Improve your decision-making, frameworks for learning, backcasting, and more | Annie Duke (#60 rebroadcast)

Sep 2, 2024 Episode Page ↗
Overview

Former World Series of Poker champion and decision strategist Annie Duke explains how poker models real-world decision-making under incomplete information and luck. She discusses frameworks for evaluating decisions, learning from all outcomes, and the importance of 'backcasting' for long-term planning.

At a Glance
17 Insights
2h 32m Duration
18 Topics
9 Concepts

Deep Dive Analysis

Chess vs. Poker: Metaphors for Real-World Decision-Making

Human Tendency to Avoid Probabilistic Thinking

Annie Duke's Journey into Professional Poker

Variable Reinforcement Schedules and Poker's Appeal

Distinguishing Luck and Skill in Games and Life

Texas Hold'em: Basic Rules and Strategic Elements

Reading Opponent Behavior and Game Dynamics

The Value of Quitting Fast in Decision-Making

Limit vs. No-Limit Poker and Game Evolution

The Advent of Analytics in Modern Poker

The 'Resulting' Heuristic and Decision Evaluation

Societal Consequences of Avoiding Bad Outcomes

Poker as a Model System for Life Decisions

Leadership Decision-Making and Analytics Adoption

Learning from Unexpected Outcomes (Y2K, D-Day)

Framework for Improving Decision-Making and Learning

The Capacity for Self-Deception in Decision Analysis

Benefits of Backcasting and Pre-Mortems

Chess vs. Poker as Decision Models

Chess, a 'perfect information' game with calculable moves and minimal luck, is a poor metaphor for real-world decision-making. Poker, conversely, involves incomplete information and significant luck, making it a more accurate model for life's stochastic nature.

Variable Reinforcement Schedule

A psychological principle where rewards are given on an unpredictable average, leading to highly persistent behavior that is difficult to extinguish. This explains the addictive nature of activities like slot machines and poker, as players feel 'due' for a win.

Skill vs. Luck Test

To determine if an activity involves skill, one can ask: 'Can you lose on purpose?' If yes, it implies skill. The influence of luck becomes more apparent as the skill gap between competitors narrows.

Resulting Heuristic

A cognitive shortcut where people judge the quality of a decision based solely on its outcome. A good outcome is assumed to stem from a good decision, and a bad outcome from a bad decision, often ignoring the underlying process or the role of luck.

Decision Matrix (Process vs. Outcome)

A 2x2 framework for evaluating decisions based on both the quality of the decision-making process (good/bad) and the quality of the outcome (good/bad). The episode highlights that people often focus only on outcomes, particularly bad ones, and neglect learning from all four quadrants.

Transparency Problem in Decisions

Decisions that are standard, expected, or made by consensus are often forgiven for bad outcomes, as luck is readily accepted as an explanation. Conversely, unusual or innovative decisions leading to bad outcomes are heavily criticized, even if the process was sound, due to a lack of perceived transparency in the decision-making.

Four Levels of Thinking (Decision-Making)

A progression in how one evaluates decisions, moving from superficial analysis to deep, symmetrical scrutiny. Level 1 (Default): Examine losses through luck, assume wins are due to skill, without deep analysis. Level 2 (Improved): Examine losses through both luck and skill (taking responsibility), but still assume wins are purely skill. Level 3 (Comprehensive): Examine all four quadrants (good/bad decision, good/bad outcome) with equal scrutiny. Level 4 (Elite): Go beyond concordance in wins; even if a good decision led to a good outcome, explore if an *even better* decision could have been made, risking turning a perceived win into a learning opportunity for long-term improvement.

Backcasting

A strategic planning method where one imagines a future goal has already been achieved and then looks backward to determine the steps and factors that led to that success. This contrasts with forecasting, which plans forward from the present.

Pre-mortem

A planning technique where one imagines a future project or goal has failed and then works backward to identify all the potential reasons for that failure. This helps uncover risks and contingencies that might be missed in traditional forward-looking planning.

?
Why is poker a better metaphor for real-world decision-making than chess?

Poker is a better metaphor because, unlike chess where all information is known, poker involves incomplete information and a strong influence of luck, mirroring the stochastic nature of real-life decisions in medicine, business, and daily life.

?
Why are humans not naturally wired to think probabilistically?

Humans tend to struggle with probabilistic thinking because our gut reactions are often deterministic, and it's difficult to constantly live in a 'counterfactual world' where multiple outcomes are possible, even for the same action.

