ExxaBlock “CME Bitcoin Reference Rate” Report
Aug. 27, 2020
Contents
- Background
- ExxaBlock Bitcoin Reference Rate
- BRR Summary
- Partition Summary
- Outlier Summary
- Slippage Summary
- Market Depth Summary
- EBBO Summary
- Unified Order Book Summary
- Latency Summary
Appendices
- Benchmark Market Prices
- Prices and Volumes on Exchanges
B.1 CoinbasePro
B.2 Kraken
B.3 Gemini
B.4 Bitstamp
B.6 itBit
- Cryptocurrency Exchanges
- Outlier Detection Methodology
D.1 Z-score Method
D.2 IQR Method
Slippage
E.1 Slippage as a price difference
E.2 Slippage as a ratio
E.3 Slippage measures at the bid-side order book
E.4 Slippage measures at the ask-side order book
Market Depth
F.1 Multiple market depths at the bid-side order book
F.2 Multiple market depths at the ask-side order book
API and Database Access
List of Figures
3.1 Trade Prices and Volumes on BRR Constituent Exchanges: All trade prices and volumes between 15:00 p.m. and 16:00 p.m. on Coinbase Pro, Gemini, Kraken, Bitstamp, and itBit are shown here. time in the X-axis is ExxaBlock timestamp, not exchange timestamp. Those exchanges produce 3,107.73773844 BTC/USD trade volume with a value of 35,193,335.46 US Dollars. The volume is generated by 11,255 trades on 4,222 unique price levels during the last one-hour BRR calculation period.
4.1 Market Benchmark Prices on BRR Partitions: There is big difference among VWMP, VWAP, and TWAP on partitions. The means of VWMP, VWAP, and TWAP are 11,323.89, 11,323.01, and 11,321.71, respectively. ∆p is a difference between the highest and lowest prices in VWMP, VWAP, and TWAP.
4.2 OHLC and Price movements on BRR Partitions: Each vertical box is the range of open and close trade prices of a partition. Box color is red if the open price is higher than the close prices. Otherwise, the color is blue. The line above the top of the box extends to the highest price in the partition. the
line below the bottom of the box extends to the lowest price in the partition. ∆OC is the difference between the open and close prices of the partition in a percentage. ∆HL is the difference between the high and low prices of the partition in a percentage.
6.1 Slippage at the Bid-Side Order Book over Small($10,000 and $25,000) and Medium($50,000 and $100,000) Market Order Sizes on BRR Partitions: Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), and itBit().
6.2 Slippage at the Bid-Side Order Book over Large Market Order sizes ($250,000 and $500,000 on BRR Partitions: Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), and itBit().
6.3 Slippage Depths at the Bid-Side Order Book over Various Market Order Sizes on the first and last BRR Partitions: Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), and itBit().
6.4 Slippage at the Ask-Side Order Book over Small($10,000 and $25,000) and Medium($50,000 and $100,000) Market Order Sizes on BRR Partitions: Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦),and itBit().
6.5 Slippage at the Ask-Side Order Book over Large Market Order sizes ($250,000 and $500,000 on BRR Partitions: Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), and itBit().
6.6 Slippage Depths at the Ask-Side Order Book over Various Market Order Sizes on the first and last BRR Partitions: Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), and itBit().
7.1 Cumulative Volume available at the Bid-Side Order Book for Shallow Market Depths 0.1%, 0.2%, 0.3%, and 0,4% on BRR Partitions: Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), and itBit().
7.2 Cumulative Volume available at the Bid-Side Order Book for Medium and Deep Market Depths 1.0%, 2.5%, 5.0%, and 10.0% on BRR Partitions. Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), anditBit().
7.3 Cumulative Volume available at the Ask-Side Order Book for Shallow Market Depths 0.1%, 0.2%, 0.3%, and 0,4% on BRR Partitions: Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), and itBit().
7.4 Cumulative Volume available at the Ask-Side Order Book for Medium and Deep Market Depths 1.0%, 2.5%, 5.0%, and 10.0% on BRR Partitions. Coinbase Pro(x), Kraken(), Gemini(•), Bitstamp(◦), anditBit().
9.1 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 1: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,364.30. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,360.60. The spread is −7.40. Bid and ask prices are crossed between 11,360.6 and 11,368.
9.2 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 2: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,283.25. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,287.59. The spread is −10.30. Bid and ask prices are crossed between 11,278.1 and 11,288.4
9.3 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 3: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,283.64. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,287.39. The spread is −7.73. Bid and ask prices are crossed between 11,279.77 and 11,287.5.
9.4 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 4: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,328.40. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,331.15. The spread is −6.80. Bid and ask prices are crossed between 11,325and11,331.8.
9.5 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 5: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,319.57. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,323.29. The spread is −10.67. Bid and ask prices are crossed between 11,314.23 and 11,324.9.
9.6 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 6: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,333.88. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,335.32. The spread is −3.25. Bid and ask prices are crossed between 11,332.25 and 11,335.5.
9.7 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 7: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,338.28. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,336.49. The spread is −4.57. Bid and ask prices are crossed between 11,336 and 11,340.57.
9.8 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 8: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,332.45. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,333.98. The spread is −3.10. Bid and ask prices are crossed between 11,330.9 and 11,334.
9.9 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 9: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,317.89. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,312.62. The spread is −12.03. Bid and ask prices are crossed between 11,311.87 and 11,323.9.
9.10 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 10: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,310.36. The micro price, which is
the volume-weighted price on best bid and ask on the order book, is 11,313.09. The spread is −5.78. Bid and ask prices are crossed between 11,307.47 and 11,313.25.
9.11 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 11: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,313.42. The micro price, which is
the volume-weighted price on best bid and ask on the order book, is 11,314.51. The spread is −4.65. Bid and ask prices are crossed between 11,311.1 and 11,315.75.
