91% / 8%.

I did find an updated USC/LA Times tracking poll showing Clinton +1, and a poll showing Trump +7 in Indiana. Both polls are slightly good for Clinton. The Indiana poll barely matters, but the state averages there had her down more than 7 so her odds of winning Indiana improved slightly. But if she wins Indiana she's winning the election in a landslide.

The USC/LA Times poll is more interesting. Nate Silver wrote a long piece on it. Most polls sample voters randomly, and ask who they'll vote for. But the LA Times poll samples the same people over and over, and asks them to assign probabilities to who they'll vote for. It's an interesting approach that might be better than traditional polling. It also forced me to adjust how I treat that poll. Because I know it's the same voters each time, when I get a new LA Times poll I zero out prior polls and only count that poll. Essentially, I'm updating what those specific voters think of the election.

The problem is they seem to have caught a sample heavy on Trump supporters. So far the LA Times poll has consistently shown worse results for Clinton than the national average. I adjust the poll using 538s house effects, but I'm not sure even they fully capture its Trumpiness.

College Football Week 2

It's time for week two. After such a stellar week one, you might think week two looks like a let down. Don't think that. Football is back and we should appreciate what's right in front of us. After all, Arkansas gets a chance to redeem the SEC, we get the holy war, WSU travels to Boise in search of redemption after their annual FCS loss, and we even kick off the week tonight with a top ten game.

I added "if they win / if they lose" scenarios to the right sidebar for each of the top ten games, enjoy!

Happy football y'all!