Skip to workspace

SecondOrder Stock [Under Construction]

v0.26

Start here

1. Real daily market data loads automatically. 2. Start with the latest 30 trading days. 3. Run all eight models and read the observations below.

Loading real market data… / 正在读取真实行情……

What data do all eight models need?

Use completed daily open, high, low, close and volume, preferably with consistent adjustment. Models 3, 4 and 8 need 3–12 assets covering the same dates. The default loads eight assets over three years, about 750 bars each.

For a latest 30-day selection, allow at least 300 aligned bars through its end; for 120 days, allow 660. This covers three separated references at the longest historical horizon. Constant returns, ambiguous matches or instability can still limit results.

Trading days exclude weekends and market closures. Comparison assets are other stocks and funds used to measure relationships; this is not your portfolio.

MARKET WORKSPACE

NVDA NVIDIA

Chart view only

Drag or scroll to browse. Ctrl + scroll to zoom. Choose Select period or hold Shift to select analysis dates.

SELECT A PERIOD

What is happening in this period?

FromTo

The dates above control every model. The shaded chart area shows this selection. Choose 10–120 trading days.

Choose individual models (optional)

READ THE CANDLE

Daily OHLC does not reveal the order of the high and low.

EIGHT MODELS

Eight views of the selected period

01–08 identify the models; they are not rankings or scores.

08 / HOMOLOGY

Homology historical comparison

Which earlier periods had similar relationships between assets?

Explanation

Historical outcome horizon

How to read Homology

This compares relationships between assets, not just the selected stock’s price shape. Each future horizon uses earlier periods whose later outcomes are already complete before your selection. Matching stability measures whether references remain similar after small changes to the input data. Low stability means caution. Changing the horizon changes the historical group; it does not forecast your selected stock.

Method

Robust linear and rank correlation distances feed exact Rips H₀/H₁ diagrams. Diagrams lose asset identity. Each outcome horizon has its own mature cohort. References are separated by the window plus horizon. Sensitivity uses eight joint block resamples against a fixed shortlist of at most 30 candidates per horizon.

How to read the eight models

Read the observation first, then the chart. Descriptive result describes these data. Other statuses explain small samples, instability or missing information, with a relevant next step. Method limitations are shown separately. These labels are not buy or sell signals. Model 8 alone searches similar historical patterns.

New to stocks? Read the numbers

O = open, H = high, L = low, C = close, V = volume. Return = price change in percent. Volume = shares traded. Volatility = how much daily returns vary, not expected profit. Drawdown = a fall from a previous peak. Correlation ranges from −1 (opposite moves) through 0 (little linear relationship) to +1 (same direction). Median = the middle observation. N/A means the data do not support this value. Unadjusted prices have not accounted for share splits: a company can divide each share into several shares, changing the quoted price without an equivalent loss in value.

These views share data. Multiscale state summarizes the same observations; it is not an independent vote.

Advanced checks (optional)

VALIDATION

Check what the comparison adds

Homology alone receives a walk-forward return comparison. Current-analysis models are checked for calculation correctness, stability, coverage and future-data isolation.

Window30 trading days

Horizon20 trading days

References3–6

Stride20 trading days

Interpretation

Candidate outcomes finish before the query begins. Queries with insufficient references or unstable local matching abstain. The baseline uses mature spaced past returns. MAE is average absolute return error; lower is better. pp means percentage points, so 5% versus 3% differs by 2 pp. Small samples and lower MAE do not establish trading value. The checks use the dataset currently loaded above.