Thinking in Bets

Thinking like Annie Duke

When making decisions under uncertainty, it’s important to think in probabilities and scenarios. It’s critical also to watch for heuristics and intuitions that might impair our reasoning. There’s actually a lot to learn here from the strategies of experienced poker players. Once such book is Thinking in Bets, in which World Series poker champion Annie Duke shares her toolkit for making better decision making under uncertainty.

This post will cover the main lessons that I took from Duke’s work, from the way she thinks about uncertainty and betting, to common decision making traps, and the distinction between luck and skill.

A bet is a decision that we make with regards to an uncertain future and comprise of five properties: choices, probabilities, risks, decisions and beliefs. If we think about our bets and their properties carefully, we can make better trade-offs, learn more effectively, and minimise the impact of emotions on our decision making.

Your opportunity cost

Incomplete information and unexpected events outside our control make our bets or investments uncertain. There is an opportunity cost associated with choosing one bet or decision over another. When we make a decision under uncertainty, we are betting against alternative future states that we haven’t chosen for ourselves.

Systems, beliefs and betting

Duke recommends we keep Daniel Kahneman‘s System 1 and 2 framework in mind when thinking about our own decision making processes. System 1 thinking relies on reflex, instinct, impulse, intuition and automatic processes to make decisions. And System 2, by contrast, involves purposeful choice, concentration and expense of mental effort.

System 1 thinking is often responsible for the formation of beliefs. When we hear something new, System 1 has a tendency to either accept or reject the idea, almost instinctively. Only occasionally do we have the time and motivation to verify the facts that underly our beliefs. Additionally, the more pressure we are under, the less likely we are to verify these beliefs.

Since our bets are only as good as our beliefs, Duke notes that an overuse of System 1 to make rational and complex decisions can lead to problems. Hence, there is enormous value in learning to calibrate our beliefs effectively. Taking each opportunity as a chance to learn can yield significant benefits that compound over time. (Mervyn King and John Kay introduce a similar idea in Radical Uncertainty, which they call a Bayesian Dial)

Decision making traps

Many heuristics and biases accompany our decision making processes. While useful in everyday life, they can lead to incorrect conclusions under conditions of high complexity and uncertainty. Duke highlights several decision making traps to watch out for:

Resulting

Resulting is the tendency to equate the quality of our decisions with the quality of its outcome. While resulting is useful for learning in games such as chess, it has less utility in chance-oriented activities like poker and investing. Duke recommends we resist the temptation to change a good strategy when only a handful of decisions have performed poorly thus far.

Hindsight and Rashomon

Many of us are unaccustomed to checking the fallibility in our observations and intepretations. Hindsight bias, for example, describes our tendency to view an outcome as inevitable after the fact has passed. Similarly, the Rashomon effect describes how accounts from different people on the same event may differ due to different facts, perspectives and incomplete information. It’s a reminder that accounts are rarely objective or completely informed.

Motivated reasoning

Motivated reasoning describes our desire to seek evidence to confirm or validate our beliefs and are quick to ignore or discredit new information that would contradict our belief. It’s a self-serving bias. We’ll cherry pick data and proclaim causation if it helps to preserve our self-narrative, intentional or not.

What’s more, we’re great at spotting these heuristics and biases in other people, but poor at recognising it in ourselves. We often think ourselves more objective and reasonable than the rest.

Tilt and path-dependency

Rarely are our feelings based on the average of how we feel about past events. Take investing in stocks for example. We tend to feel sad about breaking even on an investment that was previously valued much higher. Conversely, we tend to feel happier about identical investments that has broken even following a serious dip. In both cases, we sold the stock at break even and our net position unchanged.

Path-dependence and in-the-moment emotions are not always rational or conducive to sound decision making. Bad outcomes may sometimes lead to emotionally or irrationally made decisions. This can then lead to even more bad outcomes and a viscious cycle. In poker, this is commonly referred to as tilt. We have to be careful with the way we frame experience and results. Perspective matters.

Predictably irrational

Dan Ariely has shared similar observations on decision making traps in Predictably Irrational. The behavioural economist describes how individuals tend to measure success in terms relative to arbitrary benchmarks. Such anchors and can cause us to over or under react to success or failures in absolute terms and conflate the effects of other decision-making traps. Furthermore, our tendency to ‘self-herd’ or arrive at decisions using fixed patterns means we can have difficulties with questioning our repeated behaviours. Like System 1 thinking, this can make it difficult to identify and avoid traps such as resulting, hindsight bias and motivated reasoning when it happens.

Temporal discounting

Good processes should lead to good habits and allow for good results to compound over time. However, in-the-moment thinking can sometimes lead to irrational or impulsive decisions. Temporal discounting occurs when we take an irrationally large discount for a reward today over a far larger expected reward in the future. Duke recommends we imagine our future self when making decisions to minimise the risk of temporal discounting. We are more likely to pursue decisions consistent with our long-term goals when we consider the present and future.

