Baseball and MLB was where it all started for me, so the sport and league has a special meaning and are the ones that I enjoy most handicapping. It’s now ten years since I built my first simple baseball model. It was an Excel-only solution and with 99% certainty I didn’t have an edge on the market. From that first model it got more sophisticated quite quickly and I started building a full scale baseball simulator, which basically played through every plate appearance in a matchup. It also included a bullpen tool, that projected who would pitch next in the simulation. Very cool, if I am aloud to say it myself! :) This framework was then transferred to an Excel application (add-in), which was available for others to download and use for free on my website. I called it The Handicapper.
Description of old versions of The Handicapper can be found here:
https://www.eastsidesportsanalysis.com/the-handicapper-v1.html
https://www.eastsidesportsanalysis.com/the-handicapper-v2.html
https://www.eastsidesportsanalysis.com/the-handicapper-v3.html
https://www.eastsidesportsanalysis.com/the-handicapper-v4.html
https://www.eastsidesportsanalysis.com/the-handicapper-v5.html
At the time I also provided betting picks at the website and on the Finnish betting forum Ylikerroin with some good results to start of. Later on I relied to heavily on just the quantitative projections and the edge disappeared. I simultaneously continued to develop The Handicapper to a web application, which is where we are today. From a baseball/MLB only app The Handicapper has evolved to currently cover five sports/leagues (EPL, NBA, NFL, NHL and MLB). The handicapping framework for the app has been based on bottom up player evaluation to create projections for matchups. As experience and discussions with users have shown, this is not an optimal way of do things and users have shown an interest in more straight forward user interaction. That’s why a new version of The Handicapper will soon be launched with a bit different approach, where the projections are based on team level data (baseball is different, still doing it with a player based approach). A new football (soccer) league will also be added as Serie A will be included.
thehandicapper.net have been providing Premier League betting picks from the beginning of February. The handicapping is done by Valggro, who have had success in betting the Premier League for almost three seasons now. His track record can be found here: EPL Picks Result. I’ll be handicapping this MLB season and picks are going to for sale from the beginning of the Regular Season. Below is a description of how I work and what can be expected.
Last season was pretty good. I made 1098 bets, most on them “early” bets on game day morning (8AM-10AM Finnish time, 1AM-3AM EST). At that time the limits were around 2.5k for Moneylines and 1k for Totals. When the lines gets “sharper” in the afternoon the limits gets higher. The Totals were really good and yielded 32.9 units/106.9% with an 1 unit equal bet strategy. The Sides yielded 16.7 units/102.5%. I beat the closing line with 66.0% of the bets (cumulative CLV 32.3), which was a bit less than I expected. One reason to this is that most bets were “early” bets when lineups are not published and there is a degree of uncertainty with weather conditions. I don’t find this an issue as I believe I have an edge on the market because of the way I work when there is a lot uncertainty. At the same time this means there will be some bets placed I wouldn’t place later on during the day but that’s part of process.
My daily routines start with updating data, which is a automated process. I use several sources to collect the raw data I need, MLB.com, Fangraphs, RotoWire to name a few. I use an own “in-house” model to create player projections and join these with a few projection models from external sources. The key part in my framework is, as I see it, the part where I create probable lineups and bullpen usage for each matchup. Fangraphs RosterResource is an excellent starting point for this. The modelling is very automatized and dynamic so what I do on daily basis is that I have dashboard that shows the projections for each matchup together with key stats I find useful. I visualize last five games for starting pitchers to look if there’s big chances in BB% or FB Velocity, I look at bullpen usage last game and last three games and I look at what kind of weather is projected, to name a few. If I don’t find anything specific that I think has an impact on the model output (projections) I compare the projections with market odds/lines and take a decision on whether to place a bet or not. If there’s something I think that should impact the projections I either adjust them or then just take a “no bet” decision. There can also be cases where everything looks OK but the projections are too off compared to what markets currently thinks, which either leads to a “no bet” decision or a reduced bet size.
