US: Never Getting Burned The Same Way Again

On Polling Misses And Pit Stop Strategies

One of the great shames of the French GP this weekend – amongst the obvious sadness of Charles Leclerc fucking himself into the wall – was that we didn’t get to see one of my favourite recurring plots of F1, Red Bull making every race possible a 2 stop.

The basic tension of a decent amount of F1 races is whether or not to run your tires longer, losing lap time as they degrade, and only make 1 pit stop, or to stop twice and have faster lap times – but have to make up that second pit stop’s time lost on track, with fresher tires able to go faster than old ones. It’s a very circuit and race dependent decision, but Red Bull has made a habit of trying to 2 stop races the last year and a half, mostly (it seems) because of the one time they stayed out and got caught.

At the 2021 Spanish Grand Prix, they stayed out for a one stop, while Lewis Hamilton 2 stopped – and mowed down the Red Bull with relative ease. When they rolled up to France that same year, with Max in the lead again, he flipped the Spanish script, went for the second stop, and rolled by Lewis with ease. At Austin, Abu Dhabi, and then this year in Canada and Austria, he pitted early, putting himself on a clear two stop trajectory.

In France this year, he was clearly looking for a two-stop, and ended up on a one just because there was no need to once Leclerc fucked it in the wall, but it’s a fascinating idea, that one moment has informed the strategic tendencies of someone moving forward.

It’s also the cardinal sin of most political analysis in the US, and why the narrative around the midterms is so fucking wildly out of shape.

…

Remember the last first term midterm of a Democratic President? We hear all about the time about 2010, but the thing is, the polls were actually really bad that year. They had Republicans winning the Generic Ballot by 9, winning Colorado and Nevada, and within 10% in a number of races they ended up having no shot in (California Governor, Connecticut and Delaware Senate, etc).

This preceded the national polls being wrong in 2012 by underestimating Democrats, and the polls spitting out dumb shit like “D+4 Michigan” (Obama won it by 9.5%) and “red Florida” (Obama won it by nearly a point). If you had just used the evidence of these elections, you’d be correct in thinking that the polls were historically slanted to Democrats.

In 2016, that was the (il)logic of the position that Democrats were safer than they were – well, that and they had an electoral college bias in their favour, which they did in 2004, 2008, and 2012 – and then the polling error changed. In 2018, the polls were broadly accurate but basically every polling error made sense (Dems beat their polls in left-trending states, GOP beat theirs in red-trending ones), and we all know what happened in 2020. But what’s fascinating to me is that, having gotten the last polling error flip wrong – after believing in 2014 and 2016 that Democrats would beat their polls (and then them very much not), why the commentariat isn’t at least open to the idea of a flip happening again?

It is the case that most of the time, polling misses are cyclical – that’s certainly been true internationally, where polling misses where the right outperformed their polls (Canada 2011, UK 2015, Australia 2019) were either met with cycles where the left outperformed their polls (Canada 2015, UK 2017) or where the left did basically as expected (Australia 2022). The reason why is simple – pollsters tend to not want to make the same mistakes as last time, and if anything, tend to overcorrect for their errors. Is there much evidence of this so far? No, but if you told me that the pollsters are working to do wonky shit with their Likely Voter screens to make them more Republican friendly – despite the fact that the GOP have traded high-propensity, always going to vote in midterms voters for low-propensity voters whose willingness to show up on a random Tuesday without the President on the ballot is at least in doubt – I’d be 0% surprised.

None of this is an argument that Democrats will win the House or some shit, but the consensus is basically that the polls will break to the right because of historical trends (as if a year where Roe gets overturned merits being treated as a “normal” year) or that they’ll be wrong again because, well, they’re usually wrong. Am I guilty of the latter treatment? Extraordinarily so, I know. But the problem is, I can’t stop thinking about the deep consensus that the polls will understate Democrats without thinking of Red Bull’s addiction to the 2 stop at this point.

Red Bull got burned one time not doing a two stop – Spain last year – and their instinct now is always to take the extra stop. In the same way, the pundits got caught believing in Democratic optimism one time, and now their instincts are never to believe in Democrats again. I get it, it worked in 2020, but it also missed heavily in 2018 – remember how there were still people pretending the GOP could win the House into the fall of 2018, or that there wasn’t a “Blue wave” because California takes a while to count votes? This instinct that Democrats are going to fall on their faces is (kind of) justified, but I also remember the Pod Save Bros telling all of us that Trump had no chance because the Blue Wall and polling misses meant we were fine.

The line between a one stop being optimal or not is a very fine one, and sometimes you commit to a two stop too early and you leave yourself as an easy mark to get caught (Austria, 2022), and sometimes it wins you a race that also enables you to win the championship (France, 2021). In the same way, the almost religious belief that Democrats will underperform their polls is similarly on a knife’s edge – it’s possible that the same shit happens again this year, but it’s also possible that 2022 is the start of another flip, and it’s nauseating that this is just being ignored.

I maintain that these Ohio polls are dogshit, and they are – but they’re also from a truly dogshit class of pollsters, except for that one Suffolk poll. But I’m less confident now that the polls have to be wrong in aggregate, because guessing a polling miss is like pit stop strategy – it’s never as obvious as we’d want it to be. And yet, here we are, with the consensus of a red wave mostly unchanged despite a month’s worth of evidence that Democrats have momentum. Will it last? No idea – but it’s probably worth asking what might happen if the polls stay where they are.