InvestigeniusMethodology

What every number on this site means

Investigenius pulls data from public market and news sources, computes standard statistical and technical indicators, and runs a small set of forecasting and classification models. Nothing here is a recommendation — every number below is descriptive, not prescriptive. Use them as inputs, not as conclusions.

Data may be delayed or incomplete.

Quotes & fundamentals

Headline numbers shown at the top of every ticker page. Pulled live from Yahoo Finance.

  • Last price, day change, day rangeYahoo Finance

    Most recent trade and intraday high/low. Reflects whatever delay the venue applies (typically 15 minutes for US equities).

  • Market capYahoo Finance

    Shares outstanding × last price.

  • P/E (trailing)Yahoo Finance

    Price divided by trailing-twelve-month earnings per share. Negative or missing for unprofitable companies.

  • P/BYahoo Finance

    Price divided by book value per share.

  • ROEYahoo Finance

    Net income divided by shareholders' equity (TTM).

  • Beta (5Y monthly)Yahoo Finance

    Sensitivity to the broad market over the last five years of monthly returns. 1.0 = moves with the market.

  • Dividend yieldYahoo Finance

    Trailing 12-month dividend per share divided by price.

  • 52-week high / lowYahoo Finance

    Highest and lowest close over the last 252 trading days.

  • Average volume (10D, 90D)Yahoo Finance

    Mean daily share volume over the trailing window.

  • Short interest, days-to-coverYahoo Finance

    Reported short shares as a percent of float, plus how many average-volume days it would take to close them.

Returns & price transforms

Building blocks every other indicator is derived from. Computed from daily OHLCV candles.

  • Simple returnComputed from candles

    (price_t − price_{t−1}) / price_{t−1}.

  • Log returnComputed from candles

    ln(price_t / price_{t−1}). Additive across time, used for volatility and most statistical models.

  • Cumulative returnComputed from candles

    Compound product of (1 + simple return) over the window.

  • Rolling drawdownComputed from candles

    Percent below the running maximum close. The minimum of this series over the window is the max drawdown.

Trend & moving averages
  • SMA (20, 50, 200)Computed from closes

    Arithmetic mean of the last N closes.

  • EMA (12, 26)Computed from closes

    Exponential moving average with α = 2/(N+1). More weight on recent prices than SMA.

  • MACD (12, 26, 9)Computed from closes

    EMA12 − EMA26, with a 9-period EMA signal line. Histogram = MACD − signal.

  • ADX (14)Computed from OHLC

    Average Directional Index. Measures trend strength (not direction) on a 0–100 scale; ~25+ is commonly read as a trending regime.

Momentum & oscillators
  • RSI (14)Computed from closes

    Relative Strength Index. Average gain / average loss over 14 periods, scaled to 0–100.

  • Stochastic (%K, %D)Computed from OHLC

    Where the close sits within the recent high–low range, with a smoothed signal line.

  • Williams %R (14)Computed from OHLC

    Inverse Stochastic on a −100 to 0 scale.

  • ROC (12)Computed from closes

    Rate of Change. Percent change vs. 12 periods ago.

  • CCI (20)Computed from typical price

    Commodity Channel Index. Distance of typical price from its SMA, scaled by mean deviation.

  • MFI (14)Computed from OHLCV

    Money Flow Index. Volume-weighted RSI; integrates how much money traded on up vs. down moves.

Volatility & range
  • Historical volatility (20D, 60D)Computed from log returns

    Standard deviation of daily log returns × √252 to annualize.

  • Bollinger Bands (20, 2σ)Computed from closes

    SMA20 ± 2 standard deviations of price. Bandwidth contracts in low-vol regimes and expands when ranges widen.

  • ATR (14)Computed from OHLC

    Average True Range. Wilder-smoothed daily range, accounting for gaps.

  • Parkinson volatilityComputed from highs/lows

    Range-based estimator using daily high–low. More efficient than close-to-close when intraday ranges are informative.

