Still Flipping a Coin? You Can Make Better Choices in Life

Still Flipping a Coin? You Can Make Better Choices in Life

Some say the greatest regret in life is that we only get to live it once. But it is precisely this uniqueness—this “only once”—that makes life so precious. No one can explore every possible version of their life. From major decisions like choosing a career, a school, or a partner, to smaller ones like what to have for dinner, which flight to take, or which turn to make at an intersection, we are constantly making choices without ever knowing where the road not taken would have led.

We all hope to choose the “better” option, but sometimes even that goal is blurry. We may list the pros and cons of each option on opposite sides of a sheet of paper, yet in the end still rely on a coin toss to make up our minds. In a binary choice, flipping a coin leaves a 50% chance of future regret—which is really no different from intuitively choosing A or B. So aside from that, do we truly have no better way to reduce the probability of regret?

According to Algorithms to Live By, the answer is yes, we do. While no algorithm can make perfect decisions, a scientific and systematic way of thinking can indeed lower the odds of future regret. Over the scale of a human lifetime, sound algorithms can bring regret back down—statistically speaking—to a more normal level. In some cases, that predictable rate is about 37.5%. In other words, even the most scientific approach can only reduce regret by about 12.5 percentage points. But I think that if life could contain even one ten-thousandth less regret, that would still be deeply meaningful.

Algorithms to Live By presents mathematical solutions to a number of classic life problems. Of course, the book does not bury the reader in complicated formulas or lengthy derivations. Instead, it explains the core concepts and then offers conclusions and actionable methods. Even if you know nothing about algorithms, simply reading the cases and takeaways in the book would be highly rewarding. That said, to fully appreciate it, some background in computer science and mathematics—especially probability—does help. If you are at least somewhat familiar with ideas like bubble sort, simulated annealing, or neural networks, you may grasp the essence of the book more deeply.

The book begins with the secretary problem—something that can just as easily be framed as choosing a romantic partner or making any other one-shot selection—and introduces the optimal stopping strategy for one-time decisions. This resembles a question attributed to Socrates in ancient Greece: if you can walk through a wheat field only once, without turning back and without getting a second chance, how do you pick the largest ear of wheat? The largest ear may appear at any point, just as the best candidate may show up at any stage of the process.

The book’s recommended stopping point is 37%. This does not mean choosing the option ranked at the 37th percentile. Rather, assuming you have some sense of the total range, you should use the first 37% of the process to set a benchmark, and after that, choose the next option that exceeds everything you have seen so far. If you plan to interview 100 candidates, do not rush to conclude during the first 37. If you want to spend 100 days looking for the right house, then the first 37 days are best used for seeing as many houses as possible and establishing your standards.

It is important to note that 37% is only the optimal strategy, not a guarantee. No one can predict exactly when the best option will appear, and even the best stopping rule only improves the probability of getting the best outcome by a small margin. But for a life we get to live only once, even a slight improvement still has positive meaning, doesn’t it?

Beyond optimal stopping, every topic in the book is engaging and memorable. Particularly striking discussions include the comparison between exploring new worlds and enjoying familiar rewards: should we keep trying new restaurants, or stick with the one we already believe is best? Another example is hierarchical storage, where the author uses modern computer cache systems as an analogy for more efficient ways to store information and materials. There is also a fascinating section on scheduling, which serves as both an analysis and deep discussion of modern GTD-style time management systems, framed through the lens of computer thread scheduling. Other topics covered in the book include Bayesian prediction, overfitting, Monte Carlo randomness, constraint relaxation, random sampling, swapping and matching problems, and game theory. Every subject is discussed through cases that feel realistic, practical, and highly representative.

The authors also point out the value of algorithms as a guide to real life: “If you’re stuck on a hard problem, remember that heuristics, approximations, and randomized strategies can help you find workable solutions.” In their interviews with computer scientists, one theme came up again and again: sometimes “good enough” truly is good enough. More importantly, understanding complexity can help us choose which problems to take on: if we have some control over the situations we face, then we should choose situations that are tractable.

I also think the book’s discussion of what might be called “computational kindness” is especially illuminating. Much of the time, human interaction can fall into the kind of infinite recursion seen in board games or strategic play: I think I know what you’re thinking; I think you know that I know what you’re thinking; I think I know that you know that I know what you’re thinking... In both computer systems and human relationships, infinite recursion often leads to destructive results. So unless there is a special reason not to, being straightforward in communication is extremely important, because it lowers the other person’s cognitive cost. Even in something as trivial as deciding what to eat tonight, saying “I’m kind of in the mood for hot pot—what about you?” has a much more positive effect, in terms of computational kindness, than saying “Anything is fine.” It reduces the other person’s need to calculate and infer, reduces the game of recursive guessing, and in a world where everyone is busy, that seems undeniably meaningful.

Looking back, I think this book would still be somewhat challenging if read without any prior background knowledge. But as extracurricular reading for students in automation, computer science, or related fields, it is hard to imagine a more fitting choice.