One of the drawbacks to relying on projected scores is that you will be using averages of performance and may not be reflecting the true range of possible outcomes. Consider these outcomes in a matchup where team A is a 10 point favorite over team B:
- A wins 35-10 (A covers)
- A wins 28-21 (B covers)
- A wins 17-14 (B covers)
If those scores were an accurate representation of the range of possibilities (and all equally likely), then Team B would actually cover 2/3 of the time. However, your average projected score would be team A winning 27-15 and from that you would think team A was the better bet! One solution to this problem is to create a program to simulate each matchup thousands of times.
The "Ultimate Prediction Machine" does just this by running our Drive Chart stats through the “simulation engine” for every matchup each week. Five thousand simulated scores are created and each score is compared against the line on the game to assess who would have covered. The numbers are then compiled to give a "confidence value" (e.g.
"Denver has a 58% chance of covering this week”).
Part of the programming of this engine was designed to reflect how teams will change their strategy based on the situation - for instance, when a team is ahead it will often become more conservative on offense, yielding fewer points per drive, but also fewer turnovers. A team that is behind conversely will tend to get more aggressive, leading to more big gains, but also more mistakes. The typical net result of this is that the “UPM” favors the underdog more often than the pure projected scores.
Here's a sample of the UPM output:
