Beyond the Bets, greatly expands on a thought, really writes a great article....... Risking your bank roll on trends you can’t explain will cost you in the long run
By beyondthebets On June 21, 2011 · 2 Comments ....
I’m taking a bit of a chance here, since it’s quite easy to misinterpret a few 140-character tweets. But I wanted to highlight a tweet made Monday by Steve Fezzik (@fezziksports), two-time winner of the Las Vegas Hilton Supercontest.
Fezzik tweeted the following:
Be very wary of trends found with data mining. In a sea of data, rare and unusual events – not predictive of the future – are easy to find.
Here are a few more tweets he sent to people on Twitter who replied to him:
Where it gets dangerous is if there is no hypothesis and the data miner is just digging for random trends.
Usually you want to start with a sound hypothesis and than validate it statistically through data mining.
It might sound like too much work, but what Fezzik is suggesting isn’t much different than what you’ve probably done dozens of times as a student in college. All academic research begins with a point of inquiry — you start with a research question and form a hypothesis.
The research question is usually developed through observation, trends, statistics and ideas. For example, let’s say there’s a small, rural town of 750 people who use the Internet at unusually high levels. A researcher won’t say, “Wow, this number is high, but hey, everybody is using the Internet more nowadays. I’m sure lots of other similar towns have similar Internet use levels.”
Instead, they will try to explain why this town is using the Internet so much. Perhaps there are logical explanations, explanations that can be found through either qualitative or quantitative research. Maybe the residents of this town are wealthy. Or maybe the town was hit by a tornado in the past, and these people have an inflated desire to keep themselves informed.
Once you’ve collected your data, you can then perform the same study in other similar towns to confirm or disprove the results. If the hypothesis is sound — and the logic, too —there is a good chance your results are good. But at a minimum, you want to have a theory to explain your findings.
Generalizing findings is dangerous in all research. Random events, odd occurrences, coincidences, etc., happen all the time. Some of them can be explained and others can’t be.
So how do you apply this to sports betting?
You can start by not putting your bank roll at risk on trends you can’t explain. For instance, let’s say the Los Angeles Dodgers have lost seven straight games in NL East ball parks. Do you bet against them the next time they play in an NL East park?
Well, maybe.
But first, you should try to explain why this is happening. One explanation could be that the Dodgers don’t travel west to east very well. If that’s the case, you might want to see how other NL West teams have performed in AL East ball parks. Another explanation — and perhaps a more likely explanation — could be that the Dodgers had to face Roy Halladay, Roy Oswalt and Cliff Lee, and were underdogs in four of their other games against a red-hot team.
In that context, the seven-game skid could be explainable.
Lots of trends are explainable.
But many of them are not. Most of them are random. And even if they aren’t, we rarely take the time to understand why these occurrences are happening. We don’t bother to find out if they are random outcomes or outcomes that have predictive quality going forward.
So, again, don’t put your bank roll at risk on a trend you can’t explain. Don’t rely on sample sizes that are too small. And, of course, don’t be lazy in your research.
There’s a good chance it will cost you in the long run.
