1 Introduction
A long time ago, I watched a movie called "21", and the plot of team cooperation in card counting while playing blackjack left a deep impression on me. However, at that time, I was just watching for entertainment and did not delve into the deeper reasons why card counting could win.
Recently, having some free time, I watched it again. This time, for some reason, I was instead attracted by the hidden beauty of mathematics within it—those decisions that seemed to be made by intuition were actually following the laws of probability behind the scenes.
I suddenly realized that this movie is not just about intellectual battles in casinos, but about the “confrontation between rationality and intuition.” Especially those moments where you clearly feel you “should trust your intuition,” only to be easily overturned by numbers—it is very interesting.
To avoid being a spoiler, I won't introduce the plot in detail, but only mention a few things related to the theme of this article: the core figures in the team do not rely on “memory,” but on sensitivity to probability; almost every decision they make is built on exploiting “information asymmetry.” What the audience is most likely to overlook is: card counting is never about predicting the future, but about using past information to revise future probability judgments.
To understand the principles behind card counting, we might as well start fromthe famous problem proposed by the professor in the filmspeaking of—the Monty Hall Problem(also known as the “Three Doors Problem”):
Its form is much simpler than card counting, but the logic behind it is surprisingly similar:Both are about using new information to revise our judgment of unknown outcomes.
This problem seems on the surface to be just a lighthearted brain teaser, yet it has plunged countless mathematicians and audiences into fierce debate because it reveals a disturbing fact—Our intuition almost instinctively rejects the truth of probability.
2 Analysis of the Monty Hall Problem
1 What is the Monty Hall Problem?
The Monty Hall Problem originally originated from the American TV game show "Let's Make a Deal" in the last century. The host would have the contestant choose one door out of three doors: behind one door is a car (the grand prize); behind the other two doors are goats (empty prizes):
The contestant first selects a door, and then the host (who knows what is behind each door) will, among the remaining two doors,open a door with a goatAt this point, only two doors remain: one is the door you initially chose, and the other is the unopened door.
The host will then ask you: “Do you want to stick with your original choice, or switch to the other door?” This is the Monty Hall Problem:
The problem seems simple, but almost everyone who hears it for the first time willintuitively think that switching or not makes no difference—after all, with two doors, isn't the probability of winning 50% each?
However,the mathematical answer is completely the opposite: if you choose to switch, the probability of winning is 2/3; if you stick to your original choice, the probability of winning is only 1/3。
It is precisely this counter-intuitive result that has sparked countless discussions and debates over the decades: from television audiences to mathematics professors, from psychologists to probabilists, all have tried to explain why our intuition is so stubbornly “wrong” on this problem.
To understand “why switching doors makes it easier to win,” we can look at this problem from two different perspectives, ranging from complex to simple.
2 Why is the probability of winning higher after switching doors?
2.1 Standard Mathematical Solution: Calculating with Conditional Probability
Let's first solve this problem using the most “formal” mathematical way, which is the “variable change” thinking mentioned by the professor in the movie.
Suppose the three doors are A, B, and C, respectively, with a car behind only one door, and you choose A first. The host knows the answer, and he opens another door (say C), which must have a goat behind it. Now, he gives you the opportunity—do you want to switch to B?
To judge whether switching doors is worth it, we are actually comparing the probability of two scenarios:Scenario 1: Sticking to the original door (A) → What is the probability of winning the car?Scenario 2: Switching doors (B) → What is the probability of winning the car?
Let's first look at Scenario 1: You choose A at the beginning, and at this time, the probability of the car being behind A is only 1/3This will not change; no matter what the host does next, your first choice has already set the probability.
Now look at Scenario 2: If there is no car behind door A, then the car must be behind one of the two doors B and C (the sum of the probability of the car being behind the two doors is 1/3+1/3=2/3). Since the door opened by the host must be a goat, the probability of the car for the remaining unopened door is “concentrated” onto it. Thus, the probability of winning the car by switching doors = 2/3。
This logic is actually “conservation of probability”—the initial 1/3 of “good luck” remains on the original door, while the 2/3 of “bad luck,” after the host helps you filter out the wrong option, is entirely concentrated on the other door. Therefore,switching doors is equivalent to betting all of that 2/3 probability on the remaining unopened door, so the chance of winning is naturally higher.