Ulysses contract

A Ulysses contract is an action taken today that prevents your future-self from taking self-destructive actions. The contract involves a pre-commitment that can help us to minimise the risk of tilt and encourage deliberative thinking. One example of a Ulysses contract is a decision swear jar: putting $5, $10 or $50 in the swear jar each time we express an idea with unjustified certainty and confidence, or make complaints about bad outcomes that were purely out of our control. (Although swear jars are not true Ulysses contract per se, as it’s easy to abandon the commitment itself)

Skill, luck and learning

When thinking in bets, it is important to recognise the difference between the quality in decision making and luck. We are almost never 100% right or wrong, and outcomes are rarely 100% attributed to skill or luck. We should focus on the decisions we can control and let go of the events and luck we cannot.

Learning from bets

To improve our decision making, we have to learn from our past performance. Outcomes in probabilistic games such as poker or the stock market provide feedback on how our decisions have fared. However, winning and losing in such games provide only weak signals of our decision quality. It can be difficult to separate our skill from luck, and to maximise our learning from such forms of feedback. A lucky streak can make anybody look like a genius. Be careful not to fall victim to this trap.

Probabilistic accounting

For the reason above, we need to keep an accurate representation of what’s happened, and what could have happened. Reflecting on decision trees and scenario plans can help with calibration.

Ultimately, a great decision is the result of a good process and not necessarily a good outcome. A good process makes an informed attempt to incorporate our state of knowledge at the time. This involves acknowledging what we don’t know. Good decision makers must be comfortable with unpredictability. Probabilistic thinking helps to reduce the risk of resulting, hindsight bias and motivated reasoning.

We’re more likely to engage in resulting when we know the outcome. Even if the yardstick is objective, our intepretations are subjective, and sensitive to bias. Undertaking outcome-blind analysis, Duke says, can help us to evaluate decision-making quality with less risk of resulting. This involves assessing the quality of your decision-making strategy and process without knowledge of the outcome itself.

Self-narrative preservation

We have a mirror image problem when it comes to making decisions. Since we desire certainty, the fear of being wrong can cause many to overlook the role of luck and hidden information. Confidence need not be a binary yes or no. In many cases, being unsure is the most accurate representation of reality. Conscious ignorance and the minimisation of ‘black-and-white’ thinking can go a long way.

Talking to others can help us to escape the problem of self-narrative preservation. Groups with diverse viewpoints provide good protection against homogeneity and confirmatory thought. The danger, of course, is if we’re surrounded by clones of ourselves. Like-mindedness can reinforce narratives and generate ill-desired conformity.

Regret and envy

Duke has also highlighted regret as an intense emotion that does little to change what has already happened. However, if regret occurred prior to decision making, such an experience may encourage us to reconsider our choices and future states. She recommends people thinking about how their decisions would affect them 10 minutes, 10 months and 10 years from now.

Warren Buffett and Charlie Munger have also talked about envy and jealousy previously, noting these emotions as dangerous drivers of human behaviour and decision making (perhaps even more so than greed). Like regret, the intensity of such emotions can hamper perspective and the quality of decision making.

By preempting these emotions and anticipating the potential for negative outcomes, we can treat ourselves more compassionately after a poor outcome. It can help us to de-magnify the intense feelings we feel in-the-moment and to take a longer-term perspective. We should avoid reactive emotions that have little impact on our long-term happiness.

Do you want to bet?

Duke stresses that high quality decision-making strategies requires a philosophy centred around truth-seeking. This requires an open mind to new information and a willingness to update our beliefs appropriately (as opposed to altering our interpretation of new facts to fit our system of existing beliefs).

Engagement in complex and open-minded thought is more likely when we are held accountable to a well-informed audience that values accuracy and have incentive to interrogate our logic and choices. Truth-seekers should focus on accuracy over confirmation, hold one another accountable, and be open to diversity of ideas.

Figurative or literal bets can help us to determine the quality of our beliefs. It forces us ask questions such as: ‘How do I know this? Are my sources credible and timely? What are plausible alternatives? What am I missing?’ Making bets, in theory, should force us to engage System 2 thinking to verify and update our beliefs. We are more aware and objective when our choices, risks and beliefs are made explicit.

References

Ariely, D. (2008). Predictably Irrational: The Hidden Forces That Shape Our Decisions.
Duke, A. (2018). Thinking in Bets: Making Smarter Decisions When You Don’t Have All the Facts. More articles available at < https://www.annieduke.com/ >
Duhigg, C. (2012). The Power of Habit.
Kahneman, D. (2011). Thinking, Fast and Slow.
Munger, C. & Clark, D. (2017). Tao of Charlie Munger: A Compilation of Quotes from Berkshire Hathaway’s Vice Chairman on Life, Business, and the Pursuit of Wealth with Commentary by David Clark.

Thinking in Bets – Annie Duke on decision making