Remaining Objective
I had the privilege of attending the Macrosports Conference over the weekend and came away with some very interesting conclusions that can (and should) be applied to all sports betting. I figured the title of this article would lead readers to think I planned to delve into the age old adage about the best bettors not having favorite teams; eliminating sentimentality and fan bias from the equation entirely. While those of you that have read my work know that to be my mantra, it’s far from the focal point of this piece. Instead, remaining objective in this context applies to how we analyze data or choose to observe a series of events.
The key note speaker for Sunday’s conference was Nate Silver. While Nate’s best known for his predictive political models, the concepts he applies in that arena can readily be adapted to sports betting as well. When the end goal is building an objective model, the architect shouldn’t approach the situation searching only for data to support his initial hypothesis. There’s a surplus of information right at our fingertips and it’s easy to seek out the desired result if you’re hand selecting information. This concept resonates with every gambler who watches games always believing he/she had the right side. Whether it’s a fumble, ticky tack foul, or a blown call casual bettors can’t see past their vested interest to gain the useful takeaways from a game that will make them money in the future. If bettors don’t remain objective when analyzing the data points, they’re bound to make the same mistake with that exact same team again.
When viewed through a specific lens, our analysis is inevitably shaped to support preconceived notions. The easiest way to illustrate this concept in sports handicapping is predicating a betting decision entirely on trends. Every one of us has found a nice tidy nugget to support a bet we already planned to make before doing our research. We use the existing data, not to drive a calculated decision, but rather to substantiate a conclusion that may or may not be true. Numbers in sports don’t lie and it’s paramount to remove human emotion from empirical models if you’re to achieve long term success by using them. We don’t know if our research is correct until we’ve tested extensively, thus eliminating the false positives that aren’t accurate indicators of future success.
One of the most valuable lessons a bettor can take away from data analysis is not to lose site of the forest between the trees. We always talk about a team never being as good or as bad as their last performance because using just one data point (performance) to make a decision leads to rash impulsive choices. Every bettor weights variables differently (current form, match-ups, history, etc) yet the single end goal of cashing a ticket never changes. Nate Silver used 4 thought processes to illustrate his points that I’ve amended conceptually for applicability to betting sports:
Think probabilistically
Understand the data and what you’re looking at in each scenario. Use deductive reasoning to infer that just because an improbable scenario has manifested itself twice in a short time period it shouldn’t be the basis for all your future business decisions. (Aka using 8 team parlays as long term investment strategies because you hit 2 inside of 1 month)
Know where you’re coming from
Everyone has an agenda that leads you to starting projects; keep that in mind when trying to arrive at conclusions. We all want to be right in our analysis but if the research inputs you valued don’t say betting the Patriots to win the Superbowl is a good investment, trust your data and don’t make the bet. Sports bettors have a knack for “relying on their gut” when numbers don’t support a wager they planned to make, avoid the temptation.
Survey the data landscape (quality over quantity and also need variety)
Garbage in leads to garbage out. Whether its inputs, energy levels, or materials, you can’t create a revolutionary product without max effort. Use the surplus of data at your fingertips to help arrive at the most concrete conclusions rather than short changing your research. Spend the time acquiring box score inputs, historical statistics, and accurate betting data before trying to build an algorithm you think will provide you an infinite supply of winners.
Trial and error
The only way to know if a model will be a useful measure of predictive behavior is through trial and error. Let’s not delude ourselves into thinking a computer program for sports betting that produces a 93% win rate from historical data is indicative of future success. Understand reasonable confidence intervals by continued testing of data points and never grow complacent with your inputs because sports markets are fluid.
