Thinking in Bets: Making Smarter Decisions When You Don't Have All the Facts

by Annie Duke

A hand of poker takes about two minutes. Over the course of that hand, I could be involved in up to twenty decisions. And each hand ends with a concrete result: I win money or I lose money. The result of each hand provides immediate feedback on how your decisions are faring. But it’s a tricky kind of feedback because winning and losing are only loose signals of decision quality. You can win lucky hands and lose unlucky ones. Consequently, it’s hard to leverage all that feedback for learning.

Pete Carroll was a victim of our tendency to equate the quality of a decision with the quality of its outcome. Poker players have a word for this: “resulting.” When I started playing poker, more experienced players warned me about the dangers of resulting, cautioning me to resist the temptation to change my strategy just because a few hands didn’t turn out well in the short run.

Incorrectly interpreting rustling from the wind as an oncoming lion is called a type I error, a false positive. The consequences of such an error were much less grave than those of a type II error, a false negative. A false negative could have been fatal: hearing rustling and always assuming it’s the wind would have gotten our ancestors eaten, and we wouldn’t be here.

Bronowski quoted von Neumann’s response: “‘No, no,’ he said. ‘Chess is not a game. Chess is a well-defined form of computation. You may not be able to work out the answers, but in theory there must be a solution, a right procedure in any position. Now, real games,’ he said, ‘are not like that at all. Real life is not like that. Real life consists of bluffing, of little tactics of deception, of asking yourself what is the other man going to think I mean to do. And that is what games are about in my theory.’” The decisions we make in our lives—in business, saving and spending, health and lifestyle choices, raising our children, and relationships—easily fit von Neumann’s definition of “real games.” They involve uncertainty, risk, and occasional deception, prominent elements in poker. Trouble follows when we treat life decisions as if they were chess decisions.

We make this same mistake when we look for lessons in life’s results. Our lives are too short to collect enough data from our own experience to make it easy to dig down into decision quality from the small set of results we experience. If we buy a house, fix it up a little, and sell it three years later for 50% more than we paid, does that mean we are smart at buying and selling property, or at fixing up houses? It could, but it could also mean there was a big upward trend in the market and buying almost any piece of property would have made just as much money. Or maybe buying that same house and not fixing it up at all might have resulted in the same (or even better) profit. A lot of previously successful house flippers had to face that real possibility between 2007 and 2009.

The goal of the bookmaker is to make sure the amount of money bet on either side is equal, so that the losers essentially pay the winners while the bookmaker just takes their fee. They aim to have no stake in the outcome and adjust the odds accordingly.

One of the reasons we don’t naturally think of decisions as bets is because we get hung up on the zero-sum nature of the betting that occurs in the gambling world; betting against somebody else (or the casino), where the gains and losses are symmetrical. One person wins, the other loses, and the net between the two adds to zero. Betting includes, but is not limited to, those situations. In most of our decisions, we are not betting against another person. Rather, we are betting against all the future versions of ourselves that we are not choosing.

Subjects under time pressure or who had their cognitive load increased by a minor distraction made more errors in recalling whether the statements were true or false. But the errors weren’t random. The subjects were not equally likely to ignore some statements labeled “true” as they were to rely on some statements labeled “false.” Rather, their errors went in one direction: under any sort of pressure, they presumed all the statements were true, regardless of their labeling. This suggests our default setting is to believe what we hear is true.

Truthseeking, the desire to know the truth regardless of whether the truth aligns with the beliefs we currently hold, is not naturally supported by the way we process information. We might think of ourselves as open-minded and capable of updating our beliefs based on new information, but the research conclusively shows otherwise. Instead of altering our beliefs to fit new information, we do the opposite, altering our interpretation of that information to fit our beliefs.

Incorporating uncertainty in the way we think about what we believe creates open-mindedness, moving us closer to a more objective stance toward information that disagrees with us. We are less likely to succumb to motivated reasoning since it feels better to make small adjustments in degrees of certainty instead of having to grossly downgrade from “right” to “wrong.”

If a patient comes into a doctor’s office with a cough, the doctor must work backward from that one symptom, that one outcome of a possible disease process, to decide among the multiple reasons the patient might have that cough. Is it because of a virus? Bacteria? Cancer? A neurological disorder? Because a cough looks roughly the same whether it is from cancer or a virus, working backward from the symptom to the cause is difficult. The stakes are high. Misdiagnose the cause, and the patient might die. That is why doctors require years of training to properly diagnose patients. When the future coughs on us, it is hard to tell why.

To be sure, some of the bad stuff that happens is mainly due to luck. And some of the good stuff that happens is mainly due to skill. I just know that’s not true all the time. 100% of our bad outcomes aren’t because we got unlucky and 100% of our good outcomes aren’t because we are so awesome. Yet that is how we process the future as it unfolds. The predictable pattern of blaming the bad stuff on the world and taking credit for the good stuff is by no means limited to poker or car accidents. It’s everywhere.

