7. The gambler's fallacy and other reasoning errors

Level 1: Probability: the language of uncertainty

A stock has closed lower for five consecutive sessions. "It cannot keep falling," the trader reasons. "A bounce is due."

That sentence, so reasonable to the ear, contains a probability error with a name of its own. And it is not the only one this module takes apart.

The gambler's fallacy

The gambler's fallacy is the mistaken belief that, if a genuinely random and independent process has produced an unusual run of one outcome, the opposite outcome becomes more likely on the next attempt. As if the process needed to compensate or rebalance what has happened so far.

It is a misapplication of an idea that is true in the long run: that proportions converge toward the real probability when the process is repeated an enormous number of times.

From there it jumps, improperly, to a different and incorrect claim about the next individual outcome. That is something the very definition of independence, developed in the previous module, explicitly rules out.

Return to the example from the opening. If that stock's daily moves were reasonably independent of one another, the probability of it rising the next day does not increase simply because of the prior run. The process has no memory of what has happened.

Independence here is a fair approximation for many liquid large-cap stocks over short horizons.

Careful with the opposite extreme

A crucial nuance is worth stressing so you do not swing to the opposite extreme: this does not mean that expecting reversion is always a fallacy.

Some markets and instruments do exhibit real, measurable, tested statistical dependence: a genuine tendency to revert toward the mean over certain horizons. That is a completely different claim from the psychological reasoning that something is 'due' with no evidence behind it.

The error is not in the direction of the belief. It is in building it on an unverified intuition about a law of averages, instead of on dependence that has actually been measured and tested against data.

The hot hand

There is also the inverse version, colloquially known as the hot hand phenomenon.

After a run of winning trades, some traders over-read the run itself as proof that a special edge or skill has suddenly appeared. It may simply be a normal streak within an essentially random or near-random process that has not changed at all.

And what separates it from genuine momentum, a documented effect, statistically tested over large samples? The same question as always: whether persistence has been rigorously measured over a lot of data, or merely asserted from a short recent run.

The clustering illusion

Another common error is the clustering illusion: people tend to perceive patterns, trends or recognizable shapes in sequences that are in fact close to pure random noise.

The cause is counterintuitive. Truly random sequences contain runs and clusters far more often than intuition expects.

A perfectly fair coin, tossed many times, will produce runs of five or six consecutive heads reasonably often over the course of many tosses.

Mistaking that expected by-product of pure chance for a meaningful signal is a habitual error when looking at price charts.

The conjunction fallacy

The conjunction fallacy also shows up frequently, adapted to market scenarios. It consists of judging a very specific and detailed scenario as more likely than a broader, more general scenario that fully contains it, simply because the detailed story feels more vivid and coherent.

Compare these two claims:

• That the Federal Reserve cuts rates in September, and that this triggers a strong tech-led rally which fades before December.
• That the Federal Reserve cuts rates in September.

Judging the first as more likely than the second is a mathematical error. The compound scenario adds extra conditions, and an additional condition can never make a scenario more likely than any of its parts taken separately. At most equally likely.

Regression to the mean is not correction

Finally, confusion around regression to the mean is especially relevant when evaluating managers and strategies.

After an unusually extreme result — a fund manager's best quarter on record, a stock's sharpest rally — subsequent results tend to move back toward the usual average.

And here is the nuance. This does not happen because there is any active correction or rebalancing mechanism, as in the gambler's fallacy.

It happens because an extreme result usually contains two components mixed together:

• The stable, repeatable component of genuine skill or edge.
• A temporary favorable component, unlikely to repeat.

Attributing that move back toward the average to a causal explanation — the manager lost his touch, the market punished the stock — usually invents a narrative for what is largely a plain statistical artefact.

The common root

All of these errors share a common root: substituting an intuitive narrative, or a feeling that some law of balance is at work, for the formal, measurable tools built throughout this level.

Those tools are the ones you already have: conditional probability, random variables, expected value, variance and independence.

The rest of the quantitative mathematics track that follows this level is, to a large extent, devoted to the same thing: replacing that narrative instinct with numbers you can verify against data.