Looking for correlation and biases in parlays

There has been a couple of threads recently in the Handicapper Think Tank in sportsbookreview.com forum dealing with parlays. They raise the possibility of a correlation or bias between the two bets included in the parlay. Yes, there can be if the two wagers are on the same game. Some sports books will not allow this type of a parlay. Others may limit the ranges involved. For instance, in football, parlays with a spread and total are generally allowed. However, a wager with team A -35 and the total over 45 would not be allowed. That’s because if team A covers, then it’s obviously going to result in a high scoring game. Therefore, there is a relationship between team A and over 45. In MLB, you cannot find any parlays with a run line and the total being allowed. However, you can have a parlay with the money line and total at most sports books.

This raises the questions being discussed on the SBR forum. I’ve looked at closing line data from 2010 to 2017 and created two tables of counts of parlay combinations. Following is the one with wagers on favorites or dogs and totals. I dropped the 407 games where there were no favorites.

Win-Total	Games	Won	Push	Lost	Hit Pct	Good	Pct
Fav-Over	19,236	5,289	576	13,371	30.49%	5,577.0	28.99%
Fav-Under	19,236	5,239	576	13,421	30.23%	5,527.0	28.73%
Dog-Over	19,236	3,733	396	15,107	21.46%	3,931.0	20.44%
Dog-Under	19,236	4,003	396	14,837	22.87%	4,201.0	21.84%
All	19,236	18,264	972	56,736	105.05%	19,236	100.00%

The table has the four possible combination of parlays. The top line represents parlays picking the favorite and over. The counts are the number of times it won both sides, the number of times the favorite won, and the total pushed, and the number lost because one of the side lost. The parlay would have made money if it “Won” or “Pushed”. This is represented in the “Hit Pct” column. Notice that this column adds up to more than 100%. That’s because If the total pushes then both the Fav-Over and Fav-Under are winners. The “Good” column represents the times won plus half the pushes. This addis up to 100%.

There is also some speculation that combining the game location and total might provide some bias. So here is a table representing these combinations.

Loc-Total	Games	Won	Push	Lost	Hit Pct	Good	Pct
Home-Over	19,643	4,766	539	14,338	27.01%	5,035.5	25.64%
Home-Under	19,643	5,250	539	13,854	29.47%	5,519.5	28.10%
Visit-Over	19,643	4,458	453	14,732	25.00%	4,684.5	23.85%
Visit-Under	19,643	4,177	453	15,013	23.57%	4,403.5	22.42%
All	19,643	18,651	1,984	57,937	105.05%	19,643	100.00%

Looking at these tables, the top combinations are the favorites with either over or under, and the home team and under. So, there is some bias. But what isn’t clear is if there is any potential profit when the odds are introduced. I’ll address this issue in my next post. Follow me on Twitter @ole44bill to know when I post this.

If you have questions or want to discuss, you can comment on the blog, email me at ole44bill.gmail.com, or better yet post in the Handicapper Think Tank. There are some smart guys that hang out there and their insights can be valuable.

Two filtering techniques for MLB run line system

This post will use 2 techniques to try to filter the wagers determined by the MLB run line system described previously. The first is to segment the wagers based on some identifiable filters. One of the potential filters involved location and plus or minus runs. I segmented the suggested wagers by location and run line and generated the following table.

Over $1.00 Exp Ret	Number	Bet	Net	Ret/$	Bets/Day
Home Team +1.5	1282	$175,482	$3,415	$1.02	1.8
Home Team -1.5	1491	$149,934	$9,419	$1.06	2.1
Visit Team +1.5	2489	$390,507	$1,314	$1.00	3.5
Visit Team -1.5	667	$66,880	$520	$1.01	0.9

In addition to providing a positive expected return, one would want increasing returns as the expected returns went higher. Of the 4 possibilities, only one showed this consistently. That one was the Home Team -1.5 segment. Also, the expected return of $1.06 was the highest. This filter would produce about 2.1 bets per day, a reasonable number.

The other technique is a graphical one. Here I was interested in looking at the odds to see what impact they might have. To do this I converted the odds to European format. Thus -140 became 1.71 and +114 became 2.14. I then sorted the selected wagers by these new odds and computed a cumulative net. Here the data used was the wagers in the selected group above.

 

Here you notice that the results starting at the lower odds meander around zero net for about the first 25% of the bets, then move dramatically up word. This point corresponds to +140 odds and higher. Combining the two filters provides the following historical results.

Selected	Number	Bet	Net	Ret/$	Bets/Day
Hom-1.5 Odds >+139	1074	$107,400	$10,273	$1.10	1.5

This is a pretty solid return with a low, but decent number of bets per day. So, I’ll be using this criteria for a while and hopefully generate some solid returns.

I’ve opened a topic on Sports Book Review in the Handicappers Think Tank forum (“Run Line Wager Analysis”) to deal with some of these questions. You are welcome to make comments, raise questions, or criticisms there. Follow me on Twitter, @ole44bill, to know when I post further analysis.