A consistent example of how we price our own happiness relative to others comes from a version of the party game “Would You Rather . . . ?” When you ask people if they would rather earn $70,000 in 1900 or $70,000 now, a significant number choose 1900. True, the average yearly income in 1900 was about $450. So we’d be doing phenomenally well compared to our peers from 1900. But no amount of money in 1900 could buy Novocain or antibiotics or a refrigerator or air-conditioning or a powerful computer we could hold in one hand. About the only thing $70,000 bought in 1900 that it couldn’t buy today was the opportunity to soar above most everyone else. We’d rather lap the field in 1900 with an average life expectancy of only forty-seven years than sit in the middle of the pack now with an average life expectancy of over seventy-six years (and a computer in our palm).

There is a simple beauty in Mill’s insight. On our own, we have just one viewpoint. That’s our limitation as humans. But if we take a bunch of people with that limitation and put them together in a group, we get exposed to diverse opinions, can test alternative hypotheses, and move toward accuracy. It is almost impossible for us, on our own, to get the diversity of viewpoints provided by the combined manpower of a well-formed decision pod. To get a more objective view of the world, we need an environment that exposes us to alternate hypotheses and different perspectives. That doesn’t apply only to the world around us: to view ourselves in a more realistic way, we need other people to fill in our blind spots.

Others aren’t wrapped up in preserving our narrative, anchored by our biases. It is a lot easier to have someone else offer their perspective than for you to imagine you’re another person and think about what their perspective might be. A diverse group can do some of the heavy lifting of de-biasing for us.

Experts engaging in traditional peer review, providing their opinion on whether an experimental result would replicate, were right 58% of the time. A betting market in which the traders were the exact same experts and those experts had money on the line predicted correctly 71% of the time. A lot of people were surprised to learn that the expert opinion expressed as a bet was more accurate than expert opinion expressed through peer review, since peer review is considered a rock-solid foundation of the scientific method. Of course, this result shouldn’t be surprising to readers of this book. We know that scientists are dedicated to truthseeking and take peer review seriously. Arguably, there is already an implied betting element in the scientific process, in that researchers and peer reviewers have a reputational stake in the quality of their review. But we know that scientists, like judges—and like us—are human and subject to these patterns of confirmatory thought. Making the risk explicit rather than implicit refocuses us all to be more objective.

We tend to think about conflicts of interest in the financial sense, like the researchers getting paid by the sugar industry. But conflicts of interest come in many flavors. Our brains have built-in conflicts of interest, interpreting the world around us to confirm our beliefs, to avoid having to admit ignorance or error, to take credit for good results following our decisions, to find reasons bad results following our decisions were due to factors outside our control, to compare well with our peers, and to live in a world where the way things turn out makes sense. We are not naturally disinterested. We don’t process information independent of the way we wish the world to be.

Telling someone how a story ends encourages them to be resulters, to interpret the details to fit that outcome. If I won a hand, it was more likely my group would assess my strategy as good. If I lost, the reverse would be true. Win a case at trial, the strategy is brilliant. Lose, and mistakes were made. We treat outcomes as good signals for decision quality, as if we were playing chess. If the outcome is known, it will bias the assessment of the decision quality to align with the outcome quality. If the group is blind to the outcome, it produces higher fidelity evaluation of decision quality. The best way to do this is to deconstruct decisions before an outcome is known. Attorneys can evaluate trial strategy before the verdict comes in. Sales teams can evaluate strategy before learning whether they’ve closed the sale. Traders can vet process prior to positions being established or prior to options expiring. After the outcome, make it a habit when seeking advice to give the details without revealing the outcome.

For all the scientific research on the battle between our immediate desires and long-term goals, a particularly succinct explanation comes from Jerry Seinfeld, on why he doesn’t get enough sleep: “I stay up late at night because I’m Night Guy. Night Guy wants to stay up late. ‘What about getting up after five hours of sleep?’ ‘That’s Morning Guy’s problem. That’s not my problem. I’m Night Guy. I stay up as late as I want.’ So you get up in the morning: you’re exhausted, you’re groggy. ‘Oooh, I hate that Night Guy.’ See, Night Guy always screws Morning Guy.” That’s a good example of how we struggle in the present to take care of our future-self. Night Jerry is always going to want to stay up late and, if Morning Jerry has no say in the decision, Night Jerry will get his way regardless of what’s in Jerry’s longer-term best interest. When we make in-the-moment decisions (and don’t ponder the past or future), we are more likely to be irrational and impulsive.* This tendency we all have to favor our present-self at the expense of our future-self is called temporal discounting.*

The future we imagine is a novel reassembling of our past experiences. Given that, it shouldn’t be surprising that the same neural network is engaged when we imagine the future as when we remember the past. Thinking about the future is remembering the future, putting memories together in a creative way to imagine a possible way things might turn out. Those brain pathways include the hippocampus (a key structure for memory) and the prefrontal cortex, which controls System 2, deliberative decision-making. It is our cognitive control center.