9.12 Unified Order Book Snapshot at the Last Data Collection Time on BRR Partition 12: All order books at BRR constituent exchanges are aggregated into the order book. The left blue and right red area are bid- and ask-side of the order book. The mid price is 11,315.05. The micro price, which is the volume-weighted price on best bid and ask on the order book, is 11,315.01. The spread is −4.10. Bid and ask prices are crossed between 11,313 and 11,317.1.
10.1 Latency Test, Measurement, and Daily Report Generation: latencies between 3 different ExxaBlock test sites(T1, T2, and T3) and exchanges (EX1..n) are measured at 15:00 p.m. and 16:00 p.m. when BRR calculation starts and ends, respectively (mt). All test results (J1t, J2t, and J3t) are copied into a main test site (T1). An individual test measures the difference between two time points — (1) time just after a “ping” message(pn) is sent to a WebSocket market data server and (2) time just after receiving a “pong” message(pn) from the server
B.1 Trade Prices and Volumes on BRR Constituent Exchange Coinbase Pro during the BRR Hour: All trade prices and volumes are captured at ExxaBlock data center between 15:00 p.m. and 16:00 p.m. The exchange produces 1,270.10605962 trade volume with a value of 14,380,367.03 US Dollars. The volume is generated by 6,265 trades on 2,681 unique price levels during the hour.
B.2 Trade Prices and Volumes on BRR Constituent Exchange Kraken during the BRR Hour: All trade prices and volumes are captured at ExxaBlock data center between 15:00 p.m. and 16:00 p.m.. The exchange produces 503.58125064 trade volume with a value of 5,702,174.49 US Dollars. The volume is generated by 1,586 trades on 479 unique price levels during the hour.
B.3 Trade Prices and Volumes on BRR Constituent Exchange Gemini during the BRR Hour: All trade prices and volumes are captured at ExxaBlock data center between 15:00 p.m. and 16:00 p.m.. The exchange produces 84.51852312 trade volume with a value of 956,285.67 US Dollars. The volume is generatedby479tradeson359uniquepricelevelsduringthehour..
B.4 Trade Prices and Volumes on BRR Constituent Exchange Bitstamp during the BRR Hour: All trade prices and volumes are captured at ExxaBlock data center between 15:00 p.m. and 16:00 p.m. The exchange produces 1,228.52506211 trade volume with a value of 13,917,040.51 US Dollars. The volume is generated by 2,413 trades on 1,310 unique price levels during the hour.
B.5 Trade Prices and Volumes on BRR Constituent Exchange itBit during the BRR Hour: All trade prices and volumes are captured at ExxaBlock data center between 15:00 p.m. and 16:00 p.m.. The exchange produces 21.00684295 trade volume with a value of 237,467.75 US Dollars. The volume is generated by 512 trades on 237 unique price levels during the hour..
D.1 Trade Volume Outlier Detection with Z-scores 3 (Z3): All unique trade volumes that are away from the mean (μ) of their distribution (X) more than ±3 standard deviation (σ) are classified into outliers. Only 0.27% or less of unique volumes are considered to outliers in Z3
D.2 Trade Volume Outlier Detection with IQR 1.5 (IQR1.5): m is an median of volume data(X). Q1 is the median between Xmin and m. Q3 is the median between m and Xmax. IQR is Q3 — Q1. IQR1.5 is the aggregation of all trade volumes that are located at lower than Q1-1.5*IQR and higher than Q3+1.5*IQR
List of Tables
3.1 BRR and Trade Price Summary: EBRR is a daily CME Bitcoin Reference Rate that is independently calculated by ExxaBlock. The TWAP is higher than other benchmark prices. The VWMP is lower than the others. There is 28.07(0.25%) price difference between TWAP and VWMP. Trade prices moves -0.76% between the first(O) and last(C) trades of the one-hour period. The prices swing 1.48% between the highest(H) and lowest(L) prices. There is 5.56 difference between BRR and VWMP. It is mainly because the former is computed by averaging out VWMPs of 12 5-minute partitions while the later is a single VWMP over the BRR one-hour period.
3.2 Statistical Summary on Aggregated Trade Prices and Volumes of BRR Constituent Exchanges: N is a number of unique trade prices. μ and σ are the mean and standard deviation of a distribution (a.k.a. standard measures). m and σ are the median and median absolute deviation (MAD) (a.k.a. robust measures). Z3 and IQR 1.5 are outlier detection methods.
3.3 Benchmark Prices and Price Movements on BRR Constituent Exchanges: VWMP has more price discrepancies than other benchmark prices between exchanges. itBit trade prices move more than other exchanges between the first(O) and last(C) trade prices of the last one-hour BRR period. Bitstamp trade prices swing wider than other exchanges between highest(H) and lowest(L) trade pricesduringtheperiod.
3.4 Statistical Summary on Trade Prices on BRR Constituent Exchanges: Kraken and Bitstamp provide the lowest and highest average trade prices. Coinbase Pro generates a significantly more trade asset value than other exchanges. 69.48% of execution prices are within the 1 standard deviation from the mean of unique trade price levels in most exchanges. L is a number of unique trade price levels. A is an asset values in USD generated by trades. μ and σ are the mean and standard deviation of a trade price distribution. m and σ are the median and median absolute deviation of the distribution.
3.5 Statistical Summary on Aggregated Trade Volumes on BRR Constituent Exchanges: The majority of trades are executed at Coinbase Pro. Executed trade size per order is the smallest at itBit and the largest at Kraken. Coinbase Pro is the most liquid market place and Kraken is the least liquid marketplace. N is a number of trades at an exchange. L is a number of volumes that are aggregated by each price level. V is the total volume that an exchange produces. μ and σ are the mean and standard deviation of a aggregated trade volume distribution. m and σ are the median and median absolute deviation of the distribution.