Description of old versions of The Handicapper can be found here:
https://www.eastsidesportsanalysis.com/the-handicapper-v1.html
https://www.eastsidesportsanalysis.com/the-handicapper-v2.html
https://www.eastsidesportsanalysis.com/the-handicapper-v3.html
https://www.eastsidesportsanalysis.com/the-handicapper-v4.html
https://www.eastsidesportsanalysis.com/the-handicapper-v5.html
At the time I also provided betting picks at the website and on the Finnish betting forum Ylikerroin with some good results to start of. Later on I relied to heavily on just the quantitative projections and the edge disappeared. I simultaneously continued to develop The Handicapper to a web application, which is where we are today. From a baseball/MLB only app The Handicapper has evolved to currently cover five sports/leagues (EPL, NBA, NFL, NHL and MLB). The handicapping framework for the app has been based on bottom up player evaluation to create projections for matchups. As experience and discussions with users have shown, this is not an optimal way of do things and users have shown an interest in more straight forward user interaction. That’s why a new version of The Handicapper will soon be launched with a bit different approach, where the projections are based on team level data (baseball is different, still doing it with a player based approach). A new football (soccer) league will also be added as Serie A will be included.
thehandicapper.net have been providing Premier League betting picks from the beginning of February. The handicapping is done by Valggro, who have had success in betting the Premier League for almost three seasons now. His track record can be found here: EPL Picks Result. I’ll be handicapping this MLB season and picks are going to for sale from the beginning of the Regular Season. Below is a description of how I work and what can be expected.
Last season was pretty good. I made 1098 bets, most on them “early” bets on game day morning (8AM-10AM Finnish time, 1AM-3AM EST). At that time the limits were around 2.5k for Moneylines and 1k for Totals. When the lines gets “sharper” in the afternoon the limits gets higher. The Totals were really good and yielded 32.9 units/106.9% with an 1 unit equal bet strategy. The Sides yielded 16.7 units/102.5%. I beat the closing line with 66.0% of the bets (cumulative CLV 32.3), which was a bit less than I expected. One reason to this is that most bets were “early” bets when lineups are not published and there is a degree of uncertainty with weather conditions. I don’t find this an issue as I believe I have an edge on the market because of the way I work when there is a lot uncertainty. At the same time this means there will be some bets placed I wouldn’t place later on during the day but that’s part of process.
My daily routines start with updating data, which is a automated process. I use several sources to collect the raw data I need, MLB.com, Fangraphs, RotoWire to name a few. I use an own “in-house” model to create player projections and join these with a few projection models from external sources. The key part in my framework is, as I see it, the part where I create probable lineups and bullpen usage for each matchup. Fangraphs RosterResource is an excellent starting point for this. The modelling is very automatized and dynamic so what I do on daily basis is that I have dashboard that shows the projections for each matchup together with key stats I find useful. I visualize last five games for starting pitchers to look if there’s big chances in BB% or FB Velocity, I look at bullpen usage last game and last three games and I look at what kind of weather is projected, to name a few. If I don’t find anything specific that I think has an impact on the model output (projections) I compare the projections with market odds/lines and take a decision on whether to place a bet or not. If there’s something I think that should impact the projections I either adjust them or then just take a “no bet” decision. There can also be cases where everything looks OK but the projections are too off compared to what markets currently thinks, which either leads to a “no bet” decision or a reduced bet size.
I think my approach could be described as “quantamental”, where the main part is fully automated and quantitative but the final decision includes decision making based on fundamental reasoning. I have tried to minimize the fundamental part but I still think I can achieve a better result by leaving some room for it. It comes down to mostly taking a “no bet” decision when things look “too good”.
The first picks will probably be available on Opening Day, April 7th. Stay tuned and send me an email (hk@eastsidesportsanalysis.com) if you have questions.
Harri
Owner and creator of The Handicapper
The first picks will probably be available on Opening Day, April 7th. Stay tuned and send me an email (hk@eastsidesportsanalysis.com) if you have questions.
Harri
Owner and creator of The Handicapper
