- Check out the full season 2023 MLB Betting Results
August 18, 2023 MLB Baseball Betting Cards details:
Jumping straight from Model 10.0 to Model 11.0 today.
Model 11.0:
In Model 10 I removed Fangraphs WP% and focused solely on modeling games based on teams overall records and recent Home/Away splits. There is some strong correlation here, but I fear I may be discounting the impact of starting pitching overall as a result. Starting pitchers do have a strong say in what the opponents bats will do for about half or more of the game typically. There is some randomness to their performances, where a great starter could blow up occaisonally, but what makes them great is that it’s more the exception than the norm. Model 10.0 ignored pitching dynamics altogether and it needs to be factored in.
I’m trying to avoid tracking individual player stats in my model this season, as it will take some large effort to incorporate and get setup. I think the best plan going forward is to revert back to Fangraphs and use their WP% as I’m pretty sure they would incorporate pitcher dynamics into their formula although I can’t say for certain. My best estimate is that they would since they have a repository of all of MLB data and they are always trying to improve the stats they provide for better context of the game.
In Model 11.0 I will use Fangraph’s WP% as a baseline as I was doing up until 10.0. I will make a small adjustment from previous models and keep the Qualified bet criteria as this was a nice way to remove the MAIN/RUNTS splits where some favored WP% would be excluded. For Model 11.0 all bets will qualify as long as the team in the matchup has %10+ eROI and the ML odds are < 50%, meaning they are all positive odds bets.
The backtested data is similar to past models before 10.0 and this criteria shows strong results throughout the season. August may still be bad, but things can’t be perfect all the time and there will be downtrends. So I will try not to overcompensate for the rocky times this month as long as the overall trend is leading upward in profits.
Backtested results:
I’ve mentioned numbers previously in backtesting, but this time I’ll show things in chart form so the trends can be displayed. I’ll also include March and April this time to provide context overall.
Model 10.0 Backtested Results
Model 11.0 Backtested:
Reviewing the data between the two models one can see that Model 11.0 is superior in that the data is smoother and less volatile, while also ending up at a much higher profit total thus far. Surprisingly though, both seem to follow the same trend with a bad April and a great rebound in the months afterwards. Model 10.0 goes through another downtrend in June as Model 11.0 did not.
Overall this shows to me that focusing on the Winning records and splits has strong correlation to profit performance, but there is more there that needs to be accounted. Unfortunately for me that means I’ll have to continue to rely on Fangraph’s WP% to get better baseline numbers for WP% calculations that winning splits can’t account for on their own. Next season I’ll try to unravel these mysteries further, to improve profit potential, but based on teh results so far this season Model 11.0 should continute to produce profit. I’ll look to make occaisonal adjustments going forward, but it may be small tweaks to eROI cutoffs or how my adjustor factor calculates recent team performance.
Now let’s get back to the profit zone as all the previous missteps decimated the bankroll. Things such as Fade Oakland and too high of a churn level led to the quick fall back to negative realized profits. The Dodgers are just really really hot right now and decimating any underdogs we try to take against them at the moment. I’m not going to fight it, but I’ll allow the model to calculate it appropriately. Model 11.0 would’ve backed off the Milwaukee series, but led Oakland and Colorado to slaughter in the series beforehand. However it did capture a winning series against the Reds so all is not lost as Oakland and Colorado couldn’t catch a break and Milwaukee wasn’t paying enough.
The Round Robins will continue as a 10% play each day of every game as they are just leverage and if the model is profitable they should prove even more profitable over time. Again the issue with Round Robins is you have to churn through some more losses than flat betting so it’s why I’ve set the small amount per card for it.
Today’s Slate:
- Wednesday was a loss of a few units, but wasn’t too bad on a 3-5 record. It kept the bankroll above 90 units which mean the 3 unit plays have another shot before having to drop down to 2 based game churn.
- 30 game churn is threshold before a drop down to lower units, 40 game churn is threshold to raise
- Resets must go down to whole units equal to 35 games at the current bankroll.
- Cincinnati added as a STRETCH bet as I will try to qualify at least 25% of each day’s slate.
Last checked @ 1:00 PM
CARD is FINAL!
- RR placed for 4 games,
- 1.00 units
- 6 combinations of 2
- 0.17 units per combination
DAILY CARD:
ROUND ROBIN:
- RR total profit = +1.77 Units
Total Daily Results: +11.43 units
*See Glossary for details to help explain terms and other recent model changes.
___________________________________
___