To be more intuitive, we can also list all the scenarios to take a look:
| First choice | Car's location | Door opened by the host | Result of switching doors | Win or not |
|---|---|---|---|---|
| A | A | C | Switch → B | 输 |
| A | B | C | Switch → B | 赢 |
| A | C | B | Switch → C | 赢 |
You will find that among the three equally probable scenarios, switching doors wins 2 times (2/3), while not switching only wins 1 time (1/3). If you understand this logic from here, congratulations—you have mastered one of the most important ideas in probability theory:After new information appears, the probability can be revised.
2.2 The Simplest Heuristic: Originally choosing one door is now equivalent to choosing two doors
Actually, this problem can be understood even more simply—“Switching doors” is actually equivalent to the host giving you a new choice: choose one door, or choose two doors.
Think about it: if you don't switch doors, it is equivalent to always sticking to the initial one door, and the probability of winning is always 1/3But if you choose to switch doors, since the host has already helped you eliminate the door with the goat among the other two doors, the remaining door inherits the “combined” winning probability of those two doors.
In other words, although on the surface you are just switching from one door to another, in essence, it is equivalent to expanding your choice from “one door” to “two doors.” Thus—not switching doors: you are still betting on that 1/3 chance; switching doors: you are equivalent to choosing “the entirety of the two doors,” and the probability of winning the car becomes 2/3。
From this perspective, the so-called “switching doors” is not because your luck got better, but ratheryou used the new information to redistribute that original 2/3 probability of error to your sideThis is the simplest and most fundamental way of thinking about the Monty Hall problem:Not probability magic, but rather information redistribution.
Note: Many people like to use Bayes' theorem to prove this problem, which is actually correct, but—understanding how information changes probability is the key; the formula only helps you confirm the answer.
3 Analysis of the Card Counting Logic in the Movie "21"
1 Overview of Card Counting Principles in the Movie
In “21”, the card counting team's strategy is a textbook application of probability. On the surface, they are just "better at calculating" than others, but the real core lies in—they are dynamically using information to update their judgments。
At the casino's blackjack table, the dealer uses several decks of cards mixed together. As the cards are dealt one by one, the composition of the remaining deck gradually becomes unbalanced: many small cards dealt early on means the proportion of remaining high cards increases, and vice versa. “Card counting” is tracking this imbalance and adjusting betting strategies accordingly.
They used a method called “Hi-Lo Count” Simply put, it isby observing the cards that have already appeared, to infer the ratio of high cards to low cards in the remaining deckThe higher the count, the more high cards are left, and the player's win rate rises accordingly.
Through this information update, they can raise bets during advantageous periods and hold back during disadvantageous periods, ensuring that capital is only invested at the most certain moments. In other words, what they do is not prediction, but ratherdynamically revising their judgment of future outcomes based on the continuously acquired new information—this is exactly the same as the core thinking of the Monty Hall problem: information update → probability redistribution.
2 The Mathematical Logic of Card Counting: Dynamic Revision of Probability
The core of card counting is not “remembering what each card is,” but ratherafter each card is revealed, using this new information to revise the probability judgment of the remaining gameThis is consistent with the essence of what we learned using the Monty Hall problem in Chapter 2: new information appears → we should redistribute the probability of future events → adjust strategies based on the new probability.
Below, we break this process down into several more specific steps—explaining both “what it is” and “how to use it.”
Counting: compressing complex deck information into a “signal”
When one or more decks of cards are mixed together in a casino, the initial distribution of each card in the entire deck is known (such as how many 10-value cards and how many small cards there are). As cards are dealt, the composition of the remaining deck changes.
Card counting is “compressing” the information of the cards that have already appeared into a simple numerical value (calledrunning count), and the common practice is Hi-Lo counting:
| Card face | Count value | Description |
|---|---|---|
| 2 ~ 6 | +1 | The more small cards appear, the more favorable the remaining deck is to the player |
| 7 ~ 9 | 0 | Neutral, no count |
| 10, J, Q, K, A | −1 | The more high cards that appear, the more unfavorable the remaining deck is to the player |
By adding up the dealt cards one by one, a running count is obtained (which can be positive or negative). This number itself is a compressed description of the “bias of the remaining deck”: the higher the count, the higher the proportion of remaining high cards, and the better the player's chances of winning.
