Data Sources
Performance Metric (AV)
https://www.pro-football-reference.com/
Salary Information (Cap%)
Odds
https://www.thelines.com/odds/nfl/
https://www.sportsoddshistory.com/nfl-odds/
Education
Bachelor of Arts (Economics), Vanderbilt University
Master of Quantitative Management, Duke University
Experience
Data Analyst, CBS Television, Research
Glossary
General Terms
Dead Cap – financial “penalty” incurred when an organization trades or released a player prior to the expiration of their contract
This amount counts toward the Salary Cap, meaning that organizations should want to limit Dead Cap as much as possible so that they can spend more of the Salary Cap on players in their own organization.
Error – absolute value of the difference between the prediction and the correct (or Actual) value
Line – what a sportsbook predicts will be the result of a certain bet, could vary depending on the sport and the specific metric (points, margin of victory, etc.)
ML (Moneyline) – refers to the type of bet where the bettor predicts which team will win the match or game; instead of predicting the margin of victory, the results of these bets are scaled by the odds, meaning the more favored the winner of the game, the smaller the payout
O/U (Over/Under) – refers to the type of bet where the bettor predicts whether or not the total number of points, runs, etc. will be higher or lower than what the line is for a specific sporting event
However, for this project, when there is a reference to O/U or Over/Under, it is explicitly referring to picking “Over” or “Under” for Win Total Futures bets.
Odds – what sportsbooks use to balance Moneyline bets when the teams are not evenly matched (or the public believes that one team is superior to the other)
Ex: A team that sportsbooks believe should be favored might have -150 odds to win, meaning that a correct bet of $10 on them to win would pay out $16.67 ($6.67 of profit).
Result – the payout of the bet, i.e. the revenue
ROI – Return on Investment; [Revenue – Bet Amount] / Bet Amount
Ex: a bet of $10 with -150 odds would pay out $16.67, which has an ROI of 66.67% ([$16.67 – $10] / $10 = 66.67%)
RROI – Running Return on Investment; average ROI including the nth bet in a series of bets (for this project, the bets are in descending order of [ego – Line])
Salary Cap – the total budget (including Active Roster, Injured Reserve, Dead Cap, etc.) allotted to each organization each year
Can fluctuate season-to-season given certain internal (organizational) and external (league-wide) decisions but is Hard – in that it cannot be surpassed like in other leagues (Soft Cap in the NBA (“Luxury Tax“) and the MLB (“Competitive Balance Tax“))
Terms Generated Specifically for This Project
Note: Scores are scaled to fit a 60-100 grading scale, but that does not mean that there are not outliers (over and under)
Balance – metric for the salary efficiency of the roster on a position group level, represents the long term-efficiency
Teams who are toward the end of a rebuild (i.e. approaching a championship window) will ideally have a higher Balance, as they are comprised of talented recent draftees (young stars on cheaper rookie contracts).
Coach Score– metric for the performance of not only the Head Coach, but the entire the coaching staff
ego – the win total prediction for the respective season
Ex: The ego for the Giants in 2022 was 5.08 wins.
GM Score – metric for the performance of not only the General Manager (or Executive VP, VP of Personnel, etc.), but the entire front office
IMP (Impact on Model Performance) – metric for how the outcomes of the league affect the success of the model – the higher the IMP, the more successful the model
IMP = [(Stability)*(Power Utilization)*(Upset Luck)] / 8.5
ORG Score – metric for the performance of the organizations, not necessarily the best teams of that season as this takes long-term health of the organization into account
Power – the percentage of the total available salary cap being spent on the Active Roster, excluding Dead Cap
Organizations who are contending for a championship (i.e. in a championship window) will typically have a high Power, as they are loading up the roster with expensive players.
Power Utilization – metric for the correlation between Power (before the season) and Wins – the higher the Power Utilization, the more closely the level of spending (Power) relates to the level of performance (Wins)
Power Utilization = (Slope of the Best-Fit Line between Power and Wins) + 1
Stability – metric for the parity of the league – the higher the parity, the lower the Stability (and, in turn, a higher IMP)
Stability = (Standard Deviation of the Win Total Distribution) / (Percentage of Games Decided by 8 or Fewer Points)
Talent – metric that measures the salary efficiency of the roster on a player level, represents the short-term efficiency
Teams who are contending for a championship (i.e. in a championship window) will typically have a higher Talent, as they are loading up the roster with expensive, experienced players.
Upset Luck – metric for how lucky the model is in relation to upsets going the way of the model’s predictions – the higher the Upset Luck, the luckier the model (and, in turn, a higher IMP)
Upset Luck = (Percentage of Components That Win via Upset) + 0.5
*For every upset, there are two Components (teams). One component will be the Favorite and the other will be the Underdog. If the Favorite is an Over team and the Underdog is an Under team, then an upset would result in a Percentage of Components That Win via Upset of 0% as neither Component “won.” On the other hand, if the Favorite is an Under team and the Underdog is an Over team, then an upset would result in a Percentage of Components That Win via Upset of 100% as both Components “won.” The Percentage of Components That Win via Upset used in the Upset Luck is the percentage aggregated over the entire season.*