The paradox
Most strategy advice says to find the best option and choose it. In some games that advice is exactly wrong. If an opponent can learn what you will do, the “best” option becomes a weakness, because they can counter it. In those games the strongest plan may be to choose at random, on purpose, according to a carefully chosen mix.
That is the paradox: deliberately leaving the decision to chance can be a sign of good strategy, not a lack of it. Players use this idea, often called a mixed strategy, to stay unpredictable when predictability would be punished.
The simplest case: rock-paper-scissors
Rock-paper-scissors shows the idea cleanly. No option is better than the others; each beats one and loses to another. If you always play rock, a thoughtful opponent will soon play paper. If you favour rock 50% of the time, they can exploit that tilt.
Now suppose you pick each option with probability one third, independently each round. Whatever your opponent does, whether always the same thing or a clever pattern, the chance that you win equals the chance that you lose, and the long-run average comes out even. You cannot be beaten on average, because there is nothing to exploit.
Notice what this strategy does and does not do. It protects you from being exploited, but it does not win by itself against a weak opponent. If the other player is predictable, you could do better by exploiting their pattern, but that carries the risk of being exploited in turn if they are only pretending to be predictable.
Why it works: removing the pattern
Randomisation works by removing information. An opponent can only counter what they can predict. A truly random choice, made after you have decided the probabilities, contains no pattern to find. Even you do not know the choice until you make it.
This matters for a subtle reason. People are poor random number generators. When asked to “act randomly”, they tend to switch too often, avoid repeats and drift into habits. Opponents who pay attention can use those habits. For this reason, serious players who want to mix their play often rely on something external and arbitrary, such as the last digits of a clock or a die roll, to choose between options according to set proportions.
Where mixed play appears
The idea applies wherever one side benefits from guessing the other’s action:
- Bluffing in card games. If you only ever bet with strong hands, opponents learn to fold when you bet. If you sometimes bet with weak ones, they cannot be sure. In simplified models, the right proportion of bluffs depends on how much you bet relative to the pot. As a rough illustration, with a bet the size of the pot, about one in three bets being a bluff leaves an opponent indifferent about calling. Real games are far more complicated, so treat that as a concept rather than a rule. The wider ideas of poker odds and bet sizing show where such calculations come from.
- Choosing between attack and defence. In some strategy and fighting games, always opening with the same move invites a counter. Mixing up timing and choices keeps the opponent guessing, which relates to the study of opponent reading.
- Sports-style situations. The decision of where to aim or which way to move, when the other side is also guessing, has the same structure.
- Guessing and hidden-information games. Anywhere the opponent is trying to predict you, the ability to be unpredictable has value.
How to choose the mix
A mixed strategy is not “do whatever”. The proportions matter, and they come from the payoffs. A rough guide:
- List your options and the opponent’s counters. Identify which option beats which.
- Find the proportions that leave the opponent with no best reply. In a symmetrical game like rock-paper-scissors, equal shares do this. In an unbalanced game, the right shares are uneven, because the cost of being countered differs between options.
- Randomise genuinely. Use a tool rather than your sense of “random”.
- Adjust to evidence. If you see the opponent deviating from balanced play, you may shift your mix to exploit them, bearing in mind the risk of being exploited back.
In unbalanced games the right shares are usually uneven, and working them out takes some algebra, so it is worth checking a source on game theory before relying on a particular figure.
When chance is not a strategy
Randomising is not universally useful. It does not help when:
- Your best option is clearly best. If an option is better regardless of what the opponent does, take it every time. Random variation would only cost you.
- The opponent cannot adapt. Against a computer routine or a fixed set of probabilities in a puzzle, there is no one to confuse. Here, careful puzzle solving usually beats randomness.
- The game is a pure chance game. If outcomes are random by design, adding more randomness changes nothing.
- You are using it to avoid thinking. A mixed strategy is a calculated tool, not a shrug.
A note on risk and expectations
Mixing protects you against exploitation, not against bad luck. In any single round or short run, you can still lose, sometimes repeatedly, even with a perfectly balanced mix. Over long runs the average tends to match the theory, but short-term results will vary, as explained by the ideas behind random events and outcome distributions.
If real money is on the line, remember that unpredictable play does not turn a game with unfavourable odds into a favourable one. It only makes you harder to read in games where reading is possible.
The takeaway
In games of prediction, the safest way to avoid being exploited is to make your choices genuinely unpredictable, in proportions that follow from the payoffs. It sounds odd to call randomness a plan, but it is precisely a plan: decide the odds deliberately, let chance pick the move, and remove the pattern your opponent was looking for.