Back to Home

Star Ratings Explained

On this site you will see star-based Opportunity Ratings and Wager Ratings for each game ...

Star Ratings Explained
The use of star based rating systems is prevalant in many places: finanical bond ratings, movie reviews, product reviews, restaurant reviews etc. They are basically a way of converting long boring decimal numbers into something more visually recognizable and simpler to understand.

On this site, we are using two different rating systems that work in concert with each other:
  • Opportunity Rating (0-3 stars): This rating is driven purely from finding the difference in the spread that our deterministic model predicts for each game vs. the spread offered by the sportsbook that is farthest away from our predicted spread. The greater the difference, the larger the opportunity.
  • Wager Rating (0-4.5 stars): This rating is initiated directly from taking the Opportunity Rating and then running all of the game factors through multiple machine learning (ML) models that have data going back to the 2000 season in the NFL. Using a variety of conditions that have been identified as markers to increase or decrease a wager, we adjust the star rating upward or downward given these conditions. A given Opportunity Rating could move 2 stars in either direction when becoming the Wager Rating. And this Wager Rating is exactly what is used for best sizing in terms of % of bankroll (which is the same thing as a betting "unit"). For example, 2.5 stars means wager 2.5% of the bankroll (or 2.5 units).

    Looking at how this system performed last year, we can see the effect of the wagers the system placed on various star rating combinations.

    We started the system with a virtual bankroll of $10,000. So, 1 star = 1% = 1 unit = $100. Here are the results all wagers with their star rating combinations:

    Opportunity
    Rating
    Wager
    Rating
    Won/Lost
    0.5 0.5 $21.27
    1.0 0.0 -$167.41
    1.0 1.5 -$14.30
    1.0 2.0 $115.74
    1.0 1.0 $1,184.19
    1.5 0.5 $162.91
    1.5 2.0 -$154.88
    1.5 2.5 $324.46
    1.5 1.5 $382.67
    2.0 1.0 -$80.83
    2.0 2.5 -$249.54
    2.0 3.0 $130.89
    2.0 2.0 $159.45
    2.5 1.5 $256.55
    2.5 3.0 -$663.59
    2.5 2.5 $996.64
    3.0 2.0 $249.24
    3.0 2.5 $341.79
    3.0 3.0 -$551.55
    4.0 3.0 $855.40
    4.5 3.0 $1,493.41
    $4,792.51

    So we can see a profit of $4,792.51, which is 47.9% on a $10,000 bankroll over the course of a single season. This is encouraging news, but one season is a small sample size, so seeing how this and future seasons perform will be even more interesting. It is also worth noting that last seaon this system was being adjusted and fine tuned in order to make it even more stable and predictable.

    We can also look at profit based on star rating thresholds that would fit a given person's risk profile. For example, suppose you only wanted to wager on games that were 2 stars and above, or even as conservative as 3 stars and above. Here are the results of a simulation run on all of last seasons predictions:

    Star
    Threshold
    W-L
    Record
    W-L
    %
    W-L
    Amount
    1/2 star or more 52-37-1 58.4% $2,219
    1 star or more 41-30-1 57.7% $2,100
    1.5 stars or more 31-20 60.8% $2,174
    2 stars or more 26-15 63.4% $1,609
    2.5 stars or more 19-10 57.7% $1,654
    3 stars or more 13-6 68.4% $1,276

    This tells us a few things. The first is that this system was profitable at all levels of risk tolerance - all star rating thresholds returned a profit. Not surprisingly, the highest star rating threshold (3 and above) had the best win %. However, that same threshold returned the smallest profit. This tells us that wagering as often as possible when we have even a slight edge will be more profitable that making less wagers and waiting for only the most confident of game situations.

    This site provides every user with an initial "virtual bankroll" of $10,000. While looking at any detail pages for the week's games or while looking at the system's actual wagers it makes each week, you can click on the "wager" button to make your own virtual wagers. The system will track and settle them at the completion of each week during the NFL season. This gives everyone a way to safely test the advice of this system along with their own wagering strategies and preferences.

    One of the themes you will consistentlly hear on this site is that successful sportsbettng is extremely difficult. It is not about vibes or gut feelings. It takes a good deal of research. And you must have a system for choosing the best opportunities AND (just as importantly) bet sizing.

    Of course there are always some elements of luck in play that we need to navigate. Best of luck to you this season!