Welcome to the new FantasyOmatic!
This page will serve as your glossary of terms and concepts used on the FantasyOmatic website.
What are “Ratings”?
This site uses the concept of “Ratings” as a way to use determine the value of a player, a matchup, a player’s rest of season schedule and a player’s playoff schedule.
It is important to note that the ratings are determined mathematically (not by some “fantasy expert”) by using a performance prediction system based on a proprietary algorithm built upon a supervised method of machine learning. Whew! In English, it’s math.
But, I was told there would be no math!
Don’t worry, you do not need to know math. Math experts trained specifically in sports analytics worked for over 6 years to create and refine the formula that specifically creates these ratings so fantasy footballers like yourself don’t have to do the math.
All you need to do is pick the highest number
The ratings all fall under a 0-100 scale with 100 being the best rating that can be given.
The algorithm that drives the ratings is composed of six parts, one for each of the following positions:
- Quarterback (QB)
- Running back (RB)
- Wide receiver (WR)
- Tight end (TE)
- Kicker (K)
- Defense/Special Teams (DST)
The factors chosen to be included the algorithm are based on hundreds or tests to determine how this data can effect the ability to accurately predict player performance. Only factors that showed consistent positive results were kept in the algorithm, while factors that did not help accuracy were removed.
Then, rather than guessing that the algorithm will work, our team of mathematicians then test the formula against 13 years of past NFL seasons to see if predictions were accurate. Only factors that increase the accuracy of predictions are kept.
QB, RB, WR and TE
For all six positions, the algorithm crunches a ton of real game data (we call them “factors”) including:
- Opposing defense
- Location of game (home or away)
- Playing Surface
- Pass/Rush attempt
- Productive Snaps
- Indoor or outdoor kicking
- Strength of previous opponents
- …and a bunch more
..and then comes up with one, easy to understand, Player Rating.
The Player Ratings are then combined with the “weekly factors” each week to come up with a “Weekly Rating” and aggregate them for the remainder of the season to create a “Rest of Season Rating“.
“Playoff Ratings” are an aggregate of the last three weeks of the season. The closer to the playoffs, the better the accuracy. Again, just when you need it.
All ratings are updated weekly with new game performance data. The more data, the more accurate the prediction. That means that the system is at it’s best when you need it most toward the end of the season and during the fantasy playoffs!
Kicker formula uses future opp red zone TD vs FG and indoor vs outdoor kicking opportunities, FG conversion % ave + more.
For the most part, DST ratings are determined by tracking the frequency of how often they perform “score-able events” (like cause fumbles, Interceptions, sacks, low scores, low yards, etc) and weight that against how often the opposing offense they are facing “allows” those events to happen. The combination is then weighted for home field advantage as well as whether the defense is the vegas “favorite”. This allows the weighting to include vegas book algorithms by default.
What’s up with the Dots?
You may notice that some players have a series of dots under their name, while other do not. To explain, let’s first look at the mathematical concept behind the ratings.
In math, there is a concept of “data samples”. You hear sports folks talk about data samples being “too small” all the time.
In order to draw predictive conclusions, each position category requires a certain amount of weeks of player performance data in order to calculate ratings.
For any player that does not yet have a large enough data sample to meet the preferred minimums, we are still able to calculate the ratings. However, the “confidence” of the rating may be lower due to a small sample size.
Therefore, any player that falls below the preferred data sample size will have these dots. There are five dots, each representing 20% of the preferred data sample size. Once they reach 100%, we remove the dots.
So any player WITHOUT dots showing, has reached their preferred data sample for their career.
So, say you see two players that each have a 80 Player Rating. One of them has dots showing and the other does not.
In that case, we are saying we are “more confident” in the 80 rating for the player WITHOUT the dots, then we are in the other player’s 80 rating. In a certain amount of weeks, the player with the dots will have met the minimum performance data requirements and the dots will go away.