?
What psychological mechanism makes games like poker and slot machines so engaging?

These games operate on a variable ratio reinforcement schedule, where rewards are unpredictable but frequent enough on average, making the behavior highly persistent and difficult to extinguish, as players always feel 'due' for a win.

?
How can one distinguish between skill and luck in an activity?

A simple test is to ask: 'Can you lose on purpose?' If you can intentionally lose while following the rules, the activity involves skill. The influence of luck becomes more pronounced as the skill gap between participants narrows.

?
What is 'resulting' and how does it affect decision evaluation?

Resulting is a cognitive bias where people judge the quality of a decision based solely on its outcome. A good outcome is seen as proof of a good decision, and a bad outcome as proof of a bad decision, often ignoring the decision-making process itself.

?
How does the fear of negative outcomes impact decision-making in organizations and individuals?

The fear of being seen as an 'idiot' for a bad outcome, especially from an unconventional decision, drives people towards status quo decisions, seeking consensus, or delaying action, stifling innovation and progress.

?
Why is it important to analyze both good and bad outcomes, even when a decision seems successful?

Analyzing good outcomes helps identify overlooked risks or opportunities, and ensures that success wasn't just due to luck or an underestimated market. It allows for deeper learning and refinement of decision processes, preventing self-deception.

?
What is 'backcasting' and how can it improve long-term planning?

Backcasting involves imagining a future goal has already been achieved and then working backward to identify the steps and factors that led to that success. This approach helps uncover crucial elements, including potential interventions of luck, that might be missed in forward-looking forecasting.

?
How does a 'pre-mortem' enhance planning?

A pre-mortem involves imagining a future project has failed and then identifying all the potential reasons for that failure *before* it happens. This helps in proactively addressing risks, developing hedges, and preparing responses in a calm state, rather than reactively when problems arise.

1. Master Decision Learning Levels

Progress through four levels of examining outcomes: from blaming luck for losses (Level 1), to taking responsibility for losses (Level 2), to analyzing all four decision quadrants (Level 3), and finally, to critically examining even good decisions with good outcomes for potential improvements (Level 4).

2. Utilize Backcasting for Goals

When setting a goal, imagine it’s already achieved (e.g., ‘it’s a year and a day and I achieved my goal’). Then, look backward and ask, ‘How did I actually achieve this?’ This helps identify the steps and factors that led to success.

3. Conduct a Pre-mortem for Failure

For any goal, envision that you have failed to achieve it (e.g., ‘it’s a year and a day and I failed’). Then, ask ‘Why did I fail?’ This process helps identify potential pitfalls and allows you to plan for them in advance.

4. Incorporate Luck into Planning

When using backcasting or pre-mortems, allow for the influence of luck in your analysis. This helps you explore ways to increase chances of positive luck, decrease negative luck, or hedge against unavoidable bad luck scenarios.

5. Plan for Bad Luck in Advance

By anticipating potential negative ’luck’ scenarios through pre-mortems, you can make decisions and plans in a calmer state of mind. This proactive approach prevents reactive, ‘pants on fire’ decision-making when adverse events occur.

6. Cultivate Probabilistic Thinking

Recognize that outcomes are not deterministic; many things can occur with varying likelihoods. Understanding this probabilistic nature of the world is crucial for making better decisions.

7. Catch Decision Errors Faster

Actively strive to identify and correct errors in your decision-making process more quickly. Even small, frequent improvements in catching errors can significantly enhance overall decision quality.

8. Practice ‘Quit Fast’ Strategy

Embrace the philosophy of quitting quickly when something isn’t working or isn’t a good fit. This minimizes opportunity costs and allows you to reallocate resources to more promising endeavors.

9. Test for Skill vs. Luck

To determine if an activity primarily involves skill, ask yourself: ‘Can I lose on purpose?’ If you can, it indicates a significant skill component, as opposed to pure luck.

10. Explore All Outcome Directions

When evaluating outcomes, maintain equanimity regardless of whether the result was good or bad. Explore ‘could I have done worse?’ and ‘could I have done better?’ to gain comprehensive insights.

11. Think Orthogonally About Outcomes

Consider that outcomes might be due to reasons unrelated to your decision process, or that you won/lost for reasons you didn’t predict. This helps avoid misattributing success or failure solely to your actions.

