Every NFL off-season provides an opportunity for teams to assess their strengths and weaknesses and with a much publicized effort, they attempt to improve their winning prospects. This process is often a delicate balance of risking dollars for potential wins. The spring and summer of 2017 is no exception, as every roster has been updated along with extensive administrative and management reorganizations, including six head coaching changes. By the time the season kicks off on September 7th, all of the pieces will be in place as each team gets just 16 chances to secure a place in the post-season. It is during each of those battles where a very different game of risk management takes place.

In the same way a fund manager risks dollars to optimize portfolio value or a poker player risks their chip stack, NFL coaches also carefully manage a currency. It is called Game-Winning Chance or “GWC”. GWC represents the expected win probability at any point during the game and is the metric that truly matters. If you are trying to win a game, you are trying to maximize GWC. Just like a skilled poker player, a coach is trying to increase GWC with the full knowledge that it can also be squandered. Opposing coaches are essentially playing a game of heads-up poker with 100 chips at stake and each chip represents 1% of GWC. Depending on the particular matchup they may start with uneven chip stacks, but when the final horn blows, the winning team will have accrued 100% of the chips – the equivalent of a 100% win probability.

Here is the GWC progression from New England’s perspective in the 2017 Super Bowl game between the Patriots and the Falcons.


Equity Progression Chart of Super Bowl LI: Patriots vs. Falcons

From the opening kickoff and progressing through each play increment, the game is essentially a GWC tug-of-war. From the onset, New England was determined to have 52.6% GWC (more on that in a minute) and therefore needed to secure an additional 47.4% before the game came to an end. Conversely, Atlanta had the task of adding 52.6% of GWC to their bankroll. While the path of GWC progression is often a volatile and uncertain one, there is one undeniable truth; the winning team eventually acquires 100% and the losing team will have 0% at the end. GWC is affected by the quality of the opposing players, the efficiency of play execution, the play-calling decisions of the coaches as well as many random factors. Clock, score, replay challenges, timeouts and ball position all factor into the management of the GWC currency.

Understanding GWC at a high level is an intuitive process. Consider an average NFL team facing an equal opponent in the following circumstance:

Score: Leading 21 -14
Ball Position: 50 yard line
Down & Distance: 3rd and 10
Timeouts: 2 remaining
Clock: 10:00 remaining in the 3rd quarter

No one would argue against the fact the GWC of the team in this situation would be improved by adding points to their lead, placing the ball on the opponent’s 40 yard line instead of the 50, decreasing the yards-to-first, adding a timeout or reducing the clock by 10 minutes. All of those factors would clearly improve the team’s prospects for winning the game. However, if you are going to play a game of risk management in the most skillful way possible, you must be able to measure your currency accurately. In other words, how do each of these important factors translate to GWC?

Enter Zeno. A team of game experts and data scientists at EdjSports in Louisville, KY, including the creators of the ground-breaking Zeus football model, have developed a series of risk-management products for NFL and NCAA coaches and managers. The prescriptive tools are powered by a revolutionary simulation engine that was initially built on decades of play-by-play data. Zeno not only can assess GWC at any unique juncture of the game but it can accurately handicap the relative traits of the opposing teams. This is accomplished by an intensive analysis of team performance on every play throughout the season. Through literally billions of trials, Zeno compares every team’s actual performance to its expected performance in GWC on each play. Before New England took the field against Atlanta last Super Bowl, Zeno had already generated a million unique virtual outcomes and performed a pre-game expectation. Afterwards, the relative merits of every decision during that game were also evaluated by Zeno and the GWC progression chart was generated.

Zeno can accurately answer the following questions:

• When does a team’s Game-Winning Chance increase by going for it on 4th down?
• What are the relative values of the rushing and passing matchups?
• What are the GWC merits of a two-point conversion attempt vs. a PAT?
• How deep must an opponent be pinned to justify a short punt?
• When is an on-side kick justified?
• When is it correct to take an intentional safety or voluntarily allow an opponent to score a touchdown?
• How does an incremental improvement in rushing or passing affect seasonal expected wins?
• How much does a penalty or turnover affect GWC?
• What is the value of running down the play clock with a lead?

These are all measurable risk-management decisions that can materially impact the outcome of a game and the overall prospects of making the playoffs. According to Zeno’s analysis, during the last decade the average NFL team squandered between one half and two thirds of an expected win per season on suboptimal 4th down decisions alone. This is a staggering figure when considering it is only one small component of the in-game risk-management portfolio. With overwhelming evidence at hand for the past several years, we might expect there has been an upward shift toward more aggressive actions on fourth downs. However, the evidence does not support this:

4th Down Go for It Trends: 2000 – 2016

Computer models built on strategies similar to those in Zeno have already revolutionized strategic decision making in a number of industries from financial markets to healthcare. The very best humans at nearly every game of skill from Go to chess have been surpassed by artificial intelligence. The modern NFL game is no different and is simply moving at a different pace toward the inevitable insights of big data and computing power. At the end of the day, it will always be a game played by humans on the field but how those humans are guided is on the cusp of a technological revolution.