4.1 Market Benchmark and Trade Prices Summary on BRR Partitions
4.2 Statistical Summary on Aggregated Trade Volumes and Prices on BRR Partitions: The largest and smallest trade volumes are generated in the partition 2 and 9, respectively. Skewness(s) and Excess Kurtosis(κ3) of aggregated trade volumes show that trade volume distributions in most partitions are highly skewed to the left and have heavy/fat tails. Most trade price distributions in BRR partitions are close to the normal distribution and have light/thin tails (i.e., Platykurtic).
4.3 Exchange Trade Volumes on BRR Partitions: Coinbase Pro and itBit generate the largest and smallest trade volumes in the partitions. Σe is a sum of all exchange trade volumes in a partition. ∆e is the volume disparity between two exchanges that produce the largest and smallest volumes in the partition. Σp is a sum of all values in a column in this table. μp and σp are the mean and standard deviation of those values in the table.
4.4 Exchange Trade Values on BRR Partitions: Coinbase Pro and itBit generate the largest and smallest trade values in the partitions. Σe is a sum of all exchange trade values in a partition. ∆e is the price disparity between two exchanges that produce the largest and smallest trade values in the partition. Σp is a sum of all values in a column in this table. μp and σp are the mean and standard deviation ofthosevaluesinthetable.
5.1 Exchange Volume Outliers: VN is a number of aggregated trade volumes per price of each exchange. The number is equal to a number of unique trade prices at the exchange over BRR one-hour time period. VT is the trade volume generated by all trades executed at the exchange. ZN is a number of outliers found by Z3. Zmn and Zmx are the smallest and largest outliers. IN , Imn, and Imx are a number of outliers, outlier volume, min outlier, and max outlier found by IQR1.5.
5.2 Volume Outliers on BRR Partitions: VN is a number of aggregated trade volumes in a partition. The number is equal to a number of unique trade price levels in the partition. VT is the trade volume generated by those trades. ZN is a number of outliers found by Z3 in the partition. Zmn and Zmx are the smallest and largest outliers in the partition. IN, Imn, and Imx are a number of outliers, outlier volume, min outlier, and max outlier found by IQR 1.5.
6.1 Slippage at the Bid-Side Order Book on Exchanges over Various Market Order Sizes on BRR Partitions: For a given market order size <N>*k(e.g., $10,000), μp is the mean of slippages of all exchanges in a partition. μ<N>k is the mean of all slippages of an exchange over BRR partitions. Slippage is shown as a basis point(bps).
6.2 Slippage at the Ask-Side Order Book on Exchanges over Various Market Order Sizes on BRR Partitions: For a given market order size <N>*k(e.g., $10,000), μp is the mean of slippages of all exchanges in a partition. μ<N>k is the mean of all slippages of an exchange over BRR partitions. Slippage is shown as a basis point(bps)
7.1 Market Depths at the Bid-Side of Exchange Order Books on BRR Partitions: Vs, Vm, and Vd are available (cumulative) market volumes across all BRR constituent exchanges for shallow, medium, and deep market depths, respectively. Each market depth is specified by the percentage (%) from a mid price. μpt is the mean of available market volumes per partition for each market depth size (i.e., acolumninthetable).
7.2 Market Depths at the Ask-Side of Exchange Order Books on BRR Partitions: Vs, Vm, and Vd are available (cumulative) market volumes across all BRR constituent exchanges for shallow, medium, and deep market depths, respectively. Each market depth is specified by the percentage (%) from a mid price. μpt is the mean of available market volumes per partition for each market depth size (i.e., acolumninthetable).
7.3 Market Depth Volume Imbalance: dpt is a market depth specified by the percentage(%) from a mid price. Vb is a mean of BRR partition bid market depth volume, which is μpt in Table 7.1. Va is a mean of BRR partition ask market depth volume, which is μpt in Table 7.2. V∆ is volume difference between Vb and Va. A positive V∆ means more volume in the bid side. Id is market depth volume imbalance, which is defined by 100*V∆/min(Vb,Va).
8.1 Exchange Best Bid and Ask (EBBO) on BRR One-Hour Period: coinbasepro supplies more best bids and asks than other exchanges. NB is a number of best bids or asks that are newly generated by exchanges. RP is a percentage of NB. VD is a value of the volume added by NB in USD.
8.2 Frequent Order Volume Sizes in Exchange Best Bid and Offer (EBBO): VB is a volume size generated by a Exchange Best Bid (EBB) or a Exchange Best Offer (EBO). VN is a number of times generated by an exchange.
8.3 EBB on BRR Partitions: Σp is a total number of EBB in a partition. Σe is a total number of EBB at an exchange overall partitions. μe is the mean of Σe.
8.4 EBO on BRR Partitions: Σp is a total number of EBO in a partition. Σe is a total number of EBO at an exchange overall partitions. μe is the mean of Σe.
8.5 Statistical Summary on EBB Prices on BRR One-Hour Period: Np is a number of best bids that an exchange provides. Mp and Xp are minimum and maximum best prices produced by the exchange. μp and σp are mean and standard deviation of all best bids on the exchange. Wσ is the price range between ±σp. Wnx is the price range between min and max best bid prices — Coinbase Pro, Kraken, Gemini,Bitstamp,anditBit.
8.6 Statistical Summary on EBO Prices on BRR One-Hour Period: Np is a number of best asks that an exchange provides. Mp and Xp are minimum and maximum best prices produced by the exchange. μp and σp are mean and standard deviation of all best asks on the exchange. Wσ is the price range between ±σp. Wnx is the price range between min and max best ask prices — Coinbase Pro, Kraken, Gemini,Bitstamp,anditBit.
