Tag: probability

Discussions of probability theory and its application to real-world decision making, including expected value, uncertainty, calibration, and the role of statistical thinking in markets.

  • Variance Is the Price of Edge

    In the previous essay I argued that good decisions can still produce bad results. In probabilistic systems, outcomes contain noise, and even disciplined decisions can lead to losses.

    That reality stems from a deeper truth.

    The same uncertainty that produces losses is also what creates opportunity.

    Howard Marks often reminds investors that greater risk should demand greater potential return.

    The idea is simple. If an investment carries more uncertainty, the payoff must be large enough to justify accepting that risk.

    In other words, risk and reward are connected.

    This relationship also explains something many people misunderstand about probabilistic systems.

    If there were no uncertainty, there would be no opportunity.


    Why Edge Exists

    In probabilistic systems, an edge means the expected value of a decision is positive.

    Over many trials, the average outcome favors the strategy.

    But expected value does not eliminate uncertainty. It only describes the long-term tendency of outcomes.

    A strategy with a real edge can still lose repeatedly. That randomness is not evidence that the system is broken.

    It is evidence that the system operates in a world where outcomes are uncertain.


    The Role of Variance

    Variance is the mechanism that produces this uncertainty.

    Outcomes do not occur evenly or predictably. Wins and losses arrive in clusters. Runs of results can look extreme even when the probabilities remain unchanged. Results fluctuate around their long-term average.

    This behavior often feels chaotic in the moment.

    But it is also the reason opportunities exist.

    If outcomes were perfectly predictable, prices would reflect that certainty. Every opportunity would already be fully priced into the market.

    There would be no edge.

    Variance is what allows edge to exist in the first place.


    When Variance Looks Like Failure

    The presence of variance creates a challenge for anyone operating a probabilistic strategy.

    During periods when results run below expectation, it becomes difficult to tell whether something is wrong with the system or whether randomness is simply doing its job.

    This is where many people abandon good strategies.

    A losing streak feels like evidence of failure. A drawdown feels like proof that the edge has disappeared.

    But variance can easily produce sequences of results that look like failure even when the underlying probabilities remain intact.

    Variance often disguises itself as a broken system.

    Learning to recognize that distinction is one of the hardest parts of operating any probabilistic approach.


    Surviving the Inevitable

    Because variance cannot be eliminated, the goal is not to avoid it.

    The goal is to survive it.

    This is where risk management and position sizing become essential. A strategy with an edge must be structured so that it can endure the inevitable stretches when outcomes fall below expectation.

    Without discipline, even a positive expected value strategy can fail.

    Edge without discipline is indistinguishable from gambling.

    Sizing decisions determine whether variance becomes a temporary setback or a catastrophic loss.


    The Long Game

    Markets do not offer edge in spite of uncertainty.

    They offer edge because of it.

    If outcomes were predictable, prices would already reflect that certainty. Every opportunity would be arbitraged away.

    Variance is the reason edge can exist at all.

    The challenge is not avoiding variance.

    The challenge is surviving it long enough for the edge to matter.

  • Why Good Decisions Still Lose

    Good decisions still lose in probabilistic systems. Most people judge decisions by outcomes. If the result is good, the decision must have been correct. If the result is bad, the decision must have been flawed.

    In deterministic environments that reasoning often works. In probabilistic environments it fails.

    Markets, forecasting, and sports betting operate under uncertainty. Even when probabilities are calibrated and decisions are disciplined, a single outcome can still be wrong.

    In my previous essay I described several mistakes that cost me real money. Some losses came from execution errors. But losses alone do not prove the decision was wrong.

    In probabilistic systems, separating mistakes from variance is one of the hardest skills to learn.


    The Problem With Judging Results

    Consider a wager with a 60% chance of winning.

    That still means it loses four times out of ten.

    Losing streaks are normal. After all, a strategy that wins 60% of the time still loses four out of ten events. Over many trials those losses will cluster.

    The existence of a losing run does not invalidate the underlying probability.

    A good decision can produce a bad result. A bad decision can produce a good one.

    This idea is easy to understand mathematically. However, it is much harder to accept emotionally.

    Human intuition wants a simple story. When something works, we assume the reasoning was sound. When it fails, we assume a mistake must have been made.

    In probabilistic systems, that instinct often leads us in the wrong direction.


    When Outcomes Mislead Us

    After a losing streak, bettors often abandon strategies that may actually be profitable.

    After a winning streak, traders often increase risk because recent success creates confidence.

    Both reactions come from the same mistake: judging decisions by short-term outcomes.

    Short-term results are noisy.

    But outcomes alone cannot tell us whether a decision was good.

    What matters is the expected value at the moment the decision was made. If the probabilities were sound and the risk was sized appropriately, the decision was correct, regardless of the immediate result.


    Why Sizing Matters

    Even strategies with an edge experience losses.

    For this reason, risk management becomes essential.

    Position sizing determines whether variance becomes a temporary setback or a catastrophic failure. Proper sizing allows a strategy to survive the inevitable losing streaks that probabilistic systems produce.

    Uncertainty cannot be eliminated.

    It can only be managed.


    The Long Run

    Over short horizons, variance dominates.

    Over long horizons, disciplined process reveals itself.

    The goal of a probabilistic system is not to eliminate losses. Losses are unavoidable when outcomes contain uncertainty.

    The goal is to consistently make decisions with positive expected value and manage risk well enough to survive the variance along the way.

    Process over outcome.

    Because in probabilistic systems, outcomes are noisy but process compounds over time.