So why use like games?

My last post about the like games system wasn’t particularly encouraging, so why continue to use it? There are 3 reasons.

 1. Alternative wagers – Last time I said there were 3 primary wager types; spreads, totals, and money line. But there are also parlays and team totals. That means 5 wagers. All of these are also available for the 1^st half. That’s 10 potential wagers. If your model predicts a final score of 77-72, you can make an inference about the 10 wagers. But the set of likely scores provides an opportunity to count the results for each wager.

2.   You can look at the back tested results looking for filters that will eliminate potentially unprofitable wagers. I use expected returns as a filer. There could be other filters like location, favorites or dogs, spread size, or so on.

3.   Like games can be used in conjunction with other methods. The model we’re currently developing will use like games with 2 other predictors; a power rating system and KenPom’s projections.

KenPom projects the final score of every Division I basketball game on a daily basis. It has an unbelievable amount of statistics. The score projections are a subscription service, but it’s very reasonable, $19.95 per year.

Before describing the power system, I’ll look back at how approaches #1 and #2 worked out for the football season. Follow me on Twitter, @ole44bill, to know when this is posted.

Like games in action

I had previously posted a series describing my “like game” system. It’s a system that generates a batch of likely scores for a game. I’m looking at using the generated scores to recommend potential wagers. My historical files go back 11 years. There are 3 main types of bets for each game: spreads, totals, and money lines. Since the like games are selected by using the spreads and totals, one would not expect that there would be very many potential wagers generated. So, I decided to look at the 3^rd type of wager, money lines.

I generated 2 sets of wager files for each of the 4 sports. The 1^st listed all potential money line bets. The 2^nd generated like scores for each of these and dropped those with expected returns less than $1.00. Finally, I looked at all the bets not supported by like games (expected returns less than $1.00). Following is a table of the results.

Sport	Description	Number	Net	Ret/$
CFB	All potential wagers	15,030	-$814.04	$0.946
CFB	Those supported by like games	5,099	-$111.16	$0.978
CFB	Those not supported	9,931	-$702.88	$0.929
NFL	All potential wagers	5,562	-$266.47	$0.952
NFL	Those supported by like games	1,699	-$87.89	$0.948
NFL	Those not supported	3,863	-$178.58	$0.954
NBA	All potential wagers	28,372	-$1,211.22	$0.957
NBA	Those supported by like games	12,791	-$444.34	$0.965
NBA	Those not supported	15,581	-$766.88	$0.951
CBB	All potential wagers	75,124	-$6,812.46	$0.909
CBB	Those supported by like games	32,340	-$4,687.77	$0.855
CBB	Those not supported	42,784	-$2,124.69	$0.950

In 2 of the sports, CFB and NBA, the like games proved better than randomly betting. The other 2, NFL and CBB, like games produced worse results. In all cases, the results were dismal. Does this mean the whole like game concept is worthless? Not necessarily.

In my next post I’ll cover the 3 reasons why I pursued the idea in the first place. I’ll tweet on Twitter, @ole44bill, when I post. If you have comments or questions add them to the blog or Email them to me at ole44bill@gmail.com.