Where Information Lives

EMC Journal

Subscribe to EMC Journal: eMailAlertsEmail Alerts newslettersWeekly Newsletters
Get EMC Journal: homepageHomepage mobileMobile rssRSS facebookFacebook twitterTwitter linkedinLinkedIn


EMC Journal Authors: William Schmarzo, Greg Schulz, Jason Bloomberg, Jordan Knight, Mat Mathews

Related Topics: EMC Journal, 21st Century Management, cloudbigdataajaybudhraja

Blog Feed Post

How the Eagles and Patriots Can Avoid a Championship Let Down: Play the Percentages

Figure 1: ESPN NFL Gamecast

It’s Super Bowl LII week and the Philadelphia Eagles and New England Patriots are on the precipice of a championship. One game to decide it all, where one decision could swing the fortunes of either team, depending on a single play call. What can Bill Belichick or Doug Pederson do to avoid a letdown or propel his team to victory?

It’s simple – play the percentages.

We only need to look back one year to observe how a coach’s thought process may have prevented his team from claiming a Super Bowl title.

It was Super Bowl LI. Tom Brady had just thrown an interception that the Atlanta Falcons returned for a touchdown. Atlanta held a 21 – 0 lead with 2:21 left in the first half. By ESPN’s projections, the Falcons at this point in the game had a 96.6% probability of winning.

Jump to the third quarter, the Patriots, then trailing 28-3, faced fourth-and-3 at their own 46-yard line with 6:04 on the clock. Tom Brady completed a pass to Danny Amendola for 17 yards. This single play, yielding a Patriots first down, extended the Patriots offensive drive and increased their “Win Expectancy” from 0.2 percent to 0.5 percent (+0.3 percent increase).

It wasn’t until the Patriots scored with 57 seconds remaining in the game to force overtime that they rose from handicappers’ deathbed and evened the game’s win probabilities (see Figure 2). The Patriots ultimately won the game in overtime, overcoming seemingly insurmountable odds.

Understanding Probabilities to Win

How did the New England Patriots achieve such an unlikely comeback? Or maybe more relevant – how could the Atlanta Falcons commit such a mind-boggling, unprecedented choke job?

Figure 2: Patriots versus Flacons Win Expectancy Super Bowl LI

 

Let’s look to the card table to learn how basic probabilities can help humans make better decisions. From “A Blackjack Pro Explains How Ignoring the Odds Cost the Falcons the Super Bowl”, each decision in blackjack can be dictated by simple probabilities. The average blackjack player loses about 3 percent of his or her money. However if probabilities are played correctly, the house’s edge reduces to about 0.5 percent.  Unfortunately, even when humans know the right action to take, very few people actually play the probabilities correctly because humans are overwhelmed with cognitive biases.

Like casinos’ algorithms that determine the odds and outcomes of everything from slot machines to roulette, NFL front offices would benefit from applying Machine Learning to analyze thousands of football games played over the past 10 years. They could then analyze all possible situations and calculate the probability of each outcome. Then all a coach needs to do is to follow the math. But like in blackjack, it can be hard to stay focused on a statistical-based strategy under the stress and excitement of the moment.

Up 28-9 with two minutes left in the third quarter, the Atlanta Falcons had a 99 percent chance to win Super Bowl LI, but then the Falcons ignored simple probabilities that compounded bad decisions:

  • First, Atlanta quarterback Matt Ryan did not let the play clock run down to fewer than 10 seconds on every Falcons offensive possession.
  • Second, by not running the football once they had a late lead (Falcons were gaining an above-average 5.8 yards per rush), they allowed the clock to stop on incomplete passes.

Both of these decisions gave the Patriots – and Tom Brady – more time to get back into the game. All Atlanta needed to do was execute a simple “run the ball” strategy and reduce the number of Patriots’ offensive possessions by one. Unfortunately for the Atlanta Falcons, their decision making was akin to hitting on a 15 in blackjack when the dealer had a six showing. The Falcons ignored basic probabilities and the result was the biggest turnaround in Super Bowl history…at their expense.

Humans Aren’t Good at Making Decisions

Human decision-making capabilities have evolved from millions of years of survival on the savanna. Necessity dictated that we become very good at recognizing patterns and making quick, instinctive survival decisions based upon those patterns (see the blog “Human Decision-Making in a Big Data World”).

Unfortunately, humans are lousy number crunchers. Consequently, humans have learned to rely upon heuristics, rules of thumb, anecdotal information, intuition and “gut” as our decision guides.

So what can Bill Belichick or Doug Pederson do to overcome our natural decision-making liabilities and avoid their teams from becoming the next Atlanta Falcons? It starts with acknowledging and understanding our inherent decision-making or cognitive bias flaws.

