Posted on

July 21, 2023 MLB Baseball Betting Cards details:

Day 1 of Model 9.0 a success, a 2-1 victory day with +12.04 unit value. Arizona was looking good, but gave up a 3 run homer in the 8th with one out to go to relinquish their lead. Onward to the future!

No changes to the model today, but I do have something interesting notes to highlight about the 9.0 changes.

  • I previously had a hard BE ML cutoff of -185, this is no longer a feature because it is now all but impossible to have WP% exceed 75%. The reason being is that I added a coin flip factor to the WP% projections. Basically this adds a competing WP% projection of 50% to a teams projection and I average the two, so if a team was projected to win 100% the competing 50% projection would average it down to 75%
  • The reasoning for the competing coin flip add was because games were proving to be more like random coinflips and projected results would show breakage at extreme levels.
  • According to my backtested the highest finalized WP% projection for a team with a qualified MAIN bet is currently 58.6%, where Arizona won over Colorado 11-4 on 4/29. Sure seems like a big winner on its own there, but again the coin flip factor assumes parity and that in the MLB no team is ever truly an imposing favorite.
  • Along with the coin flip factor it means that the WP% of teams will be a lot closer to 50% and thus will often need large postive odds MLs to qualify for a winning bet. This is a win win for the model as we will only need to win less than 50% of games to show profit.
  • Also the model has a large hump for RUNTS to qualify as I automatically ding them 2% off their inverse WP% to the MAIN bet in a matchup to make a bet harder to qualify along with the need to show 17%+ eROI. The parameters are steep, but in nominal terms its not as hard compared to Mains to show value.
    • Take a 55% WP% MAIN bet, to qualify at 5% eROI it needs the BE ML to be 52.25%, a 2.75% difference.
    • A 35% WP% MAIN bet with a 2.75% nominal lead over the BE ML of 32.25% is an eROI of 8.5%.
    • The value of eROI gets exponentially larger the smaller the base BE ML is in regards to nominal differences to the WP%.

The recent changes really up the quality of the bets without forcing hard BE ML cutoffs afterwards. The only hard cutoff is if the eROI gets too hot at 30% for MAINS and 40% for RUNTS, again these baselines are due to nominal differences, but also backtested data shows bets have been unprofitable above these levels for each bet type.

  • The game impact is minimal from this HOT! eROI level.
    • Only 9 MAIN bets that would be removed above 30% eROI, and these bets lost 8/9.
    • 8 RUNTS bets above 40% eROI, all 8 lost
    • So you can see in the backtested data the game count is minimal and it really is a bad sign based on results thus far to bet games at these levels. There likely is something at play such as a key injury that books account for.
      • I do not take the time to seek more data than necessary to run the model as is. I’m trying to analyze every game matchup as quickly as possible each day using a few pieces of criteria to keep it simple enough. I’m willing to shed these few games from my model instead of trying to cram the extra data points such as individual player injuries as new variables, which would greatly increase the time needed for game analysis.

Lastly you may start to notice that about 80-90% of all bets going forward will be AWAY teams. I blame Fangraphs for this and this is because I have to adjust my model 2% to account for Fangraphs current imbalance against them. It could also be the line makers fault as they skew odds heavier towards HOME teams than the WP% is showing. Either way one or both are wrong and my data on the season so far shows there is value to be had by backing AWAY teams at such a high rate over HOME teams.

  • My backtested data on the season shows 170 MAIN bets qualified with the current criteria. 144 of them are AWAY Teams where their average WP% was calculated at 52.3% while their BE ML odds were averaged at 46.4%. An average eROI of 12.7%.
  • These teams ended up winning 78/144 games for a 54.2% win record exceeding the average WP% calculation. Resulting in 116 of current parameters Unit Profit over 688 Units bet, a massive ROI of 16.8%.

If I was making my own projected WP% based on the inputs of all teams I would have a home field advantage factor built in. However, I don’t know Fangraph’s initial algorithm inputs and I can only adjust based on the backtested results to counteract what may be too large of a lever pull overvaluing Home Teams.

Special note:

I’ve found the betting trend reports from OddsShark to be helpful in review also. The recent data shows Away Underdogs have won at about 42% which is similar to Home Underdogs, but there is almost $2K in profit difference if you bet Away Dogs over Home Dogs (Albeit both are negative if you were to bet $100 on every such game this season according to their data). This helps back up the analysis that we should bet more of the Away Dogs.

LATE ADDITION: 6:40 PM COLORADO ROCKIES +200, 5 UNITS/$7.00 Unit Value

*See Glossary for details to help explain terms and other recent model changes.

___________________________________

___

By Taylor Moore

2 Replies to “July 21, 2023 – MLB Daily Sports Betting Card and Odds”

  1. Just found your site a few days ago and its been an interesting follow. How are you making the 90 day rolling wins above expectation average? I want to follow along, but my lines are different, which create different opportunities going both ways.

    Reply

    1. Cool, thanks for following along. I track everything in Excel. I record every game’s score, lines, and projected WP% from Fangraphs. I usually use Win Probability from https://www.fangraphs.com/livescoreboard.aspx?date=2023-07-21, but to review old game data you need to check the team’s schedule page such as this one for Toronto https://www.fangraphs.com/teams/blue-jays/schedule.

      I then have Excel setup to track each day’s WP% for a team along with whether they won or lost a given game. Then Excel is set to do formula magic to compare the median WP% of the last 90 days to the actual Win% for a team over that time frame.

      Looking at Colorado for an example:
      I don’t know why but Fangraphs really dislikes their squad, again I think it has something to do with how they penalize Away teams as Colorado’s home park allows the most runs in the league and they likely hit nowhere near as well on the road. As a result Fangraphs likely penalizes their projected WP% each game.

      Anyways, Colorado has a median WP% from Fangraphs of 37.9% the last 90 days. This is the lowest of any team currently.

      Then reviewing their actual Win record over the L90 I see that they’ve won 31/75 games for a 41.3% actual win rate. This is +3.4% improvement to Fangraphs median WP%. So for today’s game I’m going to add that 3.4% to their projected WP% before factoring in the other team’s adjustment and the coinflip factor. This is because the overall data shows me that Fangraphs has been undervaluing them and pricing their WP% lower than actual performance.

      On the flip side, a team like the New York Mets. I have a median L90 WP% of 55.6% and an actual win rate of 43.1% during that time. Fangraphs has been overvaluing them by 12.5%. I added a bit of a supressor to my formula to cap these valuations at +10% or -10% as I didn’t want to adjust the starting WP% too much based on Fangraphs projections, but right now it seems minimal as only 5 or so teams are currently at the caps and without it the highest breakage would be Royals at -14.2%

      Reply

Leave a Reply