You’ve called roadside assistance, twice, and both times (after long hold times) spoken with operators who’ve told you someone will arrive “as soon as they get there after responding to your call.” You decide to change the tire yourself, only to discover you have no jack. You’re soaked to the skin and cold. How does it feel? It likely feels like the worst moment of your life. You are likely bemoaning how unlucky you are, wondering why these things always happen to you. You are miserable and you can’t imagine feeling any other way.* That’s how it feels in the moment. But if the flat tire had happened a year ago, do you think it would have an effect on your happiness today, or your overall happiness over the past year? Not likely. It likely wouldn’t cause your overall happiness to tick up or down. It would probably have faded to a funny story (or a story you try to make sound funny) told at cocktail parties.

Imagine you go up that same $1,000 in the first half hour but now, over the next hour and a half, you can’t seem to win a hand and lose $900 back, ending the night with a $100 win. How does that feel? Now imagine that you lost that same $1,000 in the first half hour but then went on a winning streak to end the night down only $100. How does that feel? Most likely, you’re pretty glum about the $100 win but still buying drinks for everyone after recovering from that terrible start to only lose $100. So you’re sad that you won $100 and happy that you lost $100. The way we field outcomes is path dependent.

It’s the perfect interaction between past-you, present-you, and future-you. Ulysses recognized that his future-self (along with his crew) would become entranced by the Sirens and steer toward the rocks. So he had his crew fill their ears with wax and tie his hands to the mast, literally binding his future-self to better behavior. One of the simplest examples of this kind of contract is using a ride-sharing service when you go to a bar. A past version of you, who anticipated that you might decide irrationally about whether you are okay to drive, has bound your hands by taking the car keys out of them.

The reason why we do reconnaissance is because we are uncertain. We don’t (and likely can’t) know how often things will turn out a certain way with exact precision. It’s not about approaching our future predictions from a point of perfection. It’s about acknowledging that we’re already making a prediction about the future every time we make a decision, so we’re better off if we make that explicit. If we’re worried about guessing, we’re already guessing. We are already guessing that the decision we execute will result in the highest likelihood of a good outcome given the options we have available to us. By at least trying to assign probabilities, we will naturally move away from the default of 0% or 100%, away from being sure it will turn out one way and not another. Anything that moves us off those extremes is going to be a more reasonable assessment than not trying at all.

The distorted view we get when we look into the future from the present is similar to the stereotypical view of the world by Manhattan residents, poked fun at by the famous New Yorker cover. The cover was a drawing of a map from the perspective of a person from New York. In that view, half the map covers a few blocks of the city. While you can see all the buildings on Ninth Avenue and even vehicles and people, the Hudson River and New Jersey are just horizontal strips. The entire United States occupies the same amount of space as the distances between Ninth and Tenth avenues. Beyond a strip of the Pacific Ocean are three tiny lumps, labeled “China,” “Japan,” and “Russia.” When we forecast the future, we run the risk of a similar distortion. From where we stand, the present and the immediate future loom large. Anything beyond that loses focus. Imagining the future recruits the same brain pathways as remembering the past. And it turns out that remembering the future is a better way to plan for it. From the vantage point of the present, it’s hard to see past the next step. We end up over-planning for addressing problems we have right now. Implicit in that approach is the assumption that conditions will remain the same, facts won’t change, and the paradigm will remain stable. The world changes too fast to assume that approach is generally valid. Samuel Arbesman’s The Half-Life of Facts makes a book-length case for the hazards of assuming the future is going to be like the pres...

Conducting a premortem creates a path to act as our own red team. Once we frame the exercise as “Okay, we failed. Why did we fail?” that frees everyone to identify potential points of failure they otherwise might not see or might not bring up for fear of being viewed as a naysayer. People can express their reservations without it sounding like they’re saying the planned course of action is wrong. Because of that, a planning process that includes a premortem creates a much healthier organization because it means that the people who do have dissenting opinions are represented in the planning. They don’t feel like they’re shut out or not being heard. Everyone’s voice now has more value. The organization is less likely to discourage dissent and thereby lose the value of diverse opinions. Those who have reservations are less likely to have resentment or regret build if things don’t work out; their voices were represented in the strategic plan.

One of the goals of mental time travel is keeping events in perspective. To understand an overriding risk to that perspective, think about time as a tree. The tree has a trunk, branches at the top, and the place where the trunk meets the branches. The trunk is the past. A tree has only one, growing trunk, just as we have only one, accumulating past. The branches are the potential futures. Thicker branches are the equivalent of more probable futures, thinner branches are less probable ones. The place where the top of the trunk meets the branches is the present. There are many futures, many branches of the tree, but only one past, one trunk. As the future becomes the past, what happens to all those branches? The ever-advancing present acts like a chainsaw. When one of those many branches happens to be the way things turn out, when that branch transitions into the past, present-us cuts off all those other branches that didn’t materialize and obliterates them. When we look into the past and see only the thing that happened, it seems to have been inevitable. Why wouldn’t it seem inevitable from that vantage point? Even the smallest of twigs, the most improbable of futures—like the 2%–3% chance Russell Wilson would throw that interception—expands when it becomes part of the mighty trunk. That 2%–3%, in hindsight, becomes 100%, and all the other branches, no matter how thick they were, disappear from view. That’s hindsight bias, an enemy of probabilistic thinking.