After discovering that my spreadsheet was calculating the odds wrong for every underdog all season, I was really frustrated. The worst part was that I didn’t know if the error meant that my results were better or worse than they should have been. Nothing like sitting at break-even after almost a month, realizing you’ve been executing the system wrong, and not knowing how your error impacted the outcome.
The Next Big Mistake
As I was working through the day’s games on Sunday morning, I realized that something was terribly wrong. It seemed odd to me that the projected wins for each team was decreasing on almost a daily basis, but to this point I had just attributed it to poor offense.
Baseball Prospectus calculates these projected wins based off of the projected stats for each individual player using a set of estimates with regard to playing time. I figured that the numbers were just trending down because the big offenses hadn’t gotten hot yet. I thought it was just a short term trend that would revert any day now.
However, when I added the projected losses to the projected wins for each team on Sunday morning, I started getting numbers that were much lower than 162. If each team plays 162 games in a season, they should project to have that many wins and losses. Something just wasn’t right here. The Tigers just simply aren’t a 77 win team.
After looking around the Projected Standings page for a little while, the answer struck me right in the face. If it was a snake, it would have bit me. The projected wins column represented wins for the remainder of the season. It did not include wins that were already in the books.
Once I understood what the problem was, finding the solution was actually pretty simple. I just loaded the current MLB standings in an adjacent window and added the total wins each team has recorded thus far to the number of wins BP projects them to record over the remainder of the season. After running through each team, all of a sudden the total wins looked a lot more like they did at the start of the season!
How Much Impact Did This Have?
Now that this new error has been discovered and fixed, the big question is whether or not it had an impact on the results thus far. Is it possible that a team like the Brewers isn’t getting enough credit for its hot start? Is it possible that the fund is actually overvaluing the Astros who are on an early pace to be worse than they projected?
At this point, most of the teams have played about 25 games, which represents about 15% of the season. Because the projected numbers that I was using still account for 85% of the season, it seems likely that this error made only a very slight impact on the fund at this point. Had I not caught the error, its impact would have continued to increase as the season progressed.
As I tweeted when I discovered the problem, this is why we paper trade new strategies. It is much better to get the kinks worked out now and have a workable strategy in place long before introducing real capital into the equation.