Thinking in Bets is one of the few books I’ve read for the second time. I personally believe this book is helpful in enhancing our thinking and decision-making abilities.
The author, Annie, has a unique life story. She was originally a doctoral student in psychology at Columbia University. A month before her doctoral defense, she gave up her degree to become a professional Texas Hold’em poker player. In this book, Annie uses Texas Hold’em as a starting point, combined with her expertise in cognitive psychology, to comprehensively yet accessibly introduce the thinking and decision-making methods she learned and generalized from playing Texas Hold’em.
The Chess Game of Computational Victory
Experts in chess-like games excel in calculation. It’s rumored that chess grandmaster Kasparov could foresee up to 10 moves ahead, and Ke Jie, in a particular game, trapped his opponent in a scenario he had set up 30 moves earlier. For any board or card game, we can approximately quantify their complexity (expressed using a data structure called a game tree). Take Tic-Tac-Toe, for example, with its 9 positions each capable of being X, O, or blank. Tic-Tac-Toe’s state magnitude is around 10⁵. Its game tree complexity is within human comprehension, allowing some experts to never lose.
Gomoku, more complex than Tic-Tac-Toe, was theoretically proven in 2001 to be a guaranteed win for the first player (black pieces), but due to its vast state space (complexity 10⁷⁰), even masters can only win using specific strategies in certain board positions.
For chess, with a complexity of 10¹²³, and Go, with 10³⁶⁰, their optimal solution search spaces are virtually infinite for human cognitive capacity.
In 1997, Deep Blue defeated chess grandmaster Kasparov, and AlphaGo successively beat Lee Sedol and Ke Jie in 2016/2017, followed by Alpha Zero’s undefeated record against AlphaGo, highlighting how superficial human exploration of these board games still is. In the face of Alpha Zero’s dimensional reduction attack, there’s no essential difference between world champion Ke Jie and someone like me, who knows nothing about Go. Regardless of their complexity, chess games are theoretically deterministic. If we had the capability to traverse the entire decision tree, the outcome of a game would be predetermined even before it starts.
The Uncertainty of Card Games
Card games (like Mahjong and Texas Hold’em) differ from chess-like games, as the former are non-deterministic and lack complete information.
Non-determinism means that participants do not have full control over the game. For instance, the initial deal in card games is random, and each round presents different cards to each player. With good luck, one might win right from the start, while bad luck might prompt an immediate surrender. Due to this randomness, a novice can win against a world champion in a single game by sheer “hand luck”.
Lack of complete information means that we don’t just lack knowledge about the future of the current game (which is obvious) but also do not fully know past information. In chess-like games, we can see the current state of the board and the history of moves. In card games, however, we neither know the cards drawn by our opponents nor the remaining cards in the deck.
These two factors prevent us from determining where the current card game stands in the decision tree, making it unclear what our next move should be. From the perspective of a participant with only partial information, the game exists in a state of quantum superposition.
Life is Poker, Not Chess
It’s often said that life is like a game of chess. However, I agree more with Annie’s view: life is more like a card game rather than a chess game. Compared to chess, card games are more complex and filled with uncertainties. Of course, life is even more complex than card games.
Throughout our lives, we continually make decisions, whether consciously or unconsciously (like reflex actions), based on partial information.
In the real world, “hand luck” or luck plays a significant role. Some people, due to good luck, achieve positive outcomes even with poor decisions, while others experience the opposite.
We often judge the quality of a decision based on the outcome, using this as a basis to improve future decision-making quality. This is known as decision review or post-mortem.
However, the author points out in the book that we cannot solely judge the quality of our decisions based on their outcomes. The author mentions “hindsight bias,” where people often attribute the quality of outcomes to their prior decisions. If the outcome is good, it is assumed that it’s due to a good decision. Yet, a good outcome doesn’t necessarily result from a good decision, and vice versa.
Many factors determine outcomes, and personal decision-making is just a small part of these. External factors such as trends of the times and luck significantly impact life. Consider this joke:
Three people take an elevator from the first to the tenth floor. One runs in place, another does push-ups, and the third headbutts the wall. They all reach the tenth floor. When asked how they got there, they each answer: “Ran up,” “Did push-ups,” “Headbutted my way up.”
Those who bought houses twenty years ago or entered the internet industry ten years ago have shared their “success stories” with many.
We perceive the rules of things through a process of “decision-making — observation.” If there’s no necessary causal relationship between decisions and observations, then observing outcomes does not help improve the quality of our decisions.
The situation is not as hopeless as it may seem. A “causal” relationship still exists between decision-making and observation, though it’s not as straightforward as we might think. The author suggests several methods in the book:
- Change perceptions.
- Establish positive cognitive feedback.
- Find like-minded individuals.
The first step in improving decision-making is to acknowledge that we cannot fully understand the world and to use shades of grey, rather than black-and-white thinking, to describe causal relationships. When assessing the impact of a decision on an outcome, we use probabilities instead of binary variables. For example, if I choose Plan A, the success probability of Event X might be 80%, whereas with Plan B, it might be 90%.
There are no absolutes in this world. All certainties are time-bound. A farmer feeds the turkey daily only because Thanksgiving hasn’t arrived yet. Until 2022, the worldview of post-2000 generations was that stocks always go up. We believe a day on Earth is 24 hours only because, over the span of human civilization (assuming 10,000 years), the Earth’s rotation has changed by only 1.5 seconds per year.
When we describe relationships in terms of probabilities, we leave room for chance within certainty.
Once we adopt this new way of thinking, we can establish positive cognitive feedback. We first make decisions (bets) based on beliefs formed from our understanding, then collect outcome samples. By analyzing a large number of decisions and their outcomes, we can reduce the uncertainty brought by luck. We improve our understanding and optimize our beliefs by using the refined information.
Positive feedback loop
Establishing positive cognitive feedback is a challenging and error-prone process. Therefore, it’s best to find like-minded individuals to improve together. This approach has several benefits:
- Group members can encourage each other and help overcome difficulties. Moreover, peer pressure can also enhance individual willpower.
- Collaborating can eliminate subjective judgments of others.
- Pooling ideas can generate more methods to help establish positive feedback.
The Limitations of Probabilistic Thinking
Probabilistic thinking has its limitations. For extremely rare events, we cannot estimate their probability, making it impossible to make appropriate decisions. For the impact and prevention of such rare events, Taleb’s “The Black Swan” offers a detailed discussion. Those interested might want to read my review of that book.