cloudmaxi0x.com

Deterministic OHLCV for an exact half-open range, with a SHA-256 Merkle manifest. Only closed candles are returned - a forming candle would hand you lookahead. Exchange decimal strings are preserved byte for byte, so the same query always yields the same root. Gaps and exchange revisions are reported rather than silently applied.

Merkle audit path proving one candle belongs to a range root you already hold, without re-fetching the range.

Fibonacci retracements and extensions from deterministically detected swings, with the base rate attached. Every level carries the bar at which its swing was CONFIRMED - a swing high is defined by the bars either side of it, so the level does not exist for anyone until eight bars later, and touches before that are excluded. Each ratio ships with what price did after touching it against what price did after touching a random level in the same swing.

Harmonic XABCD patterns - Gartley, Bat, Butterfly, Crab, Shark and Cypher - with the base rate AND the sensitivity to the settings that produce them. The same candles yield an order-of-magnitude different count depending on the ratio tolerance and whether pivots may be skipped; both are parameters here and every response shows the count across a grid of them. Outcomes are scored against each shape's own claimed direction, against a random entry with the same bullish/bearish mix.

Average Directional Index with +DI and -DI. Warm-up history is fetched from BEFORE the requested range and discarded, so the first value returned is already converged - without it the same dates computed from a longer range come back different. Values are published at the precision that actually reproduces.

Average True Range, Wilder smoothing. Warm-up history is fetched from BEFORE the requested range and discarded, so the first value returned is already converged - without it the same dates computed from a longer range come back different. Values are published at the precision that actually reproduces.

Bollinger bands, 2 standard deviations. Warm-up history is fetched from BEFORE the requested range and discarded, so the first value returned is already converged - without it the same dates computed from a longer range come back different. Values are published at the precision that actually reproduces.

Exponential moving average, SMA-seeded. Warm-up history is fetched from BEFORE the requested range and discarded, so the first value returned is already converged - without it the same dates computed from a longer range come back different. Values are published at the precision that actually reproduces.

MACD 12/26/9 with signal and histogram. Warm-up history is fetched from BEFORE the requested range and discarded, so the first value returned is already converged - without it the same dates computed from a longer range come back different. Values are published at the precision that actually reproduces.

Relative Strength Index, Wilder smoothing. Warm-up history is fetched from BEFORE the requested range and discarded, so the first value returned is already converged - without it the same dates computed from a longer range come back different. Values are published at the precision that actually reproduces.

Simple moving average of the close. Warm-up history is fetched from BEFORE the requested range and discarded, so the first value returned is already converged - without it the same dates computed from a longer range come back different. Values are published at the precision that actually reproduces.

Volume-weighted average price, anchored to the first bar of the requested range. Warm-up history is fetched from BEFORE the requested range and discarded, so the first value returned is already converged - without it the same dates computed from a longer range come back different. Values are published at the precision that actually reproduces.

Support and resistance levels with touch counts. Touches are distinct APPROACHES into the band, not bars inside it, so a long consolidation counts once rather than turning into the most-touched level on the chart.

Candlestick and wick patterns with the base rate attached. Returns every occurrence with exact candle references and the ratios behind it - and, for each pattern, what actually happened afterwards versus what happened after a RANDOM entry in the same range over the same horizon. Where the two match, the response says so: pattern detection is structure, not signal. Includes a multiple-testing correction, because scanning 23 patterns produces hits by luck alone.

Replay an order against what price actually did. Reports which of a stop and target came first, the excursion split at the exit - so a stop that fired early is visible - and refuses to guess when both fall inside one bar, because the bar carries no ordering.

Tier 2 filter audit. Strict rolling-origin walk-forward measured against a permutation null built from the same pipeline on scrambled outcomes, plus lookahead, tautology and data-integrity checks.

Tier 1 backtest audit. Send closed-trade returns in R and get a verdict on whether the result can be told apart from luck: bootstrap CI, statistical power, deflated Sharpe for the number of variants you tried, outlier dependence, regime stability, break-even cost.