By Decree of Gabe Newbot: Measuring the Cost of a Review Prompt

Let's color those spikes a different color, to see them clearly. We'll use a neighbor-similarity rule which, simply put, looks at the closest bars to check whether a bar is a local spike or not. This is notable , but there are two interesting things to notice before jumping to any conclusions: Our "base" curve looks log-normal-ish (like a normal distribution, but skewed and with a tail), whereas the "red spikes" curve looks more "exponential-ish". The spikes really do seem to have some cadence, which holds at any bar resolution (5, 10, 15, 20 minutes). We will address the first point directly in a different article, since this is a huge topic around player behavior and prediction models. But for now let's take one more step deeper into the second point - cadence of the spikes.