8.7 Frequent Order Values in Exchange Best Bid and Offer (EBBO): AB is an asset value in USD generated by a Exchange Best Bid (EBB) or a Exchange Best Offer (EBO). AN is a number of times generated by an exchange.
8.8 Risk-Free Arbitrage Opportunities with EBBO on BRR Partitions: pt is a BRR partition. Na is a number of risk-free arbitrage opportunities in the partition. Xa is a sum of price differences between best bids and asks in the partition when the former at one exchange are higher than the later at other exchanges. Va is the total risk-free arbitrage volumes that are available in the partition. Aa is the asset values of total risk-free arbitrages in the partition. Aa is not equivalent to Xa·Va.
9.1 Extreme Bid Price Outliers on the Unified Order Book.
9.2 Extreme Ask Price Outliers on the Unified Order Book.
10.1 Latencies and Jitters of BRR Exchange Streaming Market Data Servers: exch is an exchange name. hr is an hour when latency measurement is started. aet, awt and iet are AWS us-east-1, AWS us-west-1 and IBM wdc 1 test sites.
E.1 Slippage Measures at the Bid-Side Order Book on BRR Constituent Exchanges over Multiple Market Order Sizes: exch is a BRR constituent exchange. pt is a BRR partition number. amt is a market order size executed on an exchange order book. sp_bps is a slippage in a basis point. sp_prc is
a slippage in an absolute price difference. sp_dpt is a number of bid price levels required for the execution of amt. sp_vol is a volume generated by liquidating amt. vwap is a vwap of executed orders for a mt. mp is a mid price on the top-of-the-book.
E.2 Slippage Measures at the Ask-Side Order Book on BRR Constituent Exchanges over Multiple Market Order Sizes: exch is a BRR constituent exchange. pt is a BRR partition number. amt is a market order size executed on an exchange order book. sp_bps is a slippage in a basis point. sp_prc is
a slippage in an absolute price difference. sp_dpt is a number of ask price levels required for the execution of amt. sp_vol is a volume generated by liquidating amt. vwap is a vwap of executed orders for a mt. mp is a mid price on the top-of-the-book.
F.1 Multiple Market Depths at the Bid-Side Order Book on BRR Constituent Exchanges: exch is a BRR constituent exchange. pt is a BRR partition. dpt is a market depth in a percentage from a mid price. mp is the mid price. dpt_prc is a bid price that meets a market depth. dpt_vol is cumulative bid volumeuptodpt_prc. obk_lvl is a number of bid price levels required for dpt.
F.2 Multiple Market Depths at the Ask-Side Order Book on BRR Constituent Exchanges: exch is a BRR constituent exchange. pt is a BRR partition. dpt is a market depth in a percentage from a mid price. mp is the mid price. dpt_prc is a ask price that meets a market depth. dpt_vol is cumulative ask volume up to dpt_prc. obk_lvl is a number of ask price levels required for dpt.
Background
Chicago Mercantile Exchange (CME)1 publishes the daily Bitcoin Reference Rate (BRR) at 4:01 p.m. London time Hereafter, all time are presented in London time). BRR determines the contract unit of CME Bitcoin Futures. A single Future contract is the value of five times the daily BRR. BRR is also used as the final settlement price at 4:00 p.m. on the expiration day of the Futures contract. The Futures are traded on CME Globex and cleared by CME. CME Bitcoin Options are based on the Futures.
BRR is linked to BTC/USD spot trading at five exchanges between 15:00 p.m. and 16:00 p.m. The one-hour time period is divided into 12 partitions. A partition is equal to a 5-minute time window. In each partition, all pairs of trade prices and sizes are sorted in ascending order based on the prices. All trade sizes at the same price level is aggregated into a single trade volume.
BRR is computed by finding the ”volume-weighted median trade price” (VWMP) of each partition, followed by averaging out those prices over all partitions. A simple robust statistical measure, median, helps the BRR be less sensitive to a small number of big trade volumes. The VWMP of each partition becomes less sensitive to those specific volume distribution patterns. The widely used Volume Weighted Average Price (VWAP) is highly sensitive to volume outliers. However, the VWMP cannot handle trade price outliers. Price outliers exhibit a more significant impact on temporal market price movement than volume outliers. As a result, BRR is possibly manipulated by generating a lot of trade volumes at low prices. It leads the VWMP to be biased towards a lower price.
Successful trading on CME Bitcoin Futures and Options requires an deeper understanding on VWMP calculation on BRR partitions including trading activities and quantitative analysis on underline spot exchanges over the one-hour time period. ExxaBlock BRR report provides such details daily and in real time.
1 https://www.cmegroup.com/
2 https://www.cmegroup.com/trading/cryptocurrency- indices/cf- bitcoin- reference- rate.html
3 https://www.cmegroup.com/trading/equity- index/us- index/bitcoin_contract_specifications.html 4Coinbase Pro, Gemini, itBit, Kraken, and Bitstamp.
5 https://www.cryptofacilities.com/cms/storage/resources/cme- cf- reference- rate- methodology.pdf
ExxaBlock Bitcoin Reference Rate
ExxaBlock independently computes the CME BRR and publishes it 4:00 p.m. — one minute earlier than CME .
ExxaBlock BRR provides more than 99.9999% computational accuracy with respect to the official CME BRR . In addition to the BRR, ExxaBlock provides customers details on BRR partitions, trading activities, and their quantitative analysis on BRR constituent exchanges. ExxaBlock BRR report includes the following quantitative summaries.
- BRR Summary
- Partition Summary
- Outlier Summary
- Slippage Summary
- Market Depth Summary
- Exchange Best Bid and Offer (EBBO) Summary
- Unified Order Book Summary
- Latency Summary
1Upon request, unofficial BRR will be provided in other hours (e.g., 4 times per day)
2The minuscule error could come from the different latencies between constituent exchanges and data collection sites (i.e., ExxaBlock vs CME cfbenchmarks). The latencies (and time synchronization difference) decide partition boundaries and data drop-out.
