Long memory, regimes, and the limits of bell-curve models

Name the hidden contract

A quantitative rule is only as trustworthy as the market state it assumes. The archive described three assumptions that often sit inside that rule without being named: short memory, a bell-curve tail, and a continuous price path.

Short memory is the random-walk-style assumption that today's move has little effect on the next few sessions and almost none farther out. A bell-curve tail treats large moves as rare outliers. A continuous path assumes the market will trade the intermediate prices a hedge, stop, or rebalance needs.

Editorial interpretation: write those assumptions down before the trade. If the live market breaks the contract, treat that break as a change of market state, not as a rare exception the same rule can absorb.

Long memory breaks short-memory rules

A long-memory process is one in which current outcomes keep influencing later outcomes over an extended horizon instead of fading quickly. A long-memory diagnostic applied to market returns found that current outcomes continued to influence later outcomes rather than fading as a short-memory random walk would require.

Quantitative techniques that depend on short memory become unreliable if that memory assumption is false. Editorial interpretation: if a signal, filter, or forecast needs yesterday's move to wash out, a long-memory state is not noise around that rule. It is a different contract.

Fat tails at every return scale

The described return distribution has a high peak at the mean and fat tails, so both ordinary outcomes and large moves occur more often than a normal curve implies. That shape appears in daily, monthly, and annual returns. Fat tails are extra probability in the extremes relative to a normal, bell-shaped curve.

Events a normal distribution would treat as roughly once in 100 years were described as occurring every few years in the observed return distribution. Editorial interpretation: do not treat the extremes as calendar curiosities. They are part of the ordinary shape the trade has to survive.

Price discontinuities and capital timing

In fat-tailed markets, large moves tend to be abrupt and discontinuous, so prices can skip the intermediate levels a continuous-rebalancing rule needs. That skip is a price discontinuity: a jump or gap that leaves untraded the prices a continuous hedge, stop, or rebalance rule expects to use.

In the 1987 crash, trades could not be executed quickly enough on the decline for hedges that assumed continuous prices to work as designed.

A highly leveraged book can fail even when a mean-reversion assumption is eventually right, if the reversion arrives too slowly and capital is exhausted first. Editorial interpretation: a regime audit has to ask two timing questions, not one. Can the path be traded as a continuous line, and can the book stay solvent until the assumed reversion arrives?

Global structure in a complex regime

Recurring cycle mechanics such as yield-curve steepening during expansions and contractions persist while local narratives change. That persistence is global structure: recurring cross-cycle mechanics that remain even when the local economic story changes. Local randomness is the cycle-specific detail that should not be locked into a forecasting model as if it will repeat. Models fitted to one-time details do not transfer across regimes.

Markets were characterized as usually complex and self-adapting, becoming chaotic only when fluctuations become extremely large, so the internal state is never fully known. That working state is a complex regime: a self-adapting market between order and chaos, chaotic only when fluctuations become extreme.

Bear-market endings were described as fading interest rather than a single capitulation washout. A climactic selloff was assigned to bull-market reversals instead. Editorial interpretation: an ending template is a regime claim. Using a washout rule to exit a bear, or ignoring fading interest because the local story still sounds unfinished, confuses local randomness with global structure.

An editorial regime audit

Editorial interpretation: before trusting any quantitative rule, name the hidden contract. Does the rule need short memory, a bell-curve tail, and a continuous price path? Then check whether the market still honors that contract.

If long memory is present, short-memory techniques are already in the wrong state. If fat tails and price discontinuities are present, continuous-rebalancing and thin-tail risk math are in the wrong state. If the model was fitted to a one-time narrative, it has captured local randomness rather than global structure and will not transfer.

The last check is survival, not elegance. A mean-reversion story can be right in the end and still fail a leveraged book if capital is gone before the reversion arrives. Because a complex regime never makes the internal state fully known, the audit is a pre-trade habit, not a claim that the next state has been identified.