Method
Every value in a PEG brief is one of three things: calculated from retrieved market data, classified by a deterministic rule, or interpreted in language by AI from those values. Nothing is forecast.
Each family produces a 0–100 sub-score from its calculated values. The PEG Score is the weighted sum. It summarizes alignment between observable conditions and is never a probability that a stock will rise.
Tiers
A sub-score whose inputs cannot be computed is dropped from the blend and the remaining weights are renormalized — a missing value never counts as a zero.
Where price sits relative to its 20, 50 and long-term simple moving averages, and whether those averages are stacked (20 above 50 above long). Distance from each average is expressed in percent.
Rate of change over 5, 20 and 60 sessions, plus acceleration (the change in 5-day momentum versus five sessions earlier) and RSI 14 as an overextension check. RSI can only subtract from the sub-score; it never adds.
The symbol's rate of change minus SPY's over the same 20 and 60-session windows, in percentage points. QQQ is computed as a secondary reference. Benchmark windows are aligned to the symbol's own session dates, so a halted session cannot shift the comparison.
Relative volume is the latest session’s volume divided by the trailing 20-session average, excluding the latest session. Heavy volume on an up day is treated as confirmation; heavy volume on a down day is not.
Wilder ATR 14 expressed as a percentage of price, annualized 20-day close-to-close volatility, maximum 60-day drawdown and distance from 52-week extremes. Risk combines ATR, extension from the 20D, drawdown and RSI extremes into a four-step label.
SPY, QQQ and IWM are each scored on position versus their moving averages, 5 and 20-day momentum and realized volatility. The weighted blend (SPY 45%, QQQ 35%, IWM 20%) is adjusted for agreement between indices and for elevated volatility, then mapped to Risk On, Constructive, Mixed or Defensive.
The Market Terrain visualization maps calculated dimensions onto a stylized landscape: trend sub-score controls elevation, 20-day momentum controls slope, ATR controls roughness, relative strength controls peak prominence and the risk label places markers. The same inputs always draw the same terrain. It is a representation of the numbers above it, not a model of the future.
The Read, Scenario Map and supports/risks are written by a language model that receives only the calculated values, classifications, PEG Score and market regime. It does not see news, earnings, fundamentals or price history, and it is instructed to distinguish observation from inference, to use calibrated language, and never to give directives.
If the model is unavailable or no API key is configured, a deterministic text is generated from the same values so the brief always renders. Either way, every statement can be checked against the Signal Matrix.
Quotes and price history are retrieved server-side from a public market-data source that does not guarantee real-time delivery. Every brief shows its session date and retrieval timestamp, and notes that data may be delayed.
Indicators are computed only from retrieved prices and volume. When history is insufficient a value is shown as — rather than estimated.
Two limits are worth stating plainly. Where the retrieved daily history is shorter than a 200-day average, the long-term trend reference is taken from a 40-week average of the weekly series instead — and every brief labels which basis it used rather than substituting one for the other silently. And prices are not back-adjusted for splits or dividends.
Retrieved series are refreshed once per session and reused between reads, so viewing a page does not re-request data.
PEG Markets is research and education only. Nothing here is investment advice.
Analyze a ticker