- Check out the full season 2023 MLB Betting Results
August 14, 2023 MLB Baseball Betting Cards details:
Model 10.0 Debuts today!
Model 10.0:
- I’ve mentioned a certain discomfort in using Fangraphs WP% due to the unknown factor of the algorithm, well I’m getting back to basics with models I’ve built for other sports and just focusing on win rates. I will be using trended data based on recent performance to assess team’s and assign my own WP%.
- This is a big leap as there will be no focus on player stats of any kind.
- The core of the new model version looks at team’s win rates, but I’ve split them into 3 types to assess:
- HOME record L90
- AWAY record L90
- OVERALL record L40
Background:
The Los Angeles Dodgers is a team I happen to see a lot of games of, owing to them being one of the few West Coast teams that are often the only option for late night baseball watching. They are quite dominant, but what recently struck me is how much they dominated the Oakland A’s and the Colorado Rockies. These were both Home Series Sweeps for the Dodgers and I had to dig further into baseball data to see if there was something there I missed to give both those visitors wagers throughout those series.
Generally accepted baseball knowledge is that MLB has the lowest Home field advantage of any sport, somewhere between 2-4%. Looking at 2022 data Home Teams won 52.6% of the time and in 2021 it was 53.1%. Let’s just round this to +3% above .500 on average.
However there is nuance in the data if you slice into it. I prefer to dig into anomalies as they often help you identifiy variables that make sense of the noise in data. The Dodgers have the following Home/Away split records the past 3 years:
- 2021 71.3% Home win rate / 57.5% Away win rate | +13.8%
- 2022 69.9% Home win rate / 65.1% Away win rate | +4.8%
- 2023 YTD 65.5% Home win rate / 55.9% Away win rate | +9.6%
That’s an average difference of +9.4%, much higher than the league average of 6% between overall Home/Away records.
Now bring the Mile High squad of Colorado into the fold and their numbers look like this:
- 2021 59.3% Home win rate / 32.9% Away win rate | +26.4%
- 2022 50.6% Home win rate / 33.3% Away win rate | + 17.3%
- 2023 YTD 45.5% Home win rate / 31.7% Away win rate | + 13.8%
A whopping average of +19.2%. It’s no wonder this team got swept at Dodger Stadium when their large home field advantage doesn’t even get them to .500 at home.
Now obviously there’s factors at each team level that we could dig into to narrow this further. Elevation is key to the Rockies differences and being a frequent road visitor to Dodger stadium could be the counter balance for the Dodger’s success at home. But my focus is only to identify these trends as I think this is something important that needs to be included in my model to properly assess a team’s chances at winning. These splits are real and have large sample sizes to back them up. Each team has their own advantage’s here, with most being in the accepted knowledge range of + 2-4%. Washington is the only team since the start of 2022 to have negative home splits at -1.2%.
Application of knowledge:
Now understanding that Home/Away splits are a bigger factor than being accounted for I changed my model to assess each team based on their L90 performance at Home and Away. This should give a team about 40 games of recent data to use in a given matchup each day depending on which side of the box scorce they are playing.
- I subtract the Home Team’s L90 performance from the Away Team’s and add 50% to get the baseline value for the Away team’s WP%.
- Then I review the overall L40 for both teams and do a secondary adjustment based on that performance to get a final WP%.
The L40 is to identify an emerging trend for teams. This goes into a second research point I discovered. Team’s that overperform expectations are the one’s that will make money in sportsbetting.
- I used KillerSports to run season splits for each team and found there was a 40% correlation between team record improvement the previous season and overall profit betting the moneyline that season.
- The problem however, is knowing whether a team will perform better than the previous season. This is where the L40 come’s in. It adjust’s team WP% on the assumption that this new trend will stick and thus their current WP% needs to be adjusted to reflect their new performance .
- Team’s like Cubs, Baltimore, and Dodgers lead this season’s currentl L40 trends and should prove to be stronger bets than the odds do them justice.
- Team’s like the Dbacks, Miami, and the White Sox lead the bottom of L40 trends and are penalized for that poor performance and their betting spots will be rarer as a result.
Lastly all these changes mean nothing unless I can somewhat prove it out. Well I plugged them in and did some calibrating to the overall formulas and I ended up with the backtested data below.
-Set to 1 unit bet per qualified game.
I am quite satisfied with these backtested results as it shows profit each month except April, including the dreadful month of August before this new version goes live. April I’m not concerned about as I know that each season is anew with change and I’ll need to figure out something to better assess things in the early going of each season. I filled in some back data for 2022, but the results in April weren’t proving profitable and again I believe roster changes play a huge part in early season play. Also players start cold and take some time to get into that groove.
Overall Model 10.0 should bring back the profit and help filter out junk like Rockies and A’s on the road so that smarter bets are made overall.
- Also the eROI level is set to 5%+ and a bet needs to have BE ML odds <47% to qualify ( +113 or better). This is just what backtesting proved to be the optimal setup with the new algorithm.
- Lastly, I’m doing away with the MAIN/ RUNTS designation. There really is no benefit to breaking them out right now based on the criteria above . Instead I will be calling each bet Qualified as I have my sheet set to automatically list out the best bet based on eROI in each matchup to make filling out my final betting slip easier once all the day’s data is filled in.
Bankroll Management:
I’ll repeat the current betting plans
- Each game will be flat bet and the unit size of the day is determined by current bankroll size.
- If the bankroll can accomodate it the unit sizes will increase to be equal to 35 games of churn for the bankroll.
- If the bankroll ever ends a day under 30 games of churn then the unit sizes will be reset to a number where the churn is greater than 35.
- The current unit size is 2 units as the bankroll is just a hair under 105 units.
- 105/35 = 3 units, but I will not round up, only down.
- Also each day I will try to play 25% of the given slate of games, if there are not enough Qualfied bets I will Stretch for the next closest until the 25% minimum is achieved.
- I will be using Round Robin’s as additional leverage as I believe in the model and if the model makes money than leverage can be applied to increase gain’s each day with minimal extra risk. I will try to keep these around half the size of a regular game bet so that they don’t sink the card more when they fail.
- The current cards should make money each day when they go 2-2 on a typical 4 game slate, the RRs will be the top 4 games each day based on ML odds and should also produce some small profit at 2-2.
Today’s Slate:
- A smaller slate today, but already 4 games have been Qualified.
- Two games still missing that I’ll review later today when lines become available
- I’ve placed the 1 unit Round Robin bet on the 4 games for today’s card. – 6 combos @ .167 units each
Last checked @ 12:30 PM
CARD is Final!
- COL +190 last add of day at 12:30 PM (I hate backing this team right now after a disappointing 5 straight losses, but this looks like a great spot to bounce back on a return home)
DAILY CARD:
ROUND ROBIN:
- Bet: 1.00 units total
- Picks: MIA/KCR/BAL/SFG
- combos of 2 for 1.67 units each
- Profit: 1.50 UNITS
Total Daily Results: +10.70 units
*See Glossary for details to help explain terms and other recent model changes.
___________________________________
___
