Datadory notebook
Daily aluminum OHLC data in Parquet, delivered as rows
Datadory delivers aluminum data covering twelve years of daily world-aluminum OHLC bars: 6,076 rows of open, high, low and close in US dollars per tonne from May 2014 to July 2026, keyed on Gregorian and Persian calendars alike, alongside a monthly LME benchmark reaching July 1996 and China's A00 ingot assessment for the demand side - typed rows delivered as Parquet, CSV or straight into your warehouse, daily, weekly, or hourly - your call.
1,744 datasets. Pick your catch.
What is a daily aluminum OHLC dataset?
The anchor record is Hugging Face - global_aluminum_price_daily dataset: 6,076 daily world-aluminum bars in US dollars per tonne, running Gregorian 2014-05-06 to 2026-07-28 - about twelve years of trading days. It ships the same series in two shapes. mart is the wide table: date_greg, a Persian Jalali period twin, and the four price columns open, high, low, close. observations is the long-format rendering, with provenance carried on every row (freq D, location_code WLD, unit usd_per_ton) and each day's range nested in a breakdown field. Each subset holds a single train split of roughly 3,038 rows.
World-level deserves stating plainly, because it is both the feature and the limit. One series stamped WLD means no contract-month or country bookkeeping to manage - and no exchange-level detail either. Nothing on the record documents which contract month or continuous-series definition sits upstream, so treat every bar as an index-level reference price rather than a specific LME contract. Datadory delivers both shapes as typed rows: the wide table feeds charts and feature pipelines, the long table feeds joins.
What do real rows look like?
Two consecutive bars from the opening of the window, captured during the August 2026 review:
date_greg period open high low close
2014-05-06 1393-02-16 2182.75 2205.75 2165.00 2172.75
2014-05-07 1393-02-17 2152.25 2152.25 2146.00 2149.00Three things in these rows tell you how the whole record behaves. First, every day is dated twice - 2014-05-06 and 1393-02-16 are one observation - so neither calendar's reporting cycle forces conversion logic on your side. Second, the first bar already carries a 40.75 high-low band on a 2,172.75 close, roughly 1.9% of intra-day travel that a close-only feed would silently delete. Third, the unit is declared in-row on the long subset (usd_per_ton), which is what lets dollars-per-tonne bars sit beside percent-change and native-unit series later without mixing.
What does OHLC give you that a monthly average does not?
A monthly average collapses each month into one number; OHLC preserves the intra-day range, and the range is where the useful work happens. Volatility estimation needs high-low spreads, stop-loss backtests need to know whether a level was actually traded through, and candle-based features need all four prints. On this record the ranges are material: the opening week alone moves from a 2,182.75 open to a 2,146.00 low.
The monthly complement still earns its place as the long anchor. IndexMundi Aluminum Monthly Price carries the LME quote for unalloyed primary ingot - minimum 99.7% purity - as three columns per month: Month, Price in US dollars per metric tonne, and the percentage Change. The retrievable window holds 357 rows reaching July 1996, roughly thirty years of history. Recent rows read January 2026 at 3,141.89, February 2026 at 3,065.19 (-2.44%), March 2026 at 3,372.95 (+10.04%) - single-month swings large enough that a settled monthly series shows the regime move without the intra-month static.
One seam runs through the splice. The monthly record documents that the LME settlement price applies from 2005 onward and the cash price before that, so deep-history rows are not methodologically uniform with recent ones. Pair them deliberately: daily bars since 2014, monthly anchors before, and the 2005 break flagged wherever the joined series crosses it.
Who uses daily aluminum OHLC data, and for what?
Quant and volatility modelers build realized-volatility estimates from the high-low spreads and train candle-based features on the full bar. The dual Gregorian and Jalali keys mean Iranian-calendar applications consume the period column directly, with no date arithmetic bolted on afterward.
Procurement and hedging analysts track the daily bars against contract windows, then reach for the monthly LME benchmark to sanity-check the level they are negotiating against - using the purity specification notes to make sure their grade maps to the right series.
