Technical features · 142 cols feat.tech.pit
RSI, CCI, MACD, ATR, ADX, OBV, Bollinger, Donchian and 130+ more — all PIT.
PIT · Pro tier · 2003-01 → today · 12.8B rows
Coming soon — not served by the API yet.
Methodology
Every feature on a row is computed from prices observable at the close of that row's effective_ts and nothing later. That sounds obvious and is the usual place a feature table fails: a rolling window given a centred alignment, a scaling step fitted over the whole sample, or a default fill that reaches forward all leak. Each row carries the parameter set it was built with, so a result produced last quarter can be reproduced after a default changes rather than silently recomputed. The table is wide; the columns above are a representative subset and the full 142-column manifest ships with the package.
Schema
| Column | Type | Role |
|---|---|---|
as_of_ts | timestamp | knowledge time · UTC |
effective_ts | date | PIT key · trading date |
symbol | string | ticker in force on effective_ts |
feature_version | string | parameter set identity |
rsi_14 | float64 | 14-period RSI |
cci_20 | float64 | 20-period CCI |
macd_hist | float64 | MACD histogram, 12/26/9 |
atr_14 | float64 | 14-period ATR |
adx_14 | float64 | 14-period ADX |
obv | float64 | on-balance volume |
bb_pctb | float64 | Bollinger %B, 20-period |
donchian_pos | float64 | position in 20-day channel |
Use cases
- Feed a cross-sectional model without hand-rolling indicators.
- Reproduce a backtest after an indicator default changes.
- Audit a feature for forward-looking alignment.
Background
- Look-ahead bias in quantitative trading: six mechanisms and their fixes
- Point-in-time market data: as-of dates, restatements, and adjustments
Related packages
- feat.labels.pit — Labels · forward returns / triple-barrier