Stock Info Nets | Stock Trend Regression, NLP, Sentiment Analysis and News Attribution

In addition to price, volume and company information, each stock is analyzed using three independent linear regressions calculated from approximately the last 30 trading days of closing prices. Before the regressions are calculated, prices are converted into percentage returns relative to the first trading day of the period so that companies with very different share prices can be compared on the same scale. The reported slope is measured in percentage points of return per trading day (%/day). For example, a slope of +0.40%/day means the regression line increased, on average, by approximately 0.40 percentage points each trading day, while a slope of -0.25%/day means it decreased by about 0.25 percentage points per trading day.

Days 1–15 represents the linear regression fitted only to the oldest 15 trading days within the 30-day window. The reported value is the regression slope (%/day). Positive values indicate the stock was generally appreciating during the first half of the period, negative values indicate it was declining, while values close to zero indicate little or no directional trend.

Days 16–30 represents the regression fitted only to the most recent 15 trading days. Its slope is also expressed in %/day. This is generally the best estimate of the stock's current direction and recent momentum.

Days 1–30 is the regression fitted to the complete 30-day period. Its slope, also expressed in %/day, summarizes the overall direction across the month while smoothing short-term fluctuations. Although it provides useful context, it is not used to determine whether momentum is accelerating or weakening.

Delta is calculated as the difference between the recent and initial regression slopes (Days 16–30 slope − Days 1–15 slope). Since both slopes are measured in %/day, Delta is also expressed in percentage points per trading day. A positive Delta means the trend became more positive (or less negative) during the second half of the period, while a negative Delta indicates weakening momentum or an increasingly negative trend. Very small changes are treated as statistical noise and classified as Unchanged.

Delta Classification categorizes the change in momentum into three groups: Increasing, Decreasing, or Unchanged. This avoids interpreting very small differences between the two regression slopes as meaningful changes in market behavior.

Initial Trend and Current Trend simply indicate whether each 15-day regression slope is Positive, Negative, or approximately Flat. Together they describe how the direction of the trend evolved during the month.

Interpretation describes how the most recent 15-day trend has evolved relative to the overall 30-day trend. The 30-day regression establishes the dominant market direction (Uptrend, Downtrend, or Flat), while the first and last 15-day regressions are compared with the 30-day regression's ±1σ variability band. A 15-day regression that remains inside the band is considered consistent with the longer-term trend, whereas one that moves above or below the band indicates a meaningful departure from the expected trend.

Flat Trend: Stable Sideways indicates both 15-day regressions remain inside the variability band, suggesting no meaningful directional bias. Emerging Uptrend and Emerging Downtrend indicate that the latest 15-day regression has moved outside the band, providing an early indication of a new trend. Emerging Uptrend Fading and Emerging Downtrend Fading indicate that an emerging trend has returned inside the band and is losing momentum. Strong Upward Breakout and Strong Downward Breakout occur when both 15-day regressions remain outside the band on the same side, indicating sustained movement away from the previous sideways market. Reversal Toward Uptrend and Reversal Toward Downtrend indicate that the short-term trend has shifted from one side of the variability band to the other.

Uptrend: Stable Uptrend indicates both 15-day regressions remain inside the variability band, confirming a healthy and consistent uptrend. Accelerating Uptrend indicates the latest 15-day regression has moved above the band, showing stronger upward momentum. Uptrend Weakening indicates the latest 15-day regression has moved below the band, suggesting the uptrend is losing strength. Uptrend Stabilizing indicates a previous acceleration has returned inside the variability band. Uptrend Recovering indicates a previous weakening has returned inside the variability band. Strong Accelerating Uptrend indicates both 15-day regressions remain above the band, showing persistent strengthening. Strong Weakening Uptrend indicates both 15-day regressions remain below the band, showing persistent weakening. Uptrend Recovery indicates the uptrend has resumed after a temporary period of weakness. Uptrend Reversal to Downtrend indicates the short-term trend has shifted from above the variability band to below it, suggesting a possible reversal.

Downtrend: Stable Downtrend indicates both 15-day regressions remain inside the variability band, confirming a healthy and consistent downtrend. Accelerating Downtrend indicates the latest 15-day regression has moved below the band, showing stronger downward momentum. Downtrend Weakening indicates the latest 15-day regression has moved above the band, suggesting the downtrend is losing strength. Downtrend Stabilizing after Acceleration indicates an unusually strong decline has moderated back inside the variability band. Downtrend Stabilizing after Rebound indicates a temporary rebound has faded and the trend has returned inside the variability band. Strong Accelerating Downtrend indicates both 15-day regressions remain below the band, showing persistent strengthening of the downtrend. Strong Weakening Downtrend indicates both 15-day regressions remain above the band, showing persistent weakening of the downtrend. Downtrend Resuming indicates the downtrend has resumed after a temporary rebound. Uptrend Resuming indicates the short-term trend has shifted from below the variability band to above it, suggesting a possible transition toward an uptrend.

