I took a philosophy class (I think) during my freshman year of college, and I still remember a story from that class (for some reason). Socrates places a grain of sand on a table and asks his students whether it is a heap. He adds a second and asks again, then a third and fourth and so on. Ultimately, it is determined that there is no specific point at which a collection of grains of sand is a heap. (This is called the Sorites Paradox for those who are interested.)
I mention this because the crowd had a relatively average week this week:
- Straight-up: 8
- Against the spread: 5
- Over/Under: 9
So, the crowd is better than 50% with straight up picks and against the over/under but well under 50% against the spread.
What do we make of this?
Well, as I mentioned in an earlier post, the wisdom of the crowd is measured over time, so we’re not panicking yet. (As my favorite NFL writer Gregg Easterbrook says, there will be plenty of time for that later.)
But the broader idea that we want to develop is to determine when a crowd is a crowd. This week, we had five predictors. (Thank you to those who submitted predictions!) Is that enough to make a crowd?
For the crowd to have predictive skill, we believe that the highs and lows need to be normalized. This allows biases to be averaged out, and last minute predictions won’t have a strong effect in one direction the another. One game that I recall being affected last season was the Raiders-Chiefs Thursday night game. There were five total predictions, and two predictions came in pretty late that swung the crowd predictions from the Chiefs to the Raiders.
So overall, it would make sense that the prediction truly reflects the crowd when it shows a degree of stability. (I’ll do a more detailed assessment of how to address these statistics in a later post.) As you all continue to provide predictions, we’re excited to see, over the course of the season, when we’ll be able to identify the sweet spot when a critical mass of predictions has been submitted to give us a level of confidence in a given prediction.
Thanks for all your contributions so far. Keep up the good work!