By Harri Kairavuo | eastsideportsanalysis.com
This is the most exciting time of the year for the hardcore MLB fan when the first MLB Spring Training games are just around the corner. Eastside Sports Analysis and The Handicapper are ready for yet another season, which hopefully will be successful on the betting front.
Betting on baseball
Baseball and MLB is the ultimate sports and league for a systematic, high turnover betting approach. To be able to analyze properly up to fifteen (15) games a day there is no other option, than trying to automate as much as possible of the process. Many years of research has led to the baseball community being able to recognize which factors are important for measuring the number of runs scored and allowed in a baseball game. These factors are nowadays no secret to the public.
Even though high quality data is available for no cost, you need to be able to collect it, organize it and maybe enrich it just to get things going. And to be able to keep up with the pace and constantly changing numbers, you need to figure out a way to get the fresh numbers (data) with as little effort as possible.
When you have the automated data collecting figured out you need to build yourself a model which use the data to predict future events. With MLB, it’s pretty easy to get started without that much effort and build a model that predicts the future pretty well. There is a lot of analysis and research that can help you out in selecting good predictors from your data, but because there is so much free data and research available, it means that there is a high probability, that someone else has taken the same approach and built themselves a similar model. This again means, that even if your model could be quite accurate, it probably doesn’t give you an edge in finding +EV bets.
The Handicapper uses a player based model (for all sports), which roughly means that final game predictions (probabilities) are generated bottom up from individual player projections and valuations. The player projection model is a multi-factor model, where the weights for each feature can be determined by the end user. To be able to make statistically solid player ratings all data is normalized and to be able to factor in trends exponential smoothing can be applied to the time-series (eg. time-series of a 10 game moving average of FIP (Fielding Independent Pitching)).
If you take your betting seriously but don’t define yourself as a full-time pro bettor, it’s highly recommended to use some help in your evaluation process. The Handicapper is one option and there are some others on the market. The Handicapper has an edge in the flexibility the user has in model customization. Most competitors have their models fixed, which leads to all users getting the same output (projections) and wagering on same selections. If the model is good and attracts a lot of users, it leads to rapidly falling odds, which again leads to diminishing expected value.
Handicapping Spring Training games
The Spring Training games starts of on Thursday in Arizona when Seattle Mariners takes on Oakland A’s. Making accurate predictions this early in the season is really hard. You don’t know how much lineup rotation there is going to be within a game and who’s pitching and how many innings. By making a lot of research beforehand you might be able to get a clue, but is it worth the effort is a good question. The modelling (within The Handicapper) of a Spring Training matchup differs from a Regular Season matchup. For example the starting lineup is assumed to play 2/3 (6 innings) of a game and the starting pitcher between 1 and 5 innings, depending on how many innings he have done the current Spring Training. The rest (3 innings of batting/fielding and 4-8 innings of pitching) of the projection consists of league average valuations. This approach leads to moneyline probabilities (and odds) being more often close to 50/50 %.
2018 was a season that saw the stabilization of run production for the first time in four seasons (actually decreasing), betting on away teams and under was a profitable strategy and the overall MLB betting market was still very efficient but offering opportunities to the hard working punter. A few interesting notes from 2018:
- Betting 1 unit on all away teams with odds between 1.85 and 2.15 (-117 and +115) [615 bets] would have yielded 21.3 units (ROI 3.5%).
- Betting 1 unit on under on all totals lines greater than 8.0 runs [1448 bets] would have yielded 46.5 units (ROI 3.2%).
A good approach this season (and actually every season) is to be patient and cautious in Spring Training and April games. When you see what the market has to offer and which teams your model (or whatever you process is) is favoring, you can progressively start to raise your wagering to “normal” levels. With dynamic models that are heavily data dependent (like the ones The Handicapper uses), it might be a good idea to let the models adjust to the data. In practice this means, that the projections are usually “getting better” and more accurate when the season proceeds. What this does not necessarily mean, is that there is going to be more betting opportunities later in the season.