Risk and strategy desks splice the pre-2014 monthly anchors onto the daily panel for decade-scale scenario work, flagging the 2005 settlement-versus-cash break, and validate recent levels against SMM metal.com - China aluminum spot and futures: the flagship A00 ingot assessment on an in-warehouse Shanghai/Wuxi basis, shown with SHFE and LME references. At the August 2026 review the A00 average read 3,513.04 USD/tonne, +14.52 - a demand-side print no Western world-level series provides.
Economists and trade analysts join the price bars to physical flows from USA Trade Online (US Census Bureau), which counts US imports and exports in HS chapter 76 by country, state and port, monthly from January 2002 - the way to ask whether a price move tracked metal actually moving.
Why get daily aluminum data through Datadory?
Because the hard part was never the existence of these numbers - it is keeping them coherent. The same market arrives as a wide mart table and a nested long table with different key styles, dated twice in two calendars, quoted world-index-level on one record and benchmark-level on another and assessment-basis on a third, with a specification seam sitting in the middle of the long anchor. Stitched by hand, that is four ways for a group-by to quietly split.
Datadory handles it so the rows just arrive: units kept beside amounts, both calendar keys preserved, both table shapes shipped when you want them, the 2005 methodology break flagged on the joined series, and vintages stamped so every backtest is reproducible. Delivered daily, weekly, or hourly - your call - into your warehouse, your notebook or your dashboard, in Parquet or whatever shape your stack speaks.
Where to go next
This page covered the daily aluminum OHLC dataset in Parquet specifically. Continue with:
- Hugging Face - global_aluminum_price_daily dataset - the full dataset record: sample rows, the complete field dictionary for both subsets, and coverage chips.
- IndexMundi Aluminum Monthly Price - the thirty-year monthly LME anchor, for extending history past the daily panel's 2014 start.
The aluminum data hub indexes every record with coverage, granularity and delivery options side by side.
Pick up where this leaves off
Every one of these ships with sample rows before you commit to anything.
Hugging Face - global_aluminum_price_daily dataset
date_greg · period · open …+9 more
IndexMundi Aluminum Monthly Price
SMM metal.com - China aluminum spot and futures
USA Trade Online (US Census Bureau)
1480000000 · 920000000 · 985000000 …+7 more
Want rows instead of a pitch? Name the datasets.
API, files, or your warehouse. Daily, weekly, or hourly.
Get a sampleQuestions worth asking
What does a daily aluminum OHLC series give you that a close-only feed does not?
Intra-day range. The very first bar in the record - 6 May 2014 - opens at 2,182.75, touches 2,205.75 and 2,165.00, and closes at 2,172.75 US dollars per tonne, a 40.75 band that a close-only series erases entirely. Realized-volatility estimates, stop-loss backtests and candle-based features all consume exactly that band.
How far back does daily aluminum price history go?
The anchor OHLC record reaches 2014-05-06, about twelve years of trading days ending 2026-07-28. Deeper daily coverage exists in the Trading Economics aluminum record, roughly 9,500 daily observations from October 1989 to August 2026 with quarter-end and twelve-month forecast targets on the same schema. For the long view, the IndexMundi monthly LME benchmark holds 357 rows reaching July 1996.
Does the daily aluminum series cover a specific exchange or country?
Neither. It is a single world-level series stamped location_code WLD at frequency D, and no contract month or continuous-series definition is documented anywhere on the record. Treat every bar as an index-level reference price rather than a specific LME contract, and cross-check levels against China's A00 assessment and the monthly LME benchmark before a model acts on them.
Why does the dataset carry Persian calendar dates?
Every observation is dated twice: 2014-05-06 and 1393-02-16 are the same bar, the second written in the Persian Jalali calendar. Applications reporting on an Iranian calendar consume the period column directly, with no date-conversion logic to maintain. Datadory keeps both keys on the delivered rows so neither calendar forces a transform on your side.