What is truly useful is not the running count, but the “true count”
The running count is affected by how many decks of cards are currently remaining—under one deck and six decks, the same running count has different meanings. Therefore, the running count must be divided bythe estimated number of remaining decks, to obtain the True Count：
To give an intuitive example: suppose there are 6 decks of cards (312 cards), and the current running count is +6, but there are still about 3 decks left undealt on the table, then the true count ≈ +6 / 3 = +2. This +2 indicates that on a per-deck level, the remaining deck has a “+2 bias signal” for the player.
Map the true count to “player advantage”, and then map it to bet size
Empirically (with slight variations across different literature),each +1 of the true count roughly corresponds to a small percentage unit increase in the player's expected valueA common empirical value is: for every 1-point increase in the true count, the player's expected value increases by about 0.5%(this value is an empirical approximation, and actually depends on the rules and specific implementation).
Continuing with the above example, if the true count = +2, the player's expected value against the dealer might increase by about 1% (2 × 0.5%). At this point, the “big player” of the card counting team will scale up their bet accordingly: betting small when the advantage is zero or negative, and betting big when the advantage is positive—in the long run, this selective bet-raising strategy will yield positive returns.
Note: Card counting is not about making you win every hand, but about letting you scale up your bets in favorable situations and pull back in unfavorable ones, thereby turning the long-term compound expected value positive.
Risk Management (Why it is not a “sure win”)
Even if a true count of +2 indicates a 1% increase in expected value, this is only the expected value (long-term average), while individual rounds still have high variance. Therefore, the team must also consider:
- Bet spread: What chip multiplier to use when switching between high and low true counts?
- Volatility and risk of ruin: A positive expectation does not mean one won't go bankrupt in the short term; proper bankroll management (such as inspiration from the Kelly Criterion) is needed to determine the betting ratio.
- Risk of casino intervention: Large bets, frequent entry and exit, or suspicious behavior will attract casino attention (we will discuss this in detail in the reality section later).
The Relationship Between Card Counting and the Monty Hall Problem
Connecting the above steps, you will find that: Monty Hall Problem:
- The host opens a door to provide new information → we redistribute probabilities accordingly → and finally make the decision to “switch doors” or “not switch doors”;
- Card counting (Blackjack): Each card revealed is a disclosure of information → running count + true count compresses the amount of information into an actionable signal → we adjust our bets (decision) based on the true count.
The core thinking shared by both isBayesian information updating: It is not about relying on intuition of “what it should feel like”, but systematically revising probabilities when new information arrives, thereby shifting the long-term expectation to the favorable side.
3 Team Strategy and Psychological Warfare in Card Counting
In "21", card counting is not a solo effort, but a well-designed team action. Through clear division of labor and tacit cooperation, team members translate the application of mathematical probability into actual operations in the casino.
The team is usually divided into two core roles:
- Spotter: They lurk at multiple tables, observing the game with low bets. On the surface, they seem to play casually, but in reality, they are silently counting cards in their minds, tracking the ratio of high cards to low cards in the remaining deck.
- Big Player: When a spotter determines that a table has a clear advantage, they will summon the big player to the table using a secret signal. The big player appears to be a wealthy passing gambler, but in reality, they are betting on a table where the probabilities have already been calculated, usually placing large chips.
Cooperation Process:
Detecting Advantageous Tables: Spotters place low bets at various tables while tracking the Hi-Lo count in real-time, waiting for the deck to become favorable to the player.
Passing Signals: Once the count indicates a high proportion of remaining high cards, the spotter notifies the big player using a predetermined signal.
Large Betting: The big player joins the table and bets, taking full advantage of the game's favorable situation.
Retreating: When the count drops back or the advantage disappears, the spotter sends a “retreat” signal, and the big player leaves the table in time, ensuring that capital is only invested during the most confident phases.
Psychological Warfare and Casino Gaming:
The team relies not only on mathematical judgment but must also master psychological strategies. Casinos closely monitor player behavior: unusual betting, emotional reactions, or the rhythm of raising bets can all attract attention.
