ta.js 2.1.0

Released 2026-10-10. Full details in CHANGELOG.md.

⚠️ Read this before upgrading

Four indicators return different numbers, and three of them return a different number of rows. These are corrections — each was verified against the implementation it was ported from, and each fix has a test that fails against the old code — but nothing raises, so an existing pipeline will consume the new values silently.

function	what changed
halftrend	row count: n - atrlen → n. One row per input bar, so row i now covers bar i. It also returned zero trend changes on real data.
alligator, gator	row count: n - jl - js + 2 → n. Rows are now oldest-first and aligned to the input bars.
kama	row count n - len1 → n - len1 + 1, and values: the smoothing constant had lost its slow term, so every value was wrong.
er, ar, kelly	values for one-sided input: a series with no losses (or no wins) returned NaN for er/kelly; it now returns the geometric mean of the side that exists, and a Kelly fraction of 1 / negative respectively.

If you index results by bar, or compare them against stored values, re-run before trusting them.

Added

• TypeScript declarations. index.d.ts types all 144 exports plus the aroon, random and multi namespaces, and package.json points at it with types. Named imports, a default import and require are all typed. Row layouts differ per function group and a wrong order still compiles, so each group documents the layout it expects (#21, #26).
• occ — the Open Close Cross. Moving averages of open and close on a chosen basis (SMA, EMA, DEMA, TEMA, WMA, SMMA, HullMA, TMA), delayed by a configurable number of bars, with the direction of each bar. Four of the Pine original's twelve basis types are absent: LSMA needs an offset parameter this library's lsma does not have, VWMA needs volume in the row, and ALMA and SSMA are not implemented here.

Fixed

• kama computed the wrong average entirely. The smoothing constant dropped its slow term (- 2/(len3+1) + 2/(len3+1) cancelled, leaving ER * fast); the volatility window summed len1-1 one-bar changes instead of len1, which let the efficiency ratio exceed 1 and divided by zero on flat input; the seed was the SMA of the last window rather than the first, so the opening value was computed from data at the end of the series; and the loop started one bar late. The recursion now matches tulind — the reference the benchmarks compare against — bit-identically on short, flat, trending and gapping series, and to 1.4e-14 on a 100k random walk. It is also ~2.5× faster, because the volatility sum is carried across bars instead of re-summed (#22, #27).
• halftrend never flipped. The flip tests read sma(...)[0] — the SMA at the start of the window, roughly atrlen bars old — instead of at the current bar, so both conditions compared stale values and never fired: on 4901 bars of 5-minute data it produced zero trend changes. The extremes also came from the ATR window instead of amplitude bars, and the down-flip compared close with < where the Pine uses >. It now follows everget's Pine line by line, and is 8-14× faster (#28).
• er and kelly on a one-sided series. A series with no losses evaluated the geometric mean of the losing side anyway, where 1 ** Infinity is NaN in JS. An empty side carries weight zero, so it now contributes nothing: kelly returns 1 for an all-winning series and -1.0276 for an all-losing one. Mixed series are unchanged to the last bit (#29).
• alligator / gator aligned to the input bars. Both returned their rows newest-first and truncated, so the output could not be lined up with the input at all, and each line was displaced by shift - 1 bars instead of shift. gator's upper histogram also had the wrong sign.
• BLOG_CONTEXT.md and .vscode/ are no longer shipped in the published tarball (#30).

Changed

• Bundles regenerated (ta.esm.js, ta.iife.min.js) and the comparative benchmarks refreshed on the current code.
• README and package metadata rewritten for discovery: description, keywords, and a benchmark summary.

Notes

• ta.min.js is a frozen legacy bundle, not a mirror of src: 22 of its own fixtures already disagree with src (including atr, rsi, stoch, smma) and 45 functions are absent. It is still referenced from the README's CDN section. Prefer ta.iife.min.js, which is generated from source and verified against it on every build.
• atr seeds Wilder smoothing from the first bar's high−low, where TradingView's ta.atr seeds with the mean of the first length true ranges. The difference is 5.5% at the first ATR bar and decays below float resolution after a few thousand bars, so it shifts supertrend, keltner and natr at the start of a chart only.