BRR Summary
BRR summary provides the daily BRR with a set of market benchmark prices and statistical summary on the last one-hour (i.e., between 15:00 p.m. and 16:00 p.m.) trading activities at BRR constituent exchanges. BRR presented in the report is independently calculated by ExxaBlock. Market benchmark prices include Volume-Weighted Median Price(VWMP), Volume-Weighted Average Price(VWAP), and Time-Weighted Average Price(TWAP). Please see Appendix A for more details on the market benchmark prices.
Today’s BRR is 11,323.89. BRR constituent exchanges produce 3,107.73773844 BTC/USD trade volume with a value of 35,193,335.46 US Dollars. The volume is generated by 11,255 trades on 4,222 unique price levels during the last one-hour BRR calculation period. Trade prices moves -0.76% between the first and last trades of the one-hour period. The prices swing 1.48% between the highest and lowest prices. There is 5.56 difference between BRR and VWMP. It is mainly because the former is computed by averaging out VWMPs of 12 5-minute partitions while the later is a single VWMP over the BRR one-hour period. VWMP, VWAP, and TWAP are 11,318.33, 11,324.42, and 11,346.4, respectively. The TWAP is higher than other benchmark prices. The VWMP is lower than the others. There is 28.07(0.25%) price difference between TWAP and VWMP. Please see Appendix B for more information on temporal changes of trade prices and volumes at BRR constituent exchanges during the period.
For further analysis, multiple trade volumes at the same trace price are aggregated into a single volume. As a result, 11,255 raw volumes are translated into 4,222 aggregated volumes. The mean(μv) and standard deviation(σv) of the aggregated volume distribution are 0.73608189 and 1.90554765, respectively. 3,897(92.30%) of aggregated volumes are within a 1 standard deviation from the mean of the distribution (μv±1σv). 4,041(95.71%) of the volumes are within a 2 standard deviation from the mean (μv±2σv). The mean(μp) and standard deviation(σp) of the trade price distribution are 11,324.34 and 33, respectively. 2,916(69.07%) of trade prices are within a 1 standard deviation from the mean of the distribution (μp±1σp). 3,990(94.50%) of the prices are within a 2 standard deviation from the mean (μp±2σp).
Robust statistical measures median(m ) and median absolute deviation(MAD)(σ ) suggest there are a large number of outliers in aggregated volume distribution. m and σ are 0.13213832 and 0.12320765, respectively. The difference between μ and m is 0.60394357(82.05%). The difference between σ and σ is 1.78234(93.53%). Such big discrepancy implies that the volume distribution has a long right tail. That is, non-negligible outliers exist in the distribution. Two statistical outlier detection methods, Z3 and IQR1.5, identify 78(1.85%) and 603(14.28%) volume outliers, respectively, from the aggregated volumes at individual price levels. IQR1.5 finds more outliers than Z3. The former uses the robust measure m while the latter does the standard measures μ and σ . The use of the robust measure makes the 2nd moment of the distribution is less sensitive to outliers. Z3 and IQR1.5 outlier detection methods identify 0 and 93 price outliers, respectively. Please see Appendix D for more information on outlier detection methods applied in this report.
VWMP has more price discrepancies than other benchmark prices between exchanges — 34.44(0.31%) in VWMP, 23.95(0.21%) in VWAP, and 2.30(0.02%) in TWAP. itBit trade prices move more than other exchanges between the first and last trade prices of the last one-hour BRR period — 96.5(0.85%). Bitstamp trade prices swing wider than other exchanges between highest and lowest trade prices during the period — 166.46(1.48%).
The majority of trades are executed at Coinbase Pro — 6,265(55.66%). It is 5,786(1,207.93%) more trades that Gemini produces. Trades are executed over a wide range of price levels at Coinbase Pro and a tight range of the levels at itBit — 2164 vs 100. Executed trade size per order is the smallest at itBit(0.08863647) and the largest at Kraken(1.05131785). Coinbase Pro is the most liquid market place and Kraken is the least liquid market place. Bitstamp has the largest difference between the mean and median of the aggregated trade volumes (0.63780539). It suggests that many volume outliers exist at the exchange.
Partition Summary
Partition summary delivers details on trading activities between 15:00 p.m. and 16:00 p.m. on BRR constituent exchanges in a partition level. The one-hour divides into twelve equal-size partitions. A partition is equivalent to a 5-minute time window. In each partition, all pairs of trade prices and sizes are sorted in ascending order based on the prices. All trade sizes at the same price level are aggregated into a single trade volume. VWMP (and other weighted prices (e.g., VWAP and TWAP)) is computed over those prices and volumes. BRR is computed by averaging out VWMPs of those partitions. By tracking VWMPs of individual partitions and trading activities on those exchanges, the official BRR can be estimated or predicted with minimum estimation errors.
The largest and smallest trade volumes are generated in the partition 2(1384.52639666) and 9(1384.52639666), respectively. Skewness(s)1 and Excess Kurtosis(κ3)2 of trade volumes in BRR partitions, as shown in Table 4.2a, show that trade volume distributions in most partitions are highly skewed to the left (a.k.a. positively skewed)and have heavy/fat tails (i.e., leptokurtic). That is, many large (aggregated) volumes exist at at lower price levels. Larger distance between means and medians of trade volumes in BRR partitions lead more volume outliers in the partitions. Most trade price distributions in BRR partitions are close to the normal distribution and have light/thin tails (i.e., Platykurtic). Normality of trade prices and volumes in BRR partitions can be tested further using advanced statistical models.
