Binance futures historical data: what the free archive has, and when to use Aperiodic

What the archive has for USDⓈ-M futures

These are the file types data.binance.vision lists under USDⓈ-M futures (monthly, daily, listed November 5, 2026). Binance's README documents only trades, aggregate trades and klines; the descriptions of the others below come from opening the files.

Each file is a zipped CSV for one symbol and one day or month, with a .CHECKSUM file to verify it. Daily files arrive the next day and monthly files on the first Monday of the month (README, checked November 5, 2026).

Where the archive wins

• It is free, for every symbol and year. No account, no key. Our free plan covers 2025 on BTC, ETH, SOL and 10 RWA perps only.
• The raw prints. Every trade, every best bid and ask, and book depth snapshots. On May 1, 2025, the BTCUSDT bookDepth file held 2,877 snapshots, about one every 30 seconds. We sell bars, not prints.
• Positioning ratios. The daily metrics files carry top-trader and taker long/short ratios every 5 minutes. We have no long/short ratios.
• Spot, coin-margined futures and options. The archive has folders for all three; on Binance we cover USDⓈ-M perps.

For personal research on Binance alone, start there. The rest of this page is about what happens after the download, and about the licence.

Where Aperiodic is the better alternative

Commercial use is allowed. Binance's dataset terms (version 1.0, updated August 26, 2026) provide the data under CC BY-NC-SA 4.0 and say that “any commercial utilization requires a separate, written enterprise data license agreement executed with Binance” (terms, checked November 5, 2026). Personal, non-production backtesting is a permitted use. A fund or a company doing research for its own trading needs that separate licence, or a vendor whose terms cover it: our Prime Commercial plan carries a commercial licence for internal business use (licensing).

The rate between settlements, too. The archive's monthly funding file for BTCUSDT, May 2025, has 93 rows, one per 8-hour settlement. Here is the same month from Aperiodic, pulled with no API key:

# Install the Aperiodic CLI into ~/.local/bin
mkdir -p "$HOME/.local/bin"
curl -fsSL https://raw.githubusercontent.com/aperiodic-io/cli/main/install.sh | INSTALL_DIR="$HOME/.local/bin" bash
export PATH="$HOME/.local/bin:$PATH"

aperiodic funding --preview --exchange binance-futures --symbol perpetual-BTC-USDT:USDT \
  --interval 5m --timestamp exchange --start-date 2025-05-01 --end-date 2025-05-31 \
  --output-dir ./data && echo "Saved to $PWD/data"

One parquet file: 8,928 five-minute bars, each with the last funding rate observed in the bar and how often it updated. In the median bar the rate changed every 75 seconds (funding).

# pip install pandas pyarrow
import pandas as pd
df = pd.read_parquet("data/2025-05.parquet")
print(len(df))  # 8928 five-minute bars
print(df["funding_rate_avg_update_interval"].median())  # 75.0 seconds between updates

Or have your agent run it and read the file for you:

Read https://aperiodic.io/start.md and follow it. Then pull the free `funding` preview and show me how the funding rate moved between settlements.

Bars on the clock you trade on. The archive's files carry Binance's own timestamps: when the exchange matched a trade or took a snapshot. Nothing between the exchange and the file records when that event would have reached you. Aperiodic records both, at the collection point closest to each exchange, and builds every metric on either clock. With timestamp="true" a bar holds only what had arrived by its close, so a backtest cannot act on a print a live strategy had not seen yet.

Files that stay the same. Binance's README says archived files “may be updated at a later date as a result of recently discovered issues”, and its terms let it modify or remove any dataset without notice (README, terms). Aperiodic metrics are point-in-time: a published bar is not recalculated, so a backtest run next year reads the same history.

The metrics, and three venues in one schema. From the archive you would build taker flow by trade size, slippage against the quote, imbalance and price impact yourself, for each symbol. Aperiodic ships 220 of those columns across 19 datasets (catalog), for Binance futures, OKX perps and Hyperliquid perps under the same symbols, so a backtest moves between venues without a new parser.

What it costs. Aperiodic is free for every metric over 2025 on BTC, ETH, SOL and 10 RWA perps, on all three venues. Aperiodic Pro is $199 a month billed yearly ($249 month to month) for the trade and derivatives datasets with 5 years of history. The order book datasets are on Prime, at $399 a month billed yearly, and Prime Commercial ($799 billed yearly) adds the commercial licence (pricing). Every other vendor's entry price is on crypto market data pricing, vendor by vendor.

What every plan gives a backtest

• Every perpetual on Binance, OKX and Hyperliquid stays in the data after it is delisted, so a backtest sees the coins that were listed at the time, including the ones that later died.
• Published bars are never recalculated. A backtest you run next year reads the same history as one you run today.
• Every metric comes on exchange time or on local time (when the data reached us). On local time, a bar only holds what you could have seen by its close.
• Prime and above have no API rate limits, and Pro has high ones. The Python SDK downloads 10 files at a time by default.

Side by side

Archive listing, sample files, README and terms checked November 5, 2026 at data.binance.vision and binance-public-data. Binance can change them without notice.

From archive files to Aperiodic datasets

Which one to use

Use Binance’s archive for personal, non-commercial research on Binance alone, when you are happy to build and maintain the features yourself.

Use Aperiodic when the research is for a business, when you want flow, slippage, imbalance and impact as columns, or when the same backtest has to run on OKX and Hyperliquid too.