Version 1.6 · Valid from 22 April 2025

Download PDF here.

1. Introduction

Blocksize produces the Real-Time VWAP data feed as part of its crypto market data offering. All data points are converted and delivered in U.S. Dollars, Euros and other currencies across an expanding set of cryptocurrencies and digital asset classes.

The Real-Time VWAP feed is designed to be a transparent, independent pricing source that supports efficient markets, reduces information asymmetry between participants, and helps establish crypto as an asset class held to the highest standards.

It is used across financial market workflows, including:

• Pricing securities — determining fair value so transactions occur at appropriate levels

• Trade execution — executing with precision and minimising execution cost

• Algorithmic trading — supplying the input complex strategies require

• High-frequency trading — enabling rapid, high-volume execution

• Risk management — giving institutions insight into current conditions and volatility

• Research and analysis — supporting trend identification and strategy development

• Market surveillance — helping authorities detect manipulation and protect market integrity

• News analytics — measuring the market impact of news events

The feed is calculated using a transparent, robust methodology governed by the Blocksize Data Committee, with a clear framework of policies and procedures aligned to international best practices.

2. Calculation of Volume-Weighted Average Prices

The Real-Time VWAP is calculated as follows:

1. For each accepted market, calculate the volume-weighted sum of prices — Σⱼ (pⱼ × vⱼ) — from observable transactions, denominated in units of the given asset.

2. Calculate the aggregate volume across all accepted markets — Σⱼ vⱼ — referred to as the volume weight.

3. Divide the volume-weighted sum of prices by the total volume weight to obtain the VWAP:

P(VWAP) = Σⱼ (pⱼ × vⱼ) / Σⱼ vⱼ

3. Data Contingency Rules

These rules address situations where data is delayed, missing or unavailable — during low liquidity, extraordinary market conditions, or factors beyond Blocksize's control.

1. If transactions from an accepted market can't be collected due to a technical problem at that exchange (such as malformed data), those transactions are excluded from that instance of the feed.

2. If no transactions from an accepted market exist in the current timeframe, the feed relies on the other accepted markets for its calculation.

3. If none of the accepted exchanges report observable transactions, the feed does not report a newly calculated price; in that case the price is computed from the last observed transactions.

4. Data Quality Assurance

All observable transactions from accepted markets are evaluated using algorithmic anomaly and outlier detection. Suspected errors or anomalies are excluded from the calculation. Detection is applied within liquidity-adjusted timeframes of up to 120 minutes.

Alongside the automated methods, the Blocksize Data Committee periodically reviews whether data sources should be added or removed to maintain quality. New sources are vetted for integrity before inclusion; sources that begin delivering anomalous data may be removed. Any change is approved by the Committee and communicated to users where relevant.

4.1 Overview

Blocksize uses statistical methods to maintain VWAP quality automatically, applying liquidity-adjusted timeframes so that confidence intervals and computed VWAPs stay accurate and current. Two filters are used: a Price-Based Filter for individual anomalous events, and an Exchange-Based Filter for data sources delivering low-quality data. The Price-Based Filter is applied first, then the Exchange-Based Filter.

4.2 Adaptive Liquidity-Adjusted Timeframes

Timeframes adapt to the number and volume of trade events observed for each pair, aiming to satisfy four conditions at all times:

1. At least 10 events in the window

2. Aggregate volume of at least USD 1,000 (or equivalent in the quote currency)

3. All included events are less than 120 minutes old

4. At most 1,000 trade events in the window

Conditions 1 and 2 support accurate statistical estimates for the VWAP and anomaly detection. Conditions 3 and 4 exclude data that isn't sufficiently recent. In rare cases the recency limits mean conditions 1 and 2 can't be met — handled as an edge case in Section 4.5.

4.3 Price-Based Filter

For each new trade event, the filter measures how far its price deviates from the mean price of all other non-anomalous events in the liquidity-adjusted window, using the average and standard deviation of prices in that window.

Given prices already in the window, the confidence interval for including a new price is the average plus or minus a multiple (ξ) of the standard deviation:

Ψ = [ p̄ − ξσ , p̄ + ξσ ]

If the new price falls within the interval it is accepted; otherwise it is discarded as anomalous. The constant ξ sets how large a deviation counts as anomalous — Blocksize uses ξ = 3.5 to maintain high quality standards.

4.4 Exchange-Based Filter

For each data source, Blocksize compares its VWAP against the VWAP of all other sources for the same feed. The confidence interval for classifying a source as anomalous is set at two standard deviations of the combined sample excluding that source.

The VWAP excluding an exchange e′ is:

P̄(VWAP, e′) = Σ pₑ(t)vₑ(t) / Σ vₑ(t) — summed over all exchanges except e′

The distance between an exchange's own VWAP and the VWAP excluding it:

dₑ = P(VWAP, e) − P̄(VWAP, e)

The test τₑ returns +1 if dₑ falls within [ −t·σ̄ , +t·σ̄ ] and −1 otherwise. A positive result means the exchange is not an outlier; a negative result means it is, and its data is excluded per the contingency rules in Section 3.

4.5 Edge Cases and Applicability

The two filters only apply when their confidence intervals are reliable. Blocksize found empirically that meeting Condition 1 (≥ 10 events) and Condition 2 (> USD 1,000 volume) is what makes the Price-Based and Exchange-Based methods perform.

• If those conditions aren't met, conditions are treated as anomalous, the filters are not applied, and all non-erroneous events are included in the VWAP.

• If the Exchange-Based Filter flags multiple sources in one timeframe as outliers, it isn't applied — wide price spreads across exchanges can occur in abnormal conditions. All events not excluded by the Price-Based Filter are used.

• The Exchange-Based Filter only applies when three or more exchanges report data. With fewer than three, the VWAP is computed from all trades not excluded by the Price-Based Filter.

• After an abrupt price jump, a sequence of genuine trades can be wrongly flagged as anomalous. A price-jump detector handles this: if 4 or more outliers occur in sequence, above or below the interval, with ≥ 500 USD (or equivalent) of aggregated volume, the confidence interval is re-initialised to include that sequence.