| layout | title | subtitle | cover-img | thumbnail-img |
|---|---|---|---|---|
page |
NCAA Football Predictions |
/assets/img/3DFX9HiP.jpg |
I've been running different predictive models and adjusted line spreads for NFL, NCAA, MLB, and Formula 1 for a few years now. I've done well and my spreads have competed with the vegas line but this isn't always profitable. So here I'm trying to take advantage of different bet strategies on popular sites like FanDuel and Draftkings to produce high probability win's with acceptable ROI. A major shout out to Todd Beck and his Prediction Tracker website (thepredictiontracker.com). If you haven't already, make sure you check out his sight before placing ANY bets. His site pulls in predicted data from approximately 120 systems such as ESPN, Donchess Inferenence, Dunkel Index, Massey Ratings, Sagarin, Beck ELO, PerfomanZ Rating, RP-Excel and more. In addition, the prediction tracker provides an average point spread, standard deviation, median, min, and max from all of these systems. Depending on the number of systems that favor above or below the vegas line, he provides a computer adjusted line (+/- 0.5 based on the % of systems favoring that side).
Only when over 63% of the systems favor one side over the other is there a trigger to adjust the spread. What I'm doing here is focusing on just the games where the computer adjusted line differs from the vegas line (i.e., over 63% of the systems believe the spread should be higher or lower). I then find the expected point spread 2 standard deviations away from the average, which is where 95% of the game results are expected to lie.
Now for where the betting comes in. Most betting sites allow for a 'teaser' bet where you buy points or an alternative spread. Buying 6 points with two bets will give you even money. Now this also increases the chance of the parlay losing. Hence this is still in its evaluation stage. I'm also examining how this performs for just bets that favor the home team.
Below are all the games where the computer adjusted line (lineca) differs from the vegas line (line: taken from midweek line at thepredictiontracker.com). A game marked as 'yes' under the bet column means that the buy6pts (buying 6 points in favor for the betting team) falls outside 2 standard deviations from the norm. For example, say team A (home) is favored to beat team b (road) by 10.0 and the computer adjusted line says team A should instead be favored by 10.5 (thus being the "betting_team"). In this scenerio, the buy6pts column would say 4.0 (the majority of the systems believe team A is going to win by 10, so betting them to win by 4 increases the probability of winning the bet). If 4.0 (buy6pts) is outside of line_2sd (i.e., 2 standard deviations above the mean) to line_sd2 (i.e., 2 standard deviations below the mean) this can be interperted as 95% of the time team A will cover the spread at 4.0. Resulting in a 'yes' under bet. I then make pairs of 2 to place a teaser bet giving +110 to +100 odds.
