Order flow

Built from every trade on Binance, OKX and Hyperliquid perps: who crossed the spread, what it cost, and how the price moved.

8 datasets

Taker buys against sells, by size, with toxicity

Net taker flow moved with BTC's price in the same hour (correlation 0.72) and said almost nothing about the next hour (0.02).

Kyle's lambda, Amihud, and what size costs

A BTC of taker flow moved the price 1.5× as much at 00:00 UTC as at 11:00.

What taker trades paid against the quote

BTC's p95 slippage was 4.3× higher on its busiest bars than on its quietest.

Small, medium and large trades, and the tails

On BTC, 36% of trades are $1,000 or more, and they carry 97% of the notional.

How long buyers and sellers keep at it

BTC bars where trades rarely changed side moved 2.8× as far as bars where they flipped often.

Trades that moved the price up, down, or not

65% of BTC trades printed at the same price as the trade before.

Realized vol, jumps and trendiness

BTC's hourly volatility correlated 0.39 with the hour before. Its hourly return correlated -0.02.

High-low range and the shape inside the bar

A BTC 5-minute bar ranged 1.5× as far on a weekday as at the weekend.

volume_delta_notional: Buy-side traded notional minus sell-side traded notional; Flow
flow_toxicity_score: Past-normalized signed notional flow multiplied by past-normalized short-horizon realized volatility; Flow
kyle_like_lambda: Covariance of trade returns and signed trade size divided by signed-size variance; Impact
slippage_bps_mean: Average slippage in basis points versus the prevailing quote; Slippage
large_order_percentage: Ratio of large-trade notional to total traded notional; Trade size
jump_ratio: Share of realized variance not explained by bipower variation; Returns & volatility

Charge each simulated trade what real trades paid: slippage against the quote, and Kyle's lambda and Amihud for size.

Volume delta, toxicity and the mix of trade sizes, bar by bar, without storing every trade.

Realized volatility, bipower variation and the jump ratio, from every trade in the bar.

Compare your fills with what every taker paid in the same bar, buys and sells apart.

Every field, with its endpoint and code, is in the Data Catalog.

For the background, start with the microstructure guide.