  • Garman–Klass volatilityComputed from OHLC

    Uses open, high, low, and close together. Lower variance than Parkinson under most conditions.

  • Yang–Zhang volatilityComputed from OHLC

    Combines overnight, opening, and intraday components; handles overnight jumps better than range-only estimators.

  • EWMA volatility (λ = 0.94)Computed from log returns

    Exponentially weighted moving variance, RiskMetrics-style. Reacts faster to regime shifts than rolling stdev.

Risk-adjusted performance
  • Sharpe ratioComputed from returns

    (Mean return − risk-free) / standard deviation of returns. Annualized.

  • Sortino ratioComputed from returns

    Like Sharpe, but the denominator only counts downside deviation.

  • Calmar ratioComputed from returns

    Annualized return divided by absolute max drawdown.

  • Omega ratioComputed from returns

    Probability-weighted ratio of gains above a threshold to losses below it. Captures the full distribution shape.

  • Max drawdownComputed from closes

    Largest peak-to-trough percent decline over the window.

Distribution shape & tails
  • SkewnessComputed from returns

    Asymmetry of the return distribution. Negative = fat left tail (more big down days than a normal distribution predicts).

  • Kurtosis (excess)Computed from returns

    Tail heaviness vs. a normal distribution. Positive = more extreme moves than Gaussian assumes.

  • Hurst exponentComputed from returns

    Long-memory measure. ~0.5 = random walk, >0.5 = trending, <0.5 = mean-reverting.

  • Jarque–Bera testComputed from returns

    Joint test of skewness and kurtosis against the normal distribution. Reported with p-value.

  • Augmented Dickey–FullerComputed from prices

    Stationarity test. Rejecting the unit-root null suggests prices are mean-reverting on the tested horizon.

Value-at-Risk & expected shortfall

Loss estimates at common confidence levels.

  • Parametric VaR (95%, 99%)Computed from returns

    Assumes returns are normal: VaR = μ − z·σ. Simple, but understates tail risk when returns are fat-tailed.

  • Historical VaR (95%, 99%)Computed from returns

    Empirical quantile of realized returns. Makes no distributional assumption.

  • Cornish–Fisher VaRComputed from returns

    Normal VaR adjusted for observed skew and kurtosis. Closer to historical VaR when distributions are non-normal.

  • Expected shortfall (CVaR)Computed from returns

    Mean loss conditional on being beyond the VaR threshold. Captures how bad the tail gets, not just where it starts.

Forecasting models

Run against 5 years of daily candles. None of these are predictions of where the price will go — they are model outputs over distributional assumptions.

  • ARIMA(1, 1, 1)Computed

    Autoregressive integrated moving average on log prices. Reports point forecast plus confidence band over the requested horizons.

  • GARCH(1, 1)Computed

    Volatility model with persistent variance. Used to project a forward volatility cone, not a price.

  • Geometric Brownian Motion (Monte Carlo, 10,000 paths)Computed

    Simulates 10k future paths under GBM with drift and σ estimated from realized returns. Reports horizon-by-horizon percentiles (P5/P50/P95) at 5, 10, 30, 90, 180, and 360 trading days.

Probability & ML classifiers

Probability that the close N days from now is higher than today, estimated by three independent models. The ensemble averages their probabilities.

  • Random Forest classifierComputed

    100 trees, default depth. Trained per-request on the symbol's own history.

  • Logistic regressionComputed

    Linear baseline with L2 regularization. Gives a calibrated probability under linear-separability assumptions.

  • Histogram gradient boostingComputed

    Fast boosted trees over binned features. Usually the strongest of the three on noisy financial data.

  • FeaturesComputed from candles

    1-day, 5-day, and 10-day returns; SMA20 and SMA50 distance from price; 20-day realized vol; RSI(14).

Sentiment

Each message is scored individually and bucketed by source, then aggregated.

  • VADER scorerComputed

    Rule-based compound score in [−1, 1]. Tuned for short, informal text — well-suited to social posts and headlines.

