FiveThirtyEight

Nate Silver launched a new version of his FiveThirtyEight website on ESPN a couple of months ago. I liked the previous versions (most recently associated with the New York Times) which mostly covered the horse race aspect of national politics but find the new site a disappointment. In my view the problem is the new site doesn't make enough predictions. Making (and explaining) predictions is useful because it encourages you to develop models that focus on what's important. And it has the commercial advantage of driving traffic as people check back to see how the predictions are changing. During the Presidential election campaign I would check the FiveThirtyEight website regularly to get the Silver's latest odds.

It would have been straightforward to extend FiveThirtyEights politics coverage model to sports. The four major US team sports (baseball, basketball, football and hockey) crown a champion every year. So each year you have the equivalent of a Presidential election campaign. And FiveThirtyEight could offer regularly updated estimates of the chances of each team advancing to each level on the way to the championship. Along with the predictions FiveThirtyEight could have posts explaining the models used to generate them. I would find this interesting just as I found the analogous politics coverage interesting. And from a commercial point of view more people care about sports than politics.

Extending the coverage model to areas other than sports is a bit harder as you don't have the same campaign analogs. And often the data isn't as good. Still there are plenty of things you could try to predict. How will stock prices, federal tax receipts, oil prices etc. evolve over time?  It should be possible to find a variety of things about which interesting predictions could be made.

But instead of systematically developing models and using them to attempt to predict things that people care about FiveThirtyEight has too many posts like this one on how Americans like their steaks cooked.  The internet is full of random data like this and it is unclear why we are expected to find it of particular interest. 

So to sum up, in my opinion the new FiveThrityEight has changed for the worse.