4. The psychological biases that wreck portfolios

Theory 4: Risk, return and diversification

Losing a given amount of money hurts more than gaining that same amount pleases. That asymmetry, apparently harmless, explains a large part of the bad investment decisions people make.

Everything we have seen at this level — volatility, correlation, risk-adjusted return — implicitly assumes that whoever makes the decisions acts rationally and consistently.

Decades of accumulated evidence from behavioral finance show that this is rarely true, not even among experienced investors. A handful of specific psychological biases explain a large share of the bad investment decisions that mathematics alone would never predict.

Loss aversion, and the disposition effect

The most studied bias, and probably the most expensive, is loss aversion. The psychological pain of losing a given amount of money is, on average, considerably greater than the pleasure of gaining that same amount.

This imbalance was originally documented by Daniel Kahneman and Amos Tversky in their 1979 prospect theory. Kahneman received the 2002 Nobel Prize in Economics for that work; Tversky had died in 1996 and the prize is not awarded posthumously.

Out of it comes a very specific behavioral pattern, known as the disposition effect:

• Investors tend to sell their winning positions too early, for fear the gain will evaporate.
• And to hold their losing positions too long, hoping to 'get back to even' before booking the loss as final.

This is exactly the opposite of what purely rational risk management would advise. Such management would judge every hold-or-sell decision on the asset's future prospects, not on whether the position is in profit or at a loss relative to the original purchase price.

Anchoring: a number that should not matter

Anchoring is another bias with a very concrete effect on trading decisions. It is the tendency to give excessive weight to an initial reference number — typically the price at which an asset was bought — when judging future decisions.

An investor anchored to his purchase price may refuse to sell a losing position 'until it gets back to what I paid'. As though the market had some obligation to hand back exactly that price.

The reality is colder. The price you paid bears no relation to the asset's future value, nor to the rational decision to hold it or sell it today.

Overconfidence grows with every win

Overconfidence consists of systematically overestimating one's own predictive or analytical ability.

It is especially dangerous for a rather counter-intuitive reason: it tends to grow, not shrink, after a winning streak. A few consecutive profitable trades can convince anyone that they have 'discovered' a reliable pattern, when statistically this is perfectly consistent with plain luck in a small sample.

This bias feeds off confirmation bias. That is the tendency to seek out, remember and give more weight to information that confirms what we already believed. And to ignore or downplay information that contradicts it.

The example is easy to recognize. An investor convinced that a stock is going to rise will read every positive headline closely and dismiss every negative one as 'noise'. He thereby reinforces his own conviction artificially, without the actual evidence justifying it.

Herding: why bubbles form

Finally, herding, or herd behavior: the tendency to copy what the majority is doing, especially in moments of uncertainty, implicitly assuming that 'so many people cannot all be wrong at once'.

It is one of the mechanisms behind speculative bubbles and selling panics. Both are fed, to a large extent, not by changing fundamental analysis.

They are fed by each participant watching the other participants act and adjusting his own behavior accordingly. That amplifies the collective move far beyond what the available information would justify.

Why this pushes toward the algorithm

None of these biases disappears simply by knowing about them intellectually. In fact, most people who know them keep falling into them under real pressure.

And it is precisely this persistence, hard to eradicate even through willpower, that is one of the underlying reasons algorithmic trading appeals to many investors.

A set of rules defined in advance, executed without emotional intervention at the moment of decision, is structurally immune to the disposition effect, to anchoring or to overconfidence.

With one important caveat, which we will develop in the following levels: it introduces its own distinct risks. Starting with the design of the rule set itself, which is also biased by whoever builds it.