Coinbase Pro generates the largest trade volume — 1270.10605962. Coinbase Pro, Bitstamp, Kraken, Gemini, and itBit generate 1270.10605962, 1228.52506211, 503.58125064, 84.51852312, and 21.00684295, respectively. Coinbase Pro provides 294.61558270 (713,354.92%) more volume than itBit . The largest exchange partition volume is generated by Bitstamp at the partition 2–294.65688270. The smallest exchange partition volume is generated by itBit at the partition 8–0.04130000.
Trade values are directly associated with trade volumes. Coinbase Pro generates the largest trade value — $14,380,367.03. Coinbase Pro, Bitstamp, Kraken, Gemini, and itBit generate $14,380,367.03, $13,917,040.51, $5,702,174.49, $956,285.67, and $237,467.75, respectively. Coinbase Pro provides $3,339,451.92 (713,146.56%) more value than itBit. The largest exchange partition value is generated by Bitstamp at the partition 2 — $3,339,920.19. The smallest exchange partition value is generated by itBit at the partition 8 — $468.27.
Outlier Summary
Volume and price outlier summary lists outliers of aggregated trade volumes and unique trade prices in multiple time and data resolutions. Two different outlier detection methods, Z-score and IQR, are applied to trade volume and price data. Z-score uses 3 as an outlier cut-off value (Z3). That is, all trade volumes and prices that are located at 3 standard deviation away from the mean of their distributions are considered as outliers. IQR (InterQuartile Range) method uses 1.5 as an outlier cut-off value (IQR1.5). It implies that all trade volumes and prices below Q1–1.5*IQR and above Q3 + 1.5*IQR are considered as outliers. Q1 and Q3 are the first and third quartiles. VWMP is less sensitive to volume outliers compared to VWAP. TWAP is not affected by volume outliers. All of VWMP, VWAP, and TWAP are affected by price outliers. Please see Appendix D for more details on these methods.
IQR1.5 identifies more outliers than Z3 on aggregated trade volumes over BRR one-hour time period — 603 vs 78. IQR1.5 identifies more outliers than Z3 on trade prices over BRR one-hour time period — 93 vs 0. However, it does not imply that the former is better than the latter. A higher IQR cut-off threshold value produces less outliers. A lower Z-score value produces more outliers.
Coinbase Pro has a large number of outliers in both Z3 and IQR1.5 (49 in Z3 and 339 in IQR1.5). Coinbase Pro has the largest outlier in both Z3 and IQR1.5 (45). It is notable that the sum of VN in Table 5.1 (5,066) is greater than the total number of unique trade prices in BRR one-hour time period (4,222). It is mainly because the former groups trade data by exchange and the latter finds unique prices without such data decomposition.
Slippage Summary
Slippage (a.k.a. Implementation shortfall) summary shows the difference between the decision price and the real execution price from multiple time and data resolutions. A collection of real execution prices are represented by a single VWAP that “fully” executes a given market order size1 through at least one price level. As a result, the slippage is the absolute price difference between those two prices. However, it is often measured as a percentage or a basis point against the mid-price. Instead of the mid-price, the best bid price or ask price is alternately used. Please see Appendix E for more details on the method.
Slippage is a proxy of the quality of order executions. By comparing slippages among exchanges or brokerage firms, trading cost can be (significantly) reduced or minimized with or without additional slippage optimization. Slippage is one of few critical factors in successful trading stragy development and execution.
Slippage is measured for a various market order sizes that are divided into three groups — small ($10,000 and $25,000), medium ($50,000 and $100,000), and large ($250,000 and $500,000) — at both bid and ask sides on exchange order books. Slippages measured in bps are shown in the following tables. Slippages measured in absolute price differences are available in Appendix E.3 and E.4. Othe quantitative measures related to slippage such as VWAP, slippage volumes and prices are also available in those appendices.
1Market order size is commonly described by a dollar amount instead of an order quantity in slippage analysis.
Market Depth Summary
Market depth summary shows available liquidity at exchanges that can be bought and sold at different price levels. A market depth is commonly represented by the percentage away from the mid price instead of absolute price level. If the market depth is “deep”, there is sufficient market volume to be traded with less market impact. Please see Appendix F for more details on market depth calculation methodology.
Market depth is one of common measures for market liquidity finding and analysis. It is measured for a various market depth sizes that are divided into three groups — small (0.1%, 0.2%, 0.3%, 0.4%, and 0.5%), medium (0.75%, 1.0%, and 2.5%), and large (5.0% and 10.0%) — at both bid and ask sides on exchange order books. Market depths are measured and aggregated into BRR partitions as shown in Table 7.1 and 7.2. Non-aggregated market depth measurements are presented at Appendix F.1 and F.2.
from a mid price. Vb is a mean of BRR partition bid market depth volume, which is μpt in Table 7.1. Va is a mean of BRR partition ask market depth volume, which is μpt in Table 7.2. V∆ is volume difference between Vb and Va. A positive V∆ means more volume in the bid side. Id is market depth volume imbalance, which is defined by 100*V∆/min(Vb, Va).
Market volume is highly skewed into the bid-side when the market depth is depths are 0.2%, 0.3%, 0.4%, 0.5%, 0.75%, 5.0%, and 10.0%.
itbit and gemini provide more and less liquidity than other exchanges at the bid-side on exchange order books in shallow market depths as shown in Figure 7.1. itbit and gemini provide more and less liquidity than other exchanges at the bid-side on exchange order books in medium and deep market depths as shown in Figure 7.2.
itbit and gemini provide more and less liquidity than other exchanges at the ask-side on exchange order books in shallow market depths as shown in Figure 7.3. kraken and itbit provide more and less liquidity than other exchanges at the ask-side on exchange order books in medium and deep market depths as shown in Figure 7.4. itBit has less market depths than other exchanges at the ask-side of order books. It provides market depths up to 2.5% while others provide full market depths up to 10.0%.