Noise (σ) is the standard deviation of the differences between the observed percentage returns and the LOESS trend line. It measures how closely prices follow the underlying trend. Lower values indicate smoother price action, while larger values indicate greater day-to-day volatility around the trend.

These indicators are intended as descriptive statistical measures, not price forecasts. They summarize the direction, strength and evolution of recent price behavior using objective regression analysis, allowing investors to compare the momentum of different securities under a consistent framework.

Understanding the Trend Graph

The graph summarizes approximately the last 30 trading days of closing prices after converting them into percentage returns relative to the first trading day. Expressing prices as percentage returns rather than dollars allows companies with very different share prices to be compared on the same scale. A value of +10% indicates the stock closed approximately 10% above the first day of the period, while −5% indicates it closed about 5% below that starting point.

The light blue line connects the daily percentage returns and represents the actual day-to-day movement of the stock. The black dots identify each individual trading day's closing price used in the statistical calculations. Since daily prices naturally fluctuate, these observations should be viewed as the raw market data rather than the underlying trend.

The dark blue curve is a LOESS (Locally Weighted Regression) trend. Unlike a straight regression, LOESS fits many local regressions to produce a smooth curve that follows gradual accelerations, slowdowns and turning points. It estimates the underlying market trend while reducing much of the short-term price noise.

The light blue shaded region surrounding the LOESS curve represents ±1 standard deviation (±1σ) of the residuals around the trend. This band illustrates the typical daily variability of prices around the estimated trend. A narrow band indicates relatively smooth price action, while a wider band indicates greater day-to-day volatility. It should not be interpreted as a forecast or prediction interval.

The blue dotted line is the linear regression calculated from the oldest 15 trading days, while the blue dashed line represents the regression calculated from the most recent 15 trading days. Comparing these two lines helps identify whether momentum has strengthened, weakened or reversed during the month.

The solid black line is the linear regression calculated from the entire 30-day period. It represents the dominant monthly trend and serves as the reference for evaluating the shorter-term regressions. Because it uses all observations, it reacts more slowly to recent changes than the 15-day regressions.

The gray hatched region represents ±1σ around the 30-day regression and defines the range of normal variation around the longer-term trend. When a 15-day regression remains inside this band, it is considered consistent with the monthly trend. Moving above or below the band indicates that recent price behavior has deviated meaningfully from the longer-term trend and may signal strengthening, weakening or reversal of momentum.

The light gray shaded area surrounding the 30-day regression is the 95% confidence interval. It measures the statistical uncertainty in estimating the average regression line. A narrower interval indicates greater confidence in the estimated trend, while a wider interval reflects greater uncertainty. This interval describes the precision of the regression estimate and should not be interpreted as the expected range of future prices.

These graphical indicators are descriptive statistical tools rather than forecasting models. They summarize the direction, consistency and evolution of recent price behavior using objective statistical methods, allowing investors to compare trends across different securities under a consistent analytical framework.

Understanding the Sentiment Analysis

The sentiment analysis is generated through zero-shot in-context learning. Rather than relying on predefined keywords or manually constructed rules, the model evaluates the overall language, context and tone of the available textual information to estimate the prevailing market sentiment surrounding the company.

The analysis is performed exclusively from textual content. It does not use stock prices, trading volume, regression coefficients, trend indicators, technical analysis or the statistical interpretation of the price charts. The objective is to measure the sentiment expressed in the text independently from the market's recent price behavior.

The Sentiment Score is reported as a continuous value between 0.00 and 1.00. Lower values indicate language associated with deteriorating conditions and increasing downside pressure, while higher values indicate language suggesting improving conditions, stronger confidence and positive expansion. Intermediate values represent mixed or balanced sentiment.

To simplify interpretation, each numerical score is converted into one of five descriptive categories: Breakdown (0.00–0.20), Weakness (0.20–0.40), Neutral (0.40–0.60), Building (0.60–0.80), and Surge (0.80–1.00). These labels summarize the overall tone of the analyzed text and should be interpreted as qualitative descriptions rather than investment recommendations.

Because the model evaluates language probabilistically, the score should be viewed as an estimate of the sentiment conveyed by the available information at the time of analysis. It is intended to complement, rather than replace, fundamental research, financial analysis and statistical price indicators.