And why they matter more than generic poll numbers
For one second, don’t look at a singular poll number. Don’t worry whether your candidate is up by 2, 8, or 20 points. Instead, look at the trend lines. What do I mean by this? What is a trend line? Quite simply, from Google:
A trend line indicates the general course or tendency of something, e.g., a geographical feature or a set of points on a graph.
So, let’s relate this to the presidential election that just concluded in real life. Here’s a picture of the average of all polls between Hillary Clinton and Donald Trump for the last three or so months of the race:
Now, let’s look at what this shows us. First of all, look at where Trump started: he had 39.9%. By election day, he had 43.6%. That matters. Now, let’s look at where Hillary was. At the beginning of August, she had 47.8%. By November 8, she had 46.8%.
Now look at the spreads. Between those first two points, Clinton was +7.9%. By election day, however, she was only +3.2%. What these sets of values indicate to us are very clear trend lines. First, Clinton’s share of the vote barely changed, but it did change for the negative. Trump’s, however, changed a lot. Not only did Trump’s change by nearly 4 points, it also changed for the better. This indicates to us a general trend line that transcended the race’s ups and downs: Trump seemed to continue to rise, while Clinton either stayed the same or shrunk.
There exist, however, smaller trend lines within the bigger picture. Let’s take a look at them.
Here, Trump clearly rose in the polls since the beginning of August. This was because of Clinton’s pneumonia episode, the fading of her –– unusually large –– convention bump, and other factors. This is a trend line that’s positive for Trump.
Here we see the reverse of the previous data point: Clinton clearly pulling ahead and Trump collapsing. This is a positive trend line for Clinton, and was caused by her superior debate performances, the infamous “grab her by the pu**y” tape, and overall terrible month for the Donald.
This here was the final “mini” trend line of the election; it was good for Trump. After the fallout from the FBI letter, Republicans coming home to Trump, and overall race tightening, Hillary saw Trump rise in the polls and she fell.
That last trend line mattered, a lot. You see, both campaigns’ internals showed Clinton up by anywhere ranging from 5–14% points in all the important “must win” swing states. After the FBI controversy, a trend line in Donald’s favor emerged. Those same ranges fell to 3–5% points, per campaign aides for both sides. That’s important, because it signifies the relevance of trend lines versus arbitrary poll numbers. Sure, Hillary was up in all the polling (except for a very few isolated incidents), but Trump was gaining on her –– he had the momentum. The trend line is a perfect way of analyzing and predicting momentum.
So, how does this relate to our ROBLOX bubble? Don’t look at a single data point. First, average multiple points together (but, of course, take into account the weight, or sample size, of each poll). Second, look at the trend lines. Does there seem to be a positive or negative trend for your candidate? Was he or she up by 10 last week, now only by 6? Who’s gaining and who’s losing (in a three+ way race)?
All of these factors matter, and all of them will more accurately predict where a race stands. Think about these, and keep them in mind when you look at where a race stands on any particular day.