Awareness is the starting point and while I could easily write a book on the subject, let’s cover a few of the more common decision-making traps with recommendations on how to manage around these traps.

Type of Human Biases or Decision Traps

Trap: Over-confidence

Over-confidence is when a decision maker places a greater weight or value on what they know and assumes that what they don’t know isn’t important.

Corrective Action: The Falcons entered the Super Bowl with the second-ranked passing offense in the NFL in 2016, while also boasting the fifth-best running attack. However, when it became crunch time, Atlanta leaned on their passing game. To their detriment, they ignored their running attack, certain their MVP quarterback Matt Ryan could finish the job.

Had the Falcons’ coaching staff leveraged Machine Learning, they might have identified variables that might have provided better predictors of in-game performance (e.g. Tom Brady’s excellence in the 4th quarter, the Patriots late-game defensive tendencies), and avoided becoming overconfident in their passing game that wilted in the second half. As a standard operating practice, football front offices should apply Machine Learning to mine the large body of football game data to identify those “known unknowns” and “unknown unknowns” relationships buried in the data.

Trap: Anchoring Basis

An Anchoring Bias is a tendency to lock onto a single fact as a reference point for future decisions, even though that reference point may have no logical relevance to the decision at hand.

Corrective Action: The Falcons, having limited the Patriots to 215 yards plus two turnovers in the first half, had reason to feel good about their defense. However, one half does not make a football game. Buoyed by a first-half performance and a 21-3 halftime lead, the Falcons failed to adapt from what the first half showed them. The Patriots ran 23 plays over their final two first-half possessions, while Atlanta averaged 5 plays per drive in the first half despite scoring 14 offensive points (their third touchdown came via their defense). The length of the Patriots’ respective drives should have alerted Atlanta that New England was poised for greater offensive effectiveness in the second half.

Atlanta failed to identify, validate, vet and prioritize the most relevant in-game metrics to create a more effective second-half offensive game plan to keep the Patriots off the field and stymie their eventual rally.

Real-time advance metrics evaluation to monitor desired behaviors and outcomes is increasingly important to in-game success.

Figure 3: Human Design Making

Trap: Framing Effect

The Framing Effect is a cognitive bias in which a person’s decision is influenced by how the decision is presented. For example, humans tend to avoid risk when a positive frame is presented, but seek risks when a negative frame is presented.

Corrective Action: Into the 4th quarter, the Falcons’ banked on a passing game that achieved 12.3 yards per completed pass in the first half. Had they simply followed the math (and the inevitable “regression to the mean”), the Falcons would have run the ball in order to burn clock and denied the Patriots that one additional possession that ultimately decided the game.

NFL coaches may not be inclined to invest the time to carefully frame the in-game decisions or hypothesis. However had Atlanta – prior to the game – leveraged Design Thinking techniques to create in-game scoring tables and maps that would have charted the potential game flow, they could have referred to proven data, rather than rely on their gut instincts, to ensure they followed the most appropriate in-game decisions.

Trap: Risk Aversion

Risk aversion is the result of people’s preference for certainty over uncertainty and for minimizing the magnitude of the worst possible outcomes to which they are exposed.  For example, a “risk averse” investor prefers lower returns with known risks rather than higher returns with unknown risks.

Corrective Action: Again, in the 4th quarter, the Falcons turned risk aversion upside down. With nine pass attempts to four running attempts, they leaned on lower probability passing plays (passing the ball downfield is a lower probability of success than running) rather than the safely running the ball for first downs.

Falcons coaches did not take the time to understand the impact of Type I and Type II Errors of decisions under different in-game situations (e.g., kick-offs, punts, third-and-long, fourth down, 2 point conversions, overtime). They could have also applied Reinforcement Learning algorithms to create analytic models of different in-game scenarios that objectively balance rewards and risks around desired outcomes.

Trap: Sunk Costs

Sunk costs are retrospective costs that have already been incurred and cannot be recovered. Consequently, sunk costs should not factor into future decisions and should be ignored as if they never happened.

Corrective Action: Coaches in any sport rely heavily on tendencies and patterns. However every situation – even ones that mirror previous games – is unique. Making the same decision in a same situation in a different game does not guarantee the same outcome. It is a difficult habit for coaches to quit, but as the use of data science in sports increases and the use of in-game analytics grows, coaches must  ensure that sunk costs (i.e., previous in-game decisions that can’t be reversed) are identified and less influential to in-game decision making..

Trap: Endowment Effect

Endowment Effect is the hypothesis that people ascribe more value to things merely because they own them. We over-value what we have which leads to unrealistic expectations on price and terms (i.e., stock traders who become attached to a stock they own and consequently have trouble selling it).

