Sports predictive modeling involves data management, predictive models, and information systems to predict specific sport-related outcomes. Predicting the outcome of sporting events most definitely falls within the purview of the field of forecasting. Plus, the massive amount of sport data regarding the outcomes of sporting events makes it possible to undertake significant research concerning the forecasts of those events.
My predictive models use data-mining to collect, clean, process, analyze, and gain useful insight from a large amount of external data. Data-mining starts when methods to collect data is employed and ends when the results and recommendations for a specific system are given through analysis of that collected data. The most important part is the transformation of massive data into a standardized format one can comprehend more easily. My predictive models have the ability to do this daily by processing tens of thousands of data points down to single numbers for superior decision-making.
When constructing a predictive system of individuals or teams, some issues need to be addressed. Any predictive system should order all teams, compare teams, adjust for the quality of the opponents, predict outcomes of games, and predict the game scores and differentials.
SportsTrackerBot™ is a group of analytics-based predictive models made by award-winning, industry-leading, and elite level algorithms. It evaluates and rates ALL teams against daily point spreads for the NFL, NBA, NCAAF, NCAAB, and Soccer or moneylines for MLB, and the NHL. My sports forecasting provides you with automated investing solutions. My performance data provides instant analysis by estimating the likelihood (probability) of an event and ranking teams according to the day's schedule. My proprietary version of a sport beta.
My daily sports betting picks are made by these proprietary, predictive models developed by me, Larry King. All predictive models employ aspects of machine learning (all daily data is trained, gradient descent optimization, etc.) and linear regression (model parameters updated daily, all data is smoothed daily by actual outcomes vs. predicted outcomes, etc). Also, all league-specific models are dilligently monitored daily by several benchmarks, one being the log loss of each model.
First, I determined the key predictors governing how many points, goals, or runs a team is going to score. In basketball, for example, key predictors of points scored includes defensive rebound rates and assists per possessions. It should be noted, my models don't rely just on the actual stats for or allowed by a team, but instead the models use stats adjusted by the types of opponents faced and the differences in teams' scheduling. This strength of schedule adjustment involves a plus/minus ranking system according to league averages. In addition, the expected active rosters are considered by excluding injured players for game projections.
Next, the key predictors become the inputs for linear regressions to find each team's expected points, runs, or goals. This generates a projected final score. From each team's final projected score I'm able to approximate the probability of each team winning and covering the point spread. When a game's projection varies dramatically from the game's line, the predictive models grade it an "A" pick. When the projection is moderately different, it's a "B" pick. And all "C" picks occur when the game's projection is close to the game's line, while the predictive models make no pick at all when the expected results and the betting line run too close together.
How are Against-the-Spread picks made?
For football and basketball, my predictive models provide picks against the point spread. This line is a consensus of multiple sportsbooks. If a team is -5, that means it needs to win by more than 5 points to cover. If a team's projected margin of victory is more than 5 points, that team is my model's ATS (Against The Spread) pick. If the average margin of victory is less than 5 points, the opponent (with a line of +5) is the predictive model's ATS (Against The Spread) pick.
Next, when the predictive models determine which team is the better ATS pick or spread value, it compares the team's expected margin of victory versus the point spread to all the other scheduled teams' expected margin of victories. Only the teams ranked, for expected margin versus the point spread, in the top 20% of the day's schedule are considered for the day's final picks. This is the starting point for the teams that will be selected, they still need to pass a few more filters of rule-based reasoning before making your final daily list. The goal each day is to group together a package of teams that will always find you on
the correct side of probability.
How are my Moneyline picks made?
For baseball (MLB) and hockey (NHL), my predictive models determine which team is the better value. My algorithm translates the moneylines into the sportsbooks' consensus win percentage. For example, say Tampa Bay is listed at +111 and New York is -123. A line of +111 translates to a 47.4% chance of winning, and -123 will translate to 55.2%. Let's say the model has New York's winning probability as 60.2%. That would be a +6.4% value edge for New York and would make the moneyline pick New York -123. The predictive model then compares the team's expected win probability to all the other teams' win probabilities available on the day's schedule. Only the teams ranked, for expected win probability, in the top 20% of the day's schedule are considered for the final picks. This is the starting point for the teams that will be selected, they still need to pass more filters of rule-based reasoning before making your final daily list. My goal each day is to group together a package of teams that will always find you on the correct side of probability.
Information Systems are the networks of hardware and software that people and organizations use to collect, filter, process, create and also distribute data. Time for a quick word association game: If I say, “college football analytics,” what do you think of? Something involving Moneyball or when to go for it on fourth down, right? A standard definition for sports analytics is gathering information and applying it in a way that derives a competitive advantage. Translation: doing what football coaches do, only with more help from computers. Analytics are a path toward a winning edge. They see every game. They separate emotion from reality. They separate what you can control from what you cannot.
The SportsTrackerBot™ models make sports bets daily by casting a wide net. It "automagically" grades ALL games for every in-season sport, plus calculates the proper amount of units to risk (+EV). So each day you get a high volume of consistent, filtered selections.