This is a win chart. It’s used in games to describe what percentage chance each player has to win at any given point during play This is a poll. It is used to give people an idea of who’s going to win an election They might look the same but they are very different. [Intro music] Let’s start by looking at a win chart. This is a win chart for a simple best-of-three coin-flip game. We start out with an equal chance of winning. If I win the first flip then I have a 75% chance because my odds of winning best-of-three are greatly increased. But then if you win the next flip we’re back to 50-50. It’s anybody’s game. Next coin flip wins. So that was a win chart. Now here is a tracking Poll done over the last few months of the 2016 election. At first glance, this looks very much like the win chart – it’s got a lot of numbers, and each candidate’s odds of winning seem to fluctuate over time, just like in our coin flip game Now, I’m going to spend the rest of this episode explaining why win charts and polls are different, but, There is one area where they have very similar effects, That area, is Heuristics Heuristics, are rules of thumb for particular systems. Some examples of heuristics in gaming would be: “Never draw to an inside straight” Or, “Shift your infield to the left against right handed hitters.” As players learn more of a system’s heuristics, they get better at whatever the system is, And politics, naturally, have heuristics as well Heuristics comes in 2 type: Strategic and Positional. Strategic Heuristics are long term plans that you try to implement over the course of the game like Establish the running game early to open up the passing game later. Or in politics, lock down super delegates early to discourage competitors. Positional Heuristics, on the other hand, are based on specific situations that may arise during a game like: Knowing to bring in a relieve pitcher. Or in politics, knowing that when you get caught in a scandal, you should have spouse standing at your side for a while, and use a lot of Christian sounding terms like Forgiveness. Now, some games produce win charts that are clear to the players like that straight forward coin flip sample from earlier. But often it’s a lot trickier for players to figure out clear precise win charts for the game they’re playing. And we call this uncertainty “Fuzziness”. That fuzziness comes from complexity, randomness, playing against human who don’t act consistently, and having some information that’s not available to all players. In politics, win charts can sometimes be fairly clear. For example, a good party whip would know exactly how many congressional votes they can count on for an upcoming vote. But political win charts can also be extraordinarily fuzzy. Particularly when it comes to elections. Knowing where you are on a win chart is critical information for Positional Heuristics. If you don’t know how close the game is, you could easily apply the wrong Heuristics to your current situation . And unfortunately, polls are really the only tool candidates have to measure who’s currently ahead. Which brings me to my main point: Knowing where you are in the polls is NOT the same as knowing where you are on a win chart. Political candidates generally use polling information as a substitute for a win chart in order to deploy a Positional Heuristics to their current situation. Toward the end of the 2016 US Presidential election, the poll showed Hillary Clinton having a comfortable lead, so she deployed the standard Positional Heuristics for what you do when you are ahead. Her campaign actions were an incredibly logical deployment of Heuristics The only problem was she wasn’t actually ahead. Meanwhile, using that same polling information, the Donald Trump campaign followed the Heuristics for what to do when you are losing. You solidify your base, so even if you lose you at least retain political relevance post-election. Now it’s easy for us to look at the results of this election in hindsight and say that the winning was brilliant in their strategy. While the loser made a bunch of stupid mistakes, but the truth is the last election both candidates were following standard Positional Heuristics. Had they known their actual positions on the win chart, you can be sure they both would acted very differently. This is what the actual win chart would have looked like the day before the election. At that point the campaigns had already been run and the voters had made up their minds. You could have run that election day 100 times and the outcome would not change. Donald Trump may have been behind in the polls, but it turned out he was way ahead on the win chart. And the only thing that could possibly swung the election at this point would have been something truly unusual. So okay, how exactly is polling different from a win chart? WEll, let’s look at the win chart for a baseball game. Here is what the chart looks like when a team is up 1 to 0 at the end of the first inning. And here is what that win chart looks like when the team is up 1-0 with just one inning left to play. See? Even though the score is the same, the win chart is much decisive towards the end Because there’s less variability left. The losing team only has one more inning to turn things around. And the chart reflects their diminishing chances. This is something that polls just aren’t able to capture. No matter how close you get to the election, the polls won’t look anymore decisive than they appeared near the beginning. And polls have lots of others systemic issues that can make them far less reliable than they might seem. Many of these issues are the same problems that exists in win charts. Imperfect information, randomness, complexity, all of that stuff. But the bigger problems are baked into how the polls are collected. Factors like sample size, question format, the subset of people being polled, self-selective reporting, all of these systemic effects make it so that even with multiple polls to work with, polling can never quite function as a win chart. So given all of that: why do campaigns based their Heuristics on polling? Well, because polling is often the only measure campaigns have to build their relative positions. It’s not great, but it’s the best they got. That said polls also have a powerful effect on voters. And their positional Heuristics. If Hillary Clinton supporters knew just how far behind she actually was on the win chart toward the end of that election, many of them probably would have acted differently. Volunteering, giving money, actually voting. And Donald Trump supporters would have acted differently, too if they knew that he was way ahead. Now, games evolve over time I expect that either we eventually get better at removing the fuzziness from political win charts Or new Heuristics would be implemented to better adapt to that fuzziness. Either way it’s important for us to look at polls as just 1 data point when we decided how to best engage politically. It is easy to see why people put so much stock in polls though On the surface, they can seem to accurate which links us to asking
“Why do voters trust them so much?” If a poll comes out and says that party A is ahead 60-40 with a 4% margin of error Many people read that as 4% Margin of Error? HOT Dang! It’s almost mathematically impossible for party A to lose. Sadly, that is not what Margin of Error means in this case. What is actually means is if we do this poll again, using the exact same process, our next result will likely be within 4% of this result And margin of error doesn’t take into account the systemic issues with polls Or all the time between now and election day. And on top of that, a lot of people subconsciously assumed that polls stack. Meaning that if you’re ahead in more than 1 poll, or if you’re ahead in the same poll consistently overtime, then the polls must be more closely reflecting the win chart! And I mean, that’s understandable mistake to make. Consistency implies predictability in so many parts of out life. Alas, it does not work that way for polls Because polls suffers from systemic effects and those effects impacts polls in a similar way each time the poll’s done. Now don’t get me wrong, polls can be incredibly useful political tools. Particularly the nitty-gritty details that campaigns use for micro targetting which rarely gets reported on the news. There’s also some great math that goes on with the polls. And with poll aggregaters, but for national campaigns, they can terrible guides for positional Heuristics.