When Using Trends, Large Sample Size Important
by Trevor Whenham - 10/03/2008
There is nothing that the public likes more than passing in football. Heck, I would be the worst offensive coordinator in history - I'd just send five guys deep on every down. I respect good running backs, and I know that establishing the run is usually crucial, but watching a well planned, methodical running game isn't nearly as interesting as watching a team throw, throw, and throw some more. I know I'm not alone in that, either. That's why Tom Brady and Peyton Manning are the biggest stars in the league, and no one really likes watching the Ravens these days.
A funny thing happened last week, and it's a blow for the pass-loving public. In the 13 games played, the leading passer was only on the winning team twice. Let me say that again - the team that threw more passing yards lost 11 of the 13 games played last week. The spread didn't make it much better, either - the good passers were just 4-9 ATS. We hold 300 yards passing up in a game as a sign that a QB has played very well. Even that didn't hold up last week. Five QBs went for more than 300 yards, and only two of them played for winners. They only covered two of the five games as well. That doesn't say much for the importance of passing. At least last week.
So, we've learned that passing isn't as important as the public thinks, right? Hold on a second. Let's look at the first three weeks of the season just to be sure. In Week 1, the more prolific passer was 10-5 straight up (the passers were tied in the 16th game). That's the same record they had in week two. In week three they were 11-5. That means that the top passers were 31-15 coming into week four, and were still well over .500 at 33-26 on the season.
This right here is why I love sports betting so much. You can find all sorts of things that are remarkably consistent. Just as soon as you start to think that you are on to something and you have identified something that might be a meaningful predictor, you are thrown a curve ball like this one. Based on the remarkable consistency of the first three weeks, someone might have thought that the right play was to predict which teams would pass more and lay your money down on them. If you'd done that in week four, though, you'd have kissed your bankroll goodbye. Similarly, basing future decisions on what happened in week three probably isn't a road to success, either.
There are a few general lessons to be taken from this for sports bettors. First of all, if you are going to base any betting decisions on statistical patterns, you need to make sure that you are testing them over large sample sizes. You can find a small group of games in which anything happens, but until you can show that the same thing happens week after week, month after month, year after year then you don't have anything useful.
We can look to college football for an extreme example to show the point. Last weekend, Wisconsin lost to Michigan. Michigan had one of the worst first halves in history - 21 total yards and a whole pile of turnovers. Yet somehow they came back in the second half to take the win. Based on just that game you could say that teams that put together less than 25 total yards in the first half are a good second half bet, or that a team that can't hold onto the ball in the first half is a good bet, or that Top 10 teams that are up by at least 19 points at halftime are bad second half teams. None of those would prove out to be true over the long term, of course, but the evidence would seem compelling based on the small sample.
If you are serious about using stats effectively in your handicapping then you need to take these theories you have based on one game or one group of games and then test them over many, many games to see if it holds up. You also have to have a pile of patience because the vast, vast majority of times that you think you have found something meaningful, testing will prove that you don't.
The final thing that this situation reminds us is that you can't get too excited about anything that you discover when you are perusing stats. By it's very nature, sports betting is a conservative game, and those that succeed over the long term don't make decisions that haven't been carefully considered, and that aren't based on sound, proven methods of making them. You will certainly be a better bettor over the long term if you can learn t apply the brakes the next time you find something statistical to get excited about.