  • StockTwits messagesStockTwits

    Recent ticker-tagged messages from the public stream.

  • Reddit postsReddit

    Recent posts mentioning the symbol from finance subreddits.

  • X (Twitter) postsX

    Recent cashtag mentions.

  • Yahoo News headlinesYahoo Finance

    Headlines from the symbol's news feed.

  • Finnhub headlinesFinnhub

    Curated company news.

  • Alpha Vantage headlinesAlpha Vantage

    Adds Alpha Vantage's own sentiment label alongside the VADER score.

  • GDELT mentionsGDELT

    Broad-spectrum news mentions across global outlets.

  • Combined scoreComputed

    Mean VADER compound across all available messages, weighted equally per message.

Government / Congress trades
  • Recent disclosed tradesPublic US Congress disclosures

    STOCK Act disclosures aggregated from public filings. The home-page widget covers a curated universe plus today's trending tickers, merged by trade date.

  • Per-symbol detailPublic US Congress disclosures

    Politician, party, state, action (buy/sell/exchange), amount range, trade date, filing date, and reporting gap (filing − trade).

  • Caveat

    Disclosure deadlines mean filings can lag the actual trade by 30+ days. "Most recent" here means most recently disclosed, not most recently transacted.

Macro snapshot

Five macroeconomic series displayed at the top of the home page for context. Refreshed periodically.

  • CPI (YoY %)FRED

    Year-over-year change in headline US CPI. Calculated as (latest − value 12 months prior) / value 12 months prior × 100.

  • Unemployment rateFRED

    U.S. headline unemployment rate, latest monthly value.

  • 10-Year Treasury yieldFRED

    Constant-maturity yield, latest daily close.

  • Fed Funds rateFRED

    Effective federal funds rate, latest monthly average.

  • VIXFRED

    CBOE Volatility Index, latest daily close.

Earnings & filings
  • Upcoming and past earningsYahoo Finance

    Scheduled earnings date, EPS estimate, EPS actual, and surprise percent for the most recent and next quarters.

  • SEC filingsSEC EDGAR

    Recent 10-K, 10-Q, 8-K, and insider Form 4 filings linked to their EDGAR documents.

  • Upcoming IPOsFinnhub + Nasdaq

    Calendar of US listings expected to price in the next 30 days. Each entry has a dedicated page showing the company, proposed ticker, exchange, price range, share count, and an estimated raise. Once trading begins under the assigned ticker, the full analysis page becomes available at that symbol.

Options & Greeks
  • Black–Scholes priceComputed

    European-style call/put price under the standard B-S assumptions: log-normal terminal price, constant volatility, no dividends in the basic form.

  • DeltaComputed

    Sensitivity of option price to a $1 move in the underlying.

  • GammaComputed

    Rate of change of delta.

  • VegaComputed

    Sensitivity to a 1-percentage-point change in implied vol.

  • ThetaComputed

    Daily time decay.

  • RhoComputed

    Sensitivity to a 1-percentage-point change in interest rates.

Correlation & beta-to-benchmark
  • Pearson correlationComputed

    Cov(x, y) / (σ_x × σ_y) on aligned daily returns. Always in [−1, 1].

  • Beta vs. benchmarkComputed

    Cov(stock, benchmark) / Var(benchmark) on log returns.

  • R-squaredComputed

    Share of the stock's return variance explained by the benchmark.

  • Tracking errorComputed

    Standard deviation of the difference between stock and benchmark returns, annualized.

Composite scores

Internal aggregations that sit on top of the underlying metrics. The headline number is for at-a-glance comparison; every input is shown in its own section.

  • Growth scoreComputed

    Weighted blend of momentum (returns and trend strength), Monte Carlo upside probability, ML classifier probability, and sentiment. Reported on a 0–100 scale.

  • Scorecard sub-scores (Quality, Growth, Valuation, Momentum)Computed

    Each sub-score normalizes the underlying fundamentals or technicals into a 0–100 band so they can be compared. The overall card is the average of the four.