Well, I increased twice – back to back – after making changes I expected would work.
The second week, which I blogged about, I got two more games correct than the week before and went 10 for 16. Then for the 3rd week I went a whopping 11 for 16 after factoring in home field advantage slightly. Though the next week told a different story. I tried making home field advantage count for more given that it helped the week before, and I ultimately went a disappointing 7 games correct out of 16. A coin toss beats my program statistically.
My thought process was wrong. I could not prove causation (between home field advantage and winning) just from doing better week 3. My current program isn’t going to Vegas – for that matter I’m even holding back on posting it to GitHub. I’m doing this wrong.
So let’s try something else.
What I really need are team stats. By looking at player stats, I’m not doing myself any favors. It makes for far more data to digest than looking at summarizing numbers of a whole team. The problem is finding this data.
I also need to lower my scope. I’m still with the mindset of “the more data the better,” but I’m going to try only looking at games since 2010 this week because of how much professional sports teams change year to year. These last three years are undeniably more relevant to this year, the question is whether seasons 10 years ago are at all consistent. We’ll see what happens.
Now to Sublime Text. With any luck this program will be worth publishing.