EBBO Summary
Exchange Best Bid and Offer (EBBO) traces the quality of liquidity available at BBR constituent exchanges in real time . The quality is measured by tracking a number or a rate of the best bids and offers produced by those exchanges. The best bids and offers are provided by better pricing (i.e., higher prices in bid and lower prices in ask.) or more liquidity at a same price.
Best asks are offered more than best bids on BRR constituent exchanges — 25,246 vs 11,249. coinbasepro provides more best bids than other exchanges — 8,051(71.57%) It supplies highly-desirable new best bids with a value of $72,439,426.33 to the market . Those bids provide higher prices than the existing market-wide best bids or more volumes at the best bids. coinbasepro provides more best asks than other exchanges — 23,210(91.94%) It furnishes new best asks with a value of $230,091,426.13 to the market. Those asks provide lower prices than the existing market-wide best asks or more volumes at the best asks.
The most frequent order volume sizes produced by exchange best bids are 0.20000000 in Gemini, 0.25000000 in Coinbase Pro, 3.37429363 in Bitstamp, 0.20000000 in itBit, and 2.00000000 in Kraken. The most frequent order volume sizes produced by exchange best asks are 0.20000000 in Gemini, 0.15000001 in Coinbase Pro, 0.30000000 in Bitstamp, 2.00000000 in itBit, and 0.20000000 in Kraken.
1Please contact ExxaBlock for more information on real time access to ExxaBlock EBBO streaming channels.
2Here, the market is a collection of BRR constituent exchanges.
tbit provides best bids over a wider price range than other exchanges — $147.75. It also supplies most best bids within a ± 1 · standard deviation price range — $66.40.
coinbasepro provides best asks over a wider price range than other exchanges — $148.63. bitstamp supplies most best asks within a ± 1 · standard deviation price range — $60.80.
The most frequent order values produced by exchange best bids are $22,629.56 in Gemini, $137.60 in Coinbase Pro, $38,259.23 in Bitstamp, $214,931.59 in itBit, and $4,529.04 in Kraken. The most frequent order values produced by exchange best asks are $1,187.77 in Gemini, $22,608.26 in Coinbase Pro, $109.46 in Bitstamp, $11,653.68 in itBit, and $4,368.73 in Kraken.
Risk-free arbitrage is easily built with EBBO by buying at the best bid (i.e., EBB) and selling at the best offer/ask (i.e., EBO) if EBB price is higher than EBO price.
Unified Order Book Summary
Unified order book is a limit order book that is built by aggregating multiple exchange order books into a single one. The order book provides a universal bird-eye view on the current market and its temporal change.
The unified order book easily identifies the bid-ask price crossing over multiple exchanges. The crossing creates a price region/range where bid prices are higher than ask prices. It is a by-product of fragmentation and inefficiency of the current cryptocurrency market. By exploiting temporal changes of the crossing region, as shown from Figure 9.1 to 9.12, successfully risk-free arbitrage trading can be implemented.
Market order flow toxicity is measured by finding extream price outliers on the unified order book. The toxicity is quantified by the distance between those outliers and the best bid or ask price in terms of a percentage and a number of price levels as shown in Table 9.1 and 9.2. Adverse selection risk can significantly arise when order volume are temporarily evaporated on the book by major market events or the high degree of aggressive High-Frequency Trading (HFT) activities. Extreme price outliers target such events.
Latency Summary
BRR constituent exchanges publish their real time streaming market data through WebSocket1 client-server connections. A connection is established directly between an exchange and their customers. The connection is a full-duplex (i.e., bidirectional) communication channel over a TCP/IP . It allows the customer to subscribe a specific market data channel (a.k.a. endpoint). Most exchanges provide public WebSocket endpoints for their market data access. Order book updates and snapshots are accessible through the public endpoints. Trade data are also accessible through the endpoints except for itBit. itBit allows only registered customers to access their trade data.
Latencies3 to BRR constituent exchanges streaming market data servers are measured at three different ExxaBlock test sites independently when BRR calculation starts (i.e., 15:00 p.m.) and ends (i.e., 16:00 p.m.). For an individual BRR exchange market data server,
- latencies are measured 30 times, followed by removing the best and worst 5 measurements.
- Average latency is computed over 20 measurements.
- Average jitter 4 is computed over 19 differences of 20 measurements.
- Above steps are repeated at three different geological locations, which are
Amazon Web Service (AWS) us-east-1 datacenter at Ashbury, VA, US.
Amazon Web Service (AWS) us-west-1 data center at San Francisco, CA, US. IBM Cloud wdc1 datacenter at Washington D.C., US.
Coinbase Pro and Kraken provide the best/smallest and worst/largest latencies when BRR calculation starts (i.e., 15:00 p.m.) — 15.479 ms at AWS us-east-1 vs 435.122 ms at IBM wdc1. Coinbase Pro and Kraken provide the best and worst latencies when BRR calculation ends (i.e., 16:00 p.m.) — 15.447 ms at AWS us-east-1 vs 451.924 ms at IBM wdc1.
Latencies and their jitters of streaming market data play an essential role in algorithmic (or systematic) trading and real time market surveillance. They are directly involved in (risk-free) latency arbitrages, order fill ratio improvement, and market manipulation detection (e.g., wash trading and quote spoofing). As a result, identifying exchanges that provide the minimum latency with the stable jitter becomes critical for institutional and retail market participants.
Appendices
Benchmark Market Prices
Three difference market benchmark prices are used in this report. They are Volume-Weighted Median Price(VWMP), Volume-Weighted Average Price(VWAP), and Time-Weighted Average Price(TWAP).
