View New Posts
1. ## Like games system

I recently started a thread with this same name as a way of discussing the like game system posts I was making on my blog. The posts are a series describing the concept, techniques, and some analysis. I plan on continuing the series attempting to exploit the system to identify potentially good wagers.

However, after a few posts I was notified by the SBR administration that I had violated the rules by including a link to my blog. I pointed out that in fact, my blog was directing readers to this forum for comments, questions, etc. I felt it was cross promotion. Nonetheless, the thread was deleted and SBR suggested that I just copy my blog posts here.

I'm reluctant to do so because of one of my pet peeves. I don't like having to wade through long posts on SBR, and some of these are lengthy. However, I am interested in anyone's thoughts and have decided to go ahead and post these. The next 3 posts are the 1st 3 in the series.

2. Like games concept

It’s been a while since I last posted. Reality set in, health problems and a 3-week hospital stay intervened. I’m ready to resume and will start with a system I call my “like game” system. It’s a system that generates a batch of likely scores for a game. It’s based on the concept that the sports book’s markets are efficient. That implies that all available information about a game is baked in the lines. Thus, if the spread on a football home team is -6, then it is likely that half the games will result in that team covering and the other half not. Likewise, a total of 42 would imply half will finish over 42 and half under (with some pushed).

Using these two numbers, we can determine the expected score for each team. Divide the total by 2. Then subtract half the spread from the total to get the dogs expected score. And add half to get the favorites expected score. In this case, the expected scores are 18 and 24 (totaling 42 with a 6-point spread). These could be plotted on a X-Y graph, with the visit score on the y-axis and the home score on the x-axis. If I had another score set, I would have another point on the graph and could compute the distance between these points by using a simple algebraic formula.

I have data files going back 11 years (for both NFL and CFB) with the lines and the final scores. These files are used to build an internal past like games table. Each table entry contains expected scores of the 2 teams (computed from the spread and total lines), and the resulting final scores. Today’s game spreads are then compared to each of the past games to find the set of past games that are most “like” today’s game. Then the scores of these like games make up the set of likely scores that can be used to evaluate potential wagers.

My next post will deal with the matching process used to select the set of like games. I’ll tweet on Twitter, @ole44bill, when I do. Meanwhile, I’ve started a thread in the “Handicapper Think Tank” forum of www.sportsbookreview.com for comments, criticisms, questions, and suggestions. Its title is “Like game system”.

3. Like games - matching routine

This post deals with how to identify the set of games most like today’s game. I have written a routine that will do that. When first encountered for a sport, the history files for that sport are read and an internal table is built. The internal table contains the expected scores for each team (computed from the spread and total) and the final score for the game. (Note: these are big tables, 2,854 games for the NFL and 7,823 games for CFB.)

The calling code supplies 3 variables, the 2 expected scores for today’s game and the minimum number of games desired. The routine then passes every entry in the table computing the distance between today’s game and the past game for that entry. Remember, consider each pair of expected scores can be considered a point on a X/Y graph. The distance is computed as the square root of the sum of the respective difference in the scores squared. If this is reminiscent of the Pythagorean theory, there is a good reason for that.

These distances are saved in another array, which is sorted by increasing distance from today’s game. Finally, the minimum number of desired games (plus distance ties) are returned to the calling code. Then for any potential wager, these set of actual scores can be counted, percentages calculated, and expected return computed.

I hope this is a clear explanation of what occurs. If not, feel free to ask a question on this blog, Email at ole44bill.gmail.com, or in the “Handicapper Think Tank” forum of www.sportsbookreview.com under the topic “Like game system”. My next post will deal with the question of how to determine what a good minimum number of games is. I’ll tweet on Twitter, ole44bill, when I post it.

4. Like games - setting minimum number of games desired

Now it is time to deal with the question of what to set the minimum number of like games to be used. This is a parameter to the Like Game Routine. The purpose of the system is to identify a set of previous games most like today’s game, and hence predict likely final scores. One technique is to look at the correlation of the average scores of the games in the set compared to the actual final scores.

I wrote a routine to set the minimum number of games at 50, then process each of the games in the past 11 years of data and calling the find like games routine. I saved the average scores from the like games, the expected scores from the spreads and totals, and the final scores. I computed the correlation between these numbers. I repeated this for 100 games, then 150 games, and so on.

Following is a summary for the NFL.

 # Min Games Corr1 Corr2 Corr3 Corr4 50 0.381 0.383 0.360 0.361 100 0.376 0.373 0.360 0.361 150 0.369 0.364 0.360 0.361 200 0.367 0.362 0.360 0.361 250 0.366 0.359 0.360 0.361 300 0.365 0.358 0.360 0.361 350 0.359 0.357 0.360 0.361

Corre1 and Corr2 are the correlations for the like games versus the actual scores. Corr3 and Corr4 are the correlations of the expected scores from the lines and the actual scores. An interesting note is that at 50 games, the like game scores are a better predictor than the lines. It’s not until about 200 games that the lines catch up. I’ve decided to use 200 games because each score in the selected set will only have a 0.5 % impact on the probabilities, while at 50 games each would have a 2% chance.

Here is the comparable chart for CFB.

 # Min Games Corr1 Corr2 Corr3 Corr4 50 0.592 0.596 0.583 0.590 100 0.587 0.593 0.583 0.590 150 0.586 0.590 0.583 0.590 200 0.584 0.589 0.583 0.590 250 0.583 0.588 0.583 0.590

Again, the like games are a better indicator than the lines at the lower end. The line doesn’t catch up until about 200-250 games.

In my next post, I’ll put the like game system to work in looking for NFL money line wagers. I’ll post 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. The forum at sportsbookreview.com is no longer an option. The thread I started on this topic has been deleted. Apparently, I violated a rule by listing a link to my blog.

5. Hey Bill, I'm curious to see if you've pursued this further and if you were able to find any worthwhile correlation.

6. Unfortunately, health issues continue to impact the work I'm doing. I've been hospitalized twice recently (currently in hospital), but I've continued my analysis. I've fallen behind in the documentation side. Doing analysis takes precedent over writing about it. I've done a lot of work and do have plans to document the results. I'll be putting the details on my blog, and summarize them in this forum. Soon I hope for both.
Points Awarded:
 semibluff gave Bsims 1 Betpoint(s) for this post.

7. Sorry to hear about your health issues. I hope you get better soon. We all get caught up in trying to get an edge and winning money, but health trumps all of that.

Originally Posted by Bsims
Unfortunately, health issues continue to impact the work I'm doing. I've been hospitalized twice recently (currently in hospital), but I've continued my analysis. I've fallen behind in the documentation side. Doing analysis takes precedent over writing about it. I've done a lot of work and do have plans to document the results. I'll be putting the details on my blog, and summarize them in this forum. Soon I hope for both.