While I sort of ranted against gambling last post, now I can be a hypocrite and be a pick-maker. I’ll explain the how in a future post, but for now I might as well get this going while it’s hot. I made a statistical model to predict the likelihood of any given player hitting a home run in a given game. That percentage can be converted into odds to compare to sportsbooks odds now.
Here’s what I do and then I’ll show. For now, I use FanDuel for bets “To Hit a Homer”. The odds are listed as “American Odds” meaning they’re basically decimal odds minus the decimals. A 3-to-1 bet is listed as +300 for example. See below.
So those are the odds FanDuel is offering for Wednesday, April 12’s Astros-Pirates game. Now, here are those two teams’ home runs odds using my model:
A quick walkthrough of this chart. Using this model, each player is given a number. If the model number in the “Bet At” columns is lower than the FanDuel odds, that means we have an edge. In this example, the only edge is Tyler Heineman on the Pirates. He is listed at +1500, meaning a $10 bet will win you $150. The model says a “break-even” odds listing should be at +1380. This means that in the long-run, taking this bet will make you money. For one game, the odds still aren’t great. The model odds give Heineman a 7.25% chance of hitting a home run today, but the FanDuel odds imply a 6.7% chance, meaning if we made this bet 100 times, we could expect a 0.6% profit. So if you made 100 $10 bets, we could expect Heineman to hit a home run 7 times, meaning we would win $1,050. He would also lose 93 times, losing us $930 for a total profit of $120 on the original $1000 bet.
You can see where this turns into playing the long game. The sportsbook odds are usually conservative and much lower than the fair odds. Especially with the popular players, possibly because they get more volume. But the key is to run the odds and make bets on who has an edge and not who the player is. Since the model predictions compared to actual home run rates is within +/- 1%, now we have a steady baseline to bet with. But that doesn’t mean Tyler Heineman is going to win you $150 today. There’s still a 92.5% chance you go home down $10.
So that’s where bet size comes in. After messing around with different strategies, I’ve come up with this one for this year. Each bet takes 1.5% of the bankroll you have that day. Example, on April 10, we put in $100. We placed 10 bets that had an edge at $1.50 each. Two of the ten bets won and the profit for the day was $9.45. So on April 11, our beginning bankroll is now $109.45, and 1.5% of that is now our bet size, or $1.64. This limits the zero win days (of which there are several of these) while a sustained run can grow a steady profit.
This season, from April 3 through April 11 (I missed the first week of the season), the model has shown an edge 49 times, of which those bets have won 9 times for a profit year-to-date of 68%. With the bankroll method, your $100 is now at $ this morning. See the chart below (it’s going up!)
So even though from April 3 to April 6, we lost every bet, we only lost 19% of the starting bankroll. But from April 7 to April 11, when we started hitting some homers, it was a steady climb until April 11 when it was a big 3/7 day and now we are 52.87% up so far. Slow and steady wins the race.
I’m sure you have other questions. Going forward, for the next week I’ll post the chart above, with yesterday’s results and then below you will get this chart (which is the full version of the sample above).
One more thing, you may notice a few blanks next to players. These are games where the pitcher has not pitched enough to have a reliable sample size, or the hitter does not have enough plate appearances for the sample size threshold. A hitter will eventually get there. A pitcher, like a prospect making his debut, will render a whole lineup having no odds because there’s no statistics to make a prediction on. So today, you won’t see odds listed for Aaron Judge even though him going against a rookie making his Major League debut (Peyton Battenfield) seems like as close to a sure thing as there is.
Good luck!