I have a powerful lesson to teach you today about the deadly mistakes that bettors make. On Thursday of this week, I was alerted of the following scenario of a possible trend that went for a perfect 13-0:
“The Red Sox are 13-0 Straight Up (SU) in the first game of a series with no rest as a favorite after a game in which they scored in at least five separate innings and it is before the All-Star break.”
Well, on May 2nd, the Red Sox was exactly in that scenario again. They:
- Were playing the first game of a series
- Had no rest
- Came in as a favorite (-186)
- Just played a game where they scored in at least 5 separate innings
- Were before the All-Star break
See that? Not only did the Red Sox matched up with one of those factors, they matched up with ALL 5 out of 5! Historically, when all 5 of those conditions matched up, the Red Sox has gone for 13-0.
So that means the Red Sox must be a really good bet to make, right? (hint: wrong!)
You see, this is a deadly mistake that many sports handicappers fall for. I call it an amateur trap. Just because a game fits perfectly into a certain combination of different trends in the past where they’ve shown a strong performance, does not necessarily mean that it’s an accurate predictor of future results.
When looking at past data in sports, your goal is to identify a signal. This is a signal that can help you predict future outcomes. Unfortunately, a signal isn’t always that straightforward because a true signal is always surrounded by a variety of irrelevant noises. Oftentimes, it is very difficult to truly discard the noises from the signal.
I’ll give you an example. Let’s say that you’re trying to look at the historical data of how many wins a certain team has based on the number of games they’ve played to see if you can predict the day of their next win.
Below is the chart of the team’s wins vs games played. Can you find the signal?
Go ahead. Give it a shot! If you’re like many people, you would give the answer below. It’s also the correct answer:
The problem is that when sports handicappers look at past data, they find it difficult to separate the noises from the signal. Rather than creating an ideal signal that is most predictive of future results, handicappers try to be as accurate as possible with past results, believing that it helps predicts future results more accurately.
Here’s the problem: Back-fitting data will give you the most accurate past results, but it is a poor predictor of future results.
When most sports handicappers try to identify a signal from the chart above, here’s the common answer that they give. It’s also a bad answer:
![]() |
You see: By trying to be too accurate with a reading of past data, you end up being less accurate when predicting future results!
Let’s take a look again at the prediction made at the top of this email:
“The Red Sox are 13-0 Straight Up (SU) in the first game of a series with no rest as a favorite after a game in which they scored in at least five separate innings and it is before the All-Star break.”
The above is a classic example of a sports handicapper trying to be too accurate, and it ends up backfiring in a big way. By trying to be too accurate, you actually end up being worse off!
Guess what happened that day? The Boston Red Sox got destroyed by their opponent on May 2nd, even as -186 heavy favorites heading into the game.
So much for a trend that’s 13-0.On the surface, the trends together they look very strong because who can argue against 13-0, right? But the problem with this strategy is that you’re trying to be too accurate.
To see if these are signals or noises, you have to treat the data variables individually before you treat them as a whole.
For example, is the Boston Red Sox’s straight up record still exceptionally good historically if they just play another game when they “had no rest”?
In other words, you need to cull this one data variable out of the combination and test to see if it holds up.
If Boston’s record isn’t exceptionally good in general when they play a game after having no rest, then including it in as a combination to make a certain combination look better (13-0) is an example of how by back-fitting data as much as you possibly can into a signal to make it as accurate as possible historically, you end up actually creating a far less accurate predictor of future results.
Likewise, is Boston a team that just performs exceptionally better before the All-Star break than they do after it? If this one data variable alone does not hold up on its own, then including it in to make a record to be 13-0 is an example of back-fitting data to make a historical result seem more accurate.
When you back-fit data as much as you can to create a signal that it as historically accurate as possible, you end up actually creating a far less accurate predictor of future results!
Simply put: By trying to be too accurate with a reading of past data, you could end up being less accurate when predicting future outcomes. This happens more frequently when there’s an array of different arbitrary data variables. Even if they make a great trend together, it is extremely difficult to pin-point which data variables are positive contributing factors and which ones aren’t (For example: is it the playing the first game of a series? Is it the having no rest? Is it the being a favorite? Is it about just playing a game where they scored in at least 5 separate innings? Is it the before the All-Star break condition?)
The more data variables you include in order to produce a more accurate reading, the more you are actually diluting the effectiveness of the actual signal, and the more likely it is that you can open yourself up to a more probable case of a false positive. The 13-0 trend at the top of this email is exactly just that: A false positive.
Unsurprisingly, the Red Sox lost on Thursday even as -186 favorites heading into the game.
I urge you to be very careful when reading past historical data in sports. To be a successful sports handicapper, you must be able to separate irrelevant noises from an actual true signal.