Multiple Views

One of the interesting things about this project (and any complex project involving more than one person) is that the participants are going to have a variety of viewpoints, with different history, knowledge, biases, and opinions. As we uncover those, we're going to have to figure out which differences are worthy of discussion in an effort to arrive at a common way of thinking, and which can be don't get in the way of our success. I think Josh's previous post was detailed, well thought out, and was the type of statistically based thinking that I like. I disagreed with much of it.

That's not necessarily a bad thing, for either of us. It's going to give us something to talk about, and will certainly provide an opportunity for one or both of us to learn. The discussion will also serve as a basis for improving the statistical model we're using. Without getting into the specific points I disagree with, I think in general Josh may be making the common mistake of overemphasizing analysis of specific information (ie 'how has Wandy Rodriguez done at home"') that may have occurred somewhat randomly, and unemphasizing general rules based on larger sample sizes (ie 'where does home field advantage come from, and who stands to benefit the most?') I think both of us stand to benefit from learning to properly quantify the balance between the two.

There was actually a really good article about one aspect of this at The Hardball Times recently. Here is the link to

Part 1 and Part 2.

As we work through the different views on things, one of the things we'll need to determine is whether we can be effective using the exact same statistical model, or whether we each need to maintain our own to be effective. I firmly believe that unless you're comfortable with the 'system' you're using for any type of speculation or forecasting, you can't use it effectively. One of the other two Draftbug Millionaires (who will be introduced in the next day or two) has already decided that he definitely will be using his own model for that reason. I suspect that we may each ultimately maintain our own models each with its own nuances, even if we end up sharing very similar basic approaches.