How AI Read the Last Bull Market
Bull markets end in data before they end in price.
Bull markets do not end when conditions stop being good. They end when expectations for future conditions stop being consistent with the valuations that have been built around them. The data signals this transition in the same way it signals every other regime change: through divergences between what prices are embedding and what the information environment is showing. The emotional experience of a bull market peak is one of confidence and momentum. The quantitative reading is one of accumulating inconsistencies between price and data.
A bull market is defined, conventionally, as a sustained rise of 20% or more in a major market index from a recent low, persisting over at least several months. The bull market that followed the March 2020 pandemic trough, in which the S&P 500 rose from approximately 2,200 at its low to over 4,700 by January 2022, a gain of more than 110%, represents one of the most rapid and substantial recoveries in modern market history. It was also a case study in the divergence between accelerating asset prices and the data signals that quantitative systems were tracking in the months before the peak.
This article describes what systematic analysis reads during a bull market, and specifically what warning signals the data produces before the advance concludes.
Bull markets produce their own data distortions: reading through the noise
The challenge of quantitative analysis during a sustained bull market is not the absence of signals. It is the accumulation of noise that a long advance generates. In extended bull markets, multiple potential warning signals appear and then resolve without consequence: breadth deteriorates briefly and then recovers, sentiment reaches extreme positive readings and prices continue to rise, volatility spikes without producing a regime change. Each false signal conditions market participants to discount subsequent signals, including the ones that precede the actual regime transition.
The Noise Threshold concept is critical here. A quantitative system must distinguish between readings that represent genuine regime deterioration and those that fall within the normal variation of a bull market environment. The methodological challenge is that the Noise Threshold is not fixed. What constitutes meaningful divergence in a high-volatility environment differs from what constitutes meaningful divergence in a low-volatility trending environment.
During the 2021 advance, the Signal Stack was showing a sustained positive trend across major equity indices, supported by strong sentiment readings from earnings communications and constructive macro data. The Market Regime was firmly classified as trending positive. This is the correct reading for that period. Quantitative systems do not impose pessimism on constructive environments. They read the data.
Warning signals emerged in the data months before the January 2022 peak
In the autumn and winter of 2021, several quantitative indicators began to show the characteristic pattern of late-cycle divergence.
Sentiment in interest-rate-sensitive communications, including mortgage market commentary, housing sector analyst reports, and corporate issuance language, began to shift toward caution as inflation data started to exceed Federal Reserve projections. The Sentiment Layer was registering a change in the information environment around rate sensitivity that was not yet reflected in the broad equity index-level advance.
Breadth indicators began to deteriorate. The percentage of S&P 500 constituents trading above their 200-day moving averages peaked in mid-2021 and began a gradual decline through the second half of the year, even as the index continued to make new highs driven by a concentrated set of large-capitalisation names. This is the classic late-cycle breadth pattern: the advance narrowing before it ends.
Volatility Adjusted Signals for the most rate-sensitive segments of the equity market, high-multiple growth stocks and long-duration technology names, were already showing reduced confidence scores by late 2021, reflecting the elevated noise floor as volatility in those segments increased. The same trend signal in a higher-volatility environment carries less information than in a lower-volatility one.
The emotional response: why the bull market narrative outlasted the data
The emotional response to the warning signals appearing in the data during the 2021 advance was shaped by the same mechanism that operates in every late-cycle environment: the experiential prior of a multi-year bull market had conditioned market participants to treat warning signals as buying opportunities. Each prior signal that had resolved without consequence reduced the effective weight given to subsequent signals.
This is not irrationality. It is Bayesian updating from the wrong sample. In an environment where every prior warning signal had been wrong, the rational Bayesian update is to reduce the weight given to the next warning signal. The problem is that this process, applied consistently, produces maximum scepticism about warning signals precisely at the moment when the regime is actually transitioning.
The Emotionless Edge here is structural: a quantitative model does not update its response to warning signals based on the prior record of false positives. It evaluates the current data against its trained parameters and reports what it finds. When the breadth was deteriorating and sentiment was shifting in late 2021, the systematic reading was of a market entering late-cycle conditions. The emotional reading was of a market in a continuing advance.
The durable lesson: bull markets end in the data before they end in price
The consistent finding across bull market cycles is that the warning signals are visible in systematic data before they are visible in the headline index. Breadth peaks before the index. Sentiment diverges before the trend reverses. Volatility begins to rise in the most rate-sensitive or most extended segments before the broad market confirms the regime change.
The Opes Borsa Signal Stack monitors all of these dimensions simultaneously, providing a live read of where the current market environment sits within its historical pattern. The platform is not a prediction engine. It is a systematic framework for reading what the data shows about current conditions, updated continuously. You can explore the current reading at opesborsa.com.
Bull markets do not need to be predicted to be navigated systematically. They need to be read accurately as they evolve, including the phase in which the data is signalling transition while the narrative is still pointing upward.
Key Terms:
Bull Market: A sustained increase of 20% or more in a major market index from a recent low, typically persisting over several months to years. Characterised by rising prices, positive sentiment, and expanding breadth in its healthy phases.
Late-Cycle Conditions: The phase of a bull market in which the advance narrows, breadth deteriorates, and valuation assumptions begin to diverge from the data environment. Identifiable in systematic data before the index-level price confirms the regime transition.
Volatility-Adjusted Signal: A Trend Signal calibrated for the prevailing volatility environment, with confidence scores adjusted downward during high-volatility periods to reflect the elevated noise floor. Provides a more consistent signal quality comparison across different market environments.
Noise Threshold: The level above which a quantitative signal represents a genuine regime change rather than normal variation within the prevailing trend. Distinguishing genuine warning signals from bull-market noise is the central challenge of late-cycle quantitative analysis.
Breadth Deterioration: The narrowing of the proportion of market constituents participating in an index-level advance. One of the most historically consistent leading indicators of a late-cycle regime transition.