Figure 4: The Endowment Effect

 

Corrective Action: The Falcons’ quarterback Matt Ryan was appearing in his first Super Bowl. The Patriots’ quarterback Tom Brady had a history of Super Bowl heroics. Did the Falcons coaches’ overconfidence in their quarterback cloud their judgment and reliance on key predictive performance variables (e.g., quarterback rating, yards after catch, effectiveness under pressure) to guide in-game decisions? Basing analytics models on flawed variables can lead to sub-optimal and even wrong decisions.

Trap: Confirmation Bias

Confirmation Bias is the tendency to interpret new evidence as confirmation of one’s existing beliefs or theories. Confirmation biases impact how people gather information, but they also influence how we interpret and recall information.

Corrective Action: Did the Falcons’ first-half performance confirm a belief that was proven false? In sports, momentum can lead to wild swings in outcome. Did excellence in executing their game plan during the first half, and resulting confirmation bias, lead the Falcons astray in the second half?

This is why sports teams are investing heavily in in-game predictive models with data scientists with expertise in other fields in order to avoid introducing personal biases into the analytic models. The partnership between data scientists, who focus on identifying the most predictive data and algorithms, and coaches, who are responsible for the in-game decisions, represents a new dynamic in the 21st century management of athletic teams.

Other Cognitive Biases of which to be aware include:

  • Herding (Safety in Numbers)
  • Mental Accounting
  • Reluctance to Own Mistakes (Revisionist History)
  • Confusing Luck with Skill
  • Regression to the Mean
  • Don’t Respect Randomness
  • Over-emphasize the Dramatic

Summary

Figure 5: Atlanta Falcons QB, Matt Ryan

Would a basic understanding of probabilities have saved the Atlanta Falcons from compounding a series of small but bad decisions into a painful loss? Maybe not, because understanding and acting are two different things. In the excitement of the moment, humans tend to forget what they’ve been taught and react instinctively.

Awareness is step one, but training is the ultimate solution. Decision makers need to be trained to “take a breath” and consult the models and numbers before rushing into a decision. Research shows that one of the keys to making “clear headed decisions” is to have a feeling of control. NASA and the Navy Seals accomplish that with repeated training [1].

Las Vegas is built on our inherent number crunching flaws and our inability to think with a clear head when the excitement, flashing lights, and pounding music is driving us to use our gut, not our brains, to make decisions.

Don’t think for a moment that those majestic casinos are built by giving away money to gamblers.

Sources:
Figure 2: Win Probability
Figure 4: The Endowment Effect
Figure 5: Photo courtesy of atlantafalcons.com
[1]The Secret to Handling Pressure Like Astronauts, Navy Seals, and Samurai
Additional Sources on Cognitive Biases:
20 Cognitive Biases That Screw Up Your Decisions
Cognitive Bias Codex

The post How the Eagles and Patriots Can Avoid a Championship Let Down: Play the Percentages appeared first on InFocus Blog | Dell EMC Services.

Read the original blog entry...

More Stories By William Schmarzo

Bill Schmarzo, author of “Big Data: Understanding How Data Powers Big Business” and “Big Data MBA: Driving Business Strategies with Data Science”, is responsible for setting strategy and defining the Big Data service offerings for Hitachi Vantara as CTO, IoT and Analytics.

Previously, as a CTO within Dell EMC’s 2,000+ person consulting organization, he works with organizations to identify where and how to start their big data journeys. He’s written white papers, is an avid blogger and is a frequent speaker on the use of Big Data and data science to power an organization’s key business initiatives. He is a University of San Francisco School of Management (SOM) Executive Fellow where he teaches the “Big Data MBA” course. Bill also just completed a research paper on “Determining The Economic Value of Data”. Onalytica recently ranked Bill as #4 Big Data Influencer worldwide.

Bill has over three decades of experience in data warehousing, BI and analytics. Bill authored the Vision Workshop methodology that links an organization’s strategic business initiatives with their supporting data and analytic requirements. Bill serves on the City of San Jose’s Technology Innovation Board, and on the faculties of The Data Warehouse Institute and Strata.

Previously, Bill was vice president of Analytics at Yahoo where he was responsible for the development of Yahoo’s Advertiser and Website analytics products, including the delivery of “actionable insights” through a holistic user experience. Before that, Bill oversaw the Analytic Applications business unit at Business Objects, including the development, marketing and sales of their industry-defining analytic applications.

Bill holds a Masters Business Administration from University of Iowa and a Bachelor of Science degree in Mathematics, Computer Science and Business Administration from Coe College.