Card counting teams must hide their advantage: occasionally losing money on purpose, pretending to be emotional, or changing their betting tempo to maintain the appearance of an ordinary player. At the same time, they also need to observe the reactions of dealers, other players, and casino security, adjusting their behavioral strategies at any time.
The success of a card counting team relies not only on mathematical calculations, but even more onthe understanding and control of human natureThey use probability judgments to decide their betting, but use psychological strategies to control risk, making the entire gamble adual game of rationality and human nature, mathematics and psychology。
4 Digression: How Real Casinos Counter Probability Players
In the movie, the card counting team seems invincible, but the environment of a real casino is far more complex than in the movie. In the movie, the casino specifically sends people tokeep an eye on suspicious players, observing their betting patterns and behavior, which is actually a common countermeasure in real casinos.
To summarize the basic countermeasures of casinos:
- Frequent Shuffling and Mixing Multiple Decks
Real casinos usually use multiple decks mixed together for dealing and shuffle them within a short period of time. In this way, the composition of the remaining deck is quickly disrupted, making the Hi-Lo count lose its reference value. Even if players are highly proficient in probability calculations, they cannot accurately judge the advantage of the next hand.
- Betting Limits and Mid-Game Entry Regulation
Casinos not only limit the maximum betting amount at each table, but also strictly control when players can join. If someone joins mid-game and attempts to make large bets, the casino will be highly alert and may even ban them from continuing to bet. This directly counters the strategy in the movie where “Big Players join the table to place large bets at any time,” reducing the operational space for card counting teams.
- Dedicated Surveillance and Technological Means
The “casinos catching card counters” depicted in the movie also exists in reality: casinos will deploy experienced dealers, supervisors, and security guards to specifically observe players' betting patterns, behaviors, and even facial expressions. Modern casinos are even equipped with cameras and behavioral analysis systems to analyze player behavior in real-time and take immediate action upon detecting anomalies.
- Psychological and Strategic Interference
To disrupt the rhythm of probability players, casinos may manipulate the dealing tempo, dealer changes, and table rotations. Frequently changing tables or shuffling makes it difficult for players to establish a continuous advantage, while also exerting psychological pressure.
The reason why the card counting team in the movie could win is based on the idealized assumption that the casino uses fixed decks, shuffles infrequently, and that teamwork and signaling can be executed perfectly. However, based on the casino countermeasures mentioned above, in reality, these conditions are very difficult to fully meet:
- Multiple decks and frequent shuffling reduce the accuracy of the Hi-Lo count and shorten the window of advantage.
- Dedicated surveillance, betting limits, and psychological interference make it difficult to implement teamwork covertly.
- High-stakes betting easily attracts attention, and the returns are disproportionate to the risks.
- The rule prohibiting mid-game entry directly disrupts the operational model of “Big Players joining the table to bet at any time.”
In other words,Real casinos compress the probability advantage to a minimum, making any attempt to exploit probability difficult to achieve long-term stable profitability.。
5 Afterword
The reason why the movie "21" is so memorable is not just because that group of students beat the casino, but because it reveals a more universal reality:Human intuition is not always suited for handling probability problems。
In the movie, the protagonists trained themselves to break free from “gut-feeling” judgments and instead relied on mathematical models to bet—this was precisely the key to their victory over the casino. But at the same time, their success also seemed extremely “inhuman”: no passionate impulses, no intuitive high-stakes gambling, everything was built on rational calculation.
In real life, it is almost impossible for us to act like them. Even if we know that “card counting” is a rational probability strategy, we will still be swayed by emotions, wishful thinking, or short-term wins and losses. We want to “win,” but we often dislike the rational rules that make “winning” boring.
This is also where the true paradox of probability thinking lies:Mastering it means giving up a certain “human nature.”
In other words, mathematics teaches us to be calm, while human nature makes us fluctuate. Many choices in casinos, the stock market, and life are actually like a game of “21” (Blackjack)—you cannot control what cards you are dealt, but you can choose how to bet. Therefore, card counting is only a surface phenomenon.
The deeper revelation is:Can we remain rational and update our judgments when facing uncertainty? Can we, like the card counting team, constantly revise our strategies based on new information, rather than indulging in emotional ups and downs?
Perhaps, "21" is not about how to beat the casino, but rather—how to coexist with probability in an uncertain world.