12. Learn from Unexpected Wins

Don’t just celebrate unexpectedly good outcomes; actively investigate why they occurred. This helps identify new strategies or factors that contributed to success, which can then be intentionally replicated.

13. Forecast & Analyze Performance

Before an activity, forecast your expected performance (e.g., percentage of bullseyes). If actual performance deviates significantly (either better or worse), actively investigate the reasons to understand what changed.

14. Challenge Your Own Ideas

Actively seek out challenges to your decisions and beliefs, both from others and from within yourself. This helps prevent self-deception and ensures a more robust decision-making process.

15. Seek Pure, Unbiased Feedback

Actively solicit feedback from others, ensuring you communicate in a way that encourages honesty and doesn’t bias their responses. Being genuinely ‘hungry for the information’ is the crucial first step.

16. Know When to Decide Fast

Understand the broader decision-making framework to discern which decisions require careful, slow consideration versus those that can be made quickly. This prevents analysis paralysis on minor choices.

17. Embrace Relative Certainty

For bigger decisions, once you’ve identified options and one appears better than others, stop striving for absolute certainty. Instead, focus on the relative superiority of your chosen option, recognizing inherent unknowns.

If I play chess against you a few times, it becomes very clear to me why I'm losing. I'm losing because you're better than I am. I have nothing else to protect my ego.

Annie Duke

We do not spend enough time encouraging people to quit.

Annie Duke

The further you get from shore, the deeper the water.

Peter Attia

If I know what the quality of the outcome is, I can work backwards from that to the quality of the decision.

Annie Duke

When we land on the beaches of Normandy, and it's a pivot point in the war, and we win, people aren't looking as hard to try to figure out to really explore the counterfactuals.

Annie Duke

The longer the time horizon, the more capacity you have, the more leeway the world is giving you to fool yourself.

Annie Duke

This framework shows you where the curves are. And that frees you up to go faster in the straightaways.

Peter Attia

Backcasting for Life Goals (Centenarian Olympics)

Peter Attia
  1. Imagine yourself on the day before your last breath, having achieved your long-term physical goals (e.g., at age 100).
  2. Define specific, pragmatic physical tasks you want to be able to perform at that age (e.g., 30-pound goblet squat, standing from the floor without support).
  3. Work backward: Determine what physical capabilities you would need at earlier ages (e.g., 90, 80, 60) to maintain the trajectory towards your centenarian goals.
  4. Identify current actions and training necessary to build and sustain those capabilities.
  5. Continuously adjust your present actions based on this backward-looking perspective.

Pre-Mortem for Strategic Planning

Annie Duke
  1. Define a specific goal to be achieved within a set timeframe (e.g., 'I'd like to achieve X in a year').
  2. Imagine it's the day after the deadline, and the goal has *failed*.
  3. Ask: 'Why did I fail?' Brainstorm all possible reasons for failure, including both controllable factors (e.g., insufficient effort, poor decisions) and uncontrollable factors (e.g., bad luck, unforeseen events).
  4. For identified 'luck' factors, ask: 'Can I decrease the chances that this bad luck happens?' and 'Are there any hedges available to me?'
  5. Plan in advance how to respond if those bad luck scenarios occur, making decisions in a calmer state of mind.
  6. Use these insights to refine the strategic plan, making it more robust against potential failures and unexpected interventions.
80-ish percent
Win probability for aces vs. fives in poker Player with aces is 80% likely to win against fives.
20%
Win probability for fives vs. aces in poker Player with fives is 20% likely to win against aces.
19 correct
Expected correct answers in a 95% confidence interval game (out of 20 questions) The goal is to have 19 out of 20 answers bracketed correctly.
12 right and eight wrong
Peter Attia's score on the 95% confidence interval game (out of 20 questions) This score is equivalent to a 60% confidence interval.
less than 2%
Interception rate on Pete Carroll's Super Bowl pass play The probability of an interception on that specific play.
about 56% of the time
Win rate for a very good limit poker player Over an eight-hour session, the player wins 56% of the time.
44% of the time
Loss rate for opponent of a very good limit poker player Over an eight-hour session, the opponent wins 44% of the time.
1500 hours
Hours of play for a limit poker player to be 'in the money' A player with a 56% session win rate will approach 100% certainty of being profitable over 1500 hours of play.
80% or so
Percentage of ear infections that are viral Used as an example of a situation where doctors might still prescribe antibiotics to avoid blame, despite the high likelihood of a viral cause.