VWMP is computed by sorting pairs of prices and volumes in ascending order based on prices, followed by selecting the ”middle” price based on cumulative volumes. It is defined by
where a given array of price and volume pairs,
VWAP is simply computed by dividing the sum of price * volume by the total volume as follows. Unlike VWMP, it does not require data sorting.
TWAP is computed by averaging out open(O1), high(Hi), low(Lj), and close(CN) prices over a specific time range. Unlike VWMP and VWAP, It requires time causality between data elements without any data sorting. It is defined by
where i and j are at any locations within 1 and N of the array of price and volume pairs.
Prices and Volumes on Exchanges
B.1 Gemini
B.2 Kraken
B.3 Bitstamp
B.4 itBit
B.5 Coinbase Pro
Cryptocurrency Exchanges
BRR is computed over five spot exchanges that trade BTC/USD between 15:00 and 16:00 p.m. in London time. Those exchanges are Coinbase Pro, Gemini, Kraken, Bitstamp, and itBit. All exchanges allow public access to their streaming market data through WebSocket APIs. However, itBit provides streaming trade data to only customers.
- Coinbase Pro
— Homepage : https://pro.coinbase.com/
— WebSocket API : https://docs.pro.coinbase.com/#websocket-feed •
- Gemini
— Homepage : https://gemini.com/
— WebSocket API : https://docs.gemini.com/websocket-api/
- Kraken
— Homepage : https://www.kraken.com/
— WebSocket API : https://docs.kraken.com/websockets/
- Bitstamp
— Homepage : https://www.bitstamp.net/
— WebSocket API : https://www.bitstamp.net/websocket/v2/
- itBit
— Homepage : https://www.itbit.com/
— WebSocket API : https://api.itbit.com/marketdata
Outlier Detection Methodology
Two outlier detection methods, Z-score and InterQuartile Range(IQR), are applied to trade volume data
generated by BRR constituent exchanges between 15:00 p.m. and 16:00 p.m. in London time (i.e., GMT and BST).
D.1 Z-score Method
Z-score (a.k.a., standard score) is defined by
where:
X = a random variable that represents trade volume data generated by a BRR constituent exchange,
μ = a mean of the trade volume data,
σ = a standard deviation of the trade volume data
All trade volumes that their Z-scores are less than -3 or greater than +3 (Z3) are considered as outliers in the report. Alternatively, outliers are defined by all volumes located at 3 standard deviation (3σ) away from the mean (μ) of volume distribution. That is,
In a normal distribution,
has a probability of .0027 (2*00135) That is, we consider only .27% of trade volumes as outliers Z3 is an extreme measure.
D.2 IQR Method
IQR1.5 is defined by
Where:
X = a random variable that represents trade volume data generated by a BRR constituent exchange,
Q1 = the first quartile of the trade volume data,
Q3 = the third quartile of the trade volume data,
IQR = Q3 − Q1
IQR1.5 considers all trade volumes that are lower than Q1–1.5*IQR and higher than Q3+1.5*IQR. In a stan- dard normal distribution (i.e., N(0,σ2)), Q1–1.5·IQR and Q3 + 1.5·IQR are equivalent to -2.698σ and 2.698σ, respectively. As a result, IQR1.5 detects slightly more outliers than Z3 in the normal data distribution. IQR1.5 is visually described by
IQR1.5 is based on a robust measure median (m). The median is less sensitive to outliers than the mean (μ). As a small number of outliers are far away from the mean of the data (i.e., |μ − m | ≫ 0), Z3 becomes skewed to the direction where those outliers exist. Meanwhile, IQR1.5 becomes less effective from those outliers.
Slippage
Slippage refers to the difference between the mid price1 and the actual execution price on a limited order book for
a given market order. It is measured by the price difference or the ratio such as a percentage and a basis point.
E.1 Slippage as a price difference
A large market order size is matched and executed over multiple price levels (1 . . . L) on a central limit order book. Slippage for the buy market order volume/size vm (Sask) is defined by
Where:
a0 = a mid price on an order book,
ai = an ask price at the ask-side order book level i,
vi = an ask volume available at the ask-side order book level i,
vm = a buy market order volume, which is equal to
1Instead of the mid price, the best ask price (a1) can be used.
The equation E.1 is computationally equivalent (a.k.a. duality) to
W1,L is a volume-weighted ask price that executes the given market buy order volume vm. Similar to Sask, Slippage for the sell market order vm (Sbid) is defined by
Where:
b0 = a mid price on an order book,
bi = an bid price at the bid-side order book level i,
vi = an bid volume available at the bid-side order book level i,
vm = a sell market order volume, which is equal to
The equation E.3 is computationally equivalent to
W1,L is a volume-weight price that trades the market order size vm. Equations from E.1 to E.4 are generalized by
Where:
p0 = a mid price of a limited order book,
pi = a price level in the order book
E.2 Slippage as a ratio
Slippage is often measured in a ratio (e.g., a percentage and a basis point), instead of the absolute price difference, as follows.
Where:
p0 = a mid price on an order book,
W1,L = a volume-weighted price,
λ = a numeric scale factor
λ is 100 for the percentage (i.e., %) and 10,000 for the basis point (i.e., bp).
E.3 Slippage measures at the bid-side order book
E.4 Slippage measures at the ask-side order book
Market Depth
Market depth is a way to measure the degree of market liquidity by cumulative order volume associated with a specific price-level at an exchange order book. The volume is a collection of bid or ask volumes available at the order book. The price-level is specified by a percentage away from the mid price of the order book. Instead of the mid price, the best bid or ask price is alternatively used. It shows available liquidity at the exchange that can be bought and sold at different price levels. If the market depth is “deep”, there is sufficient market volume to be traded with less market impact.
For a given market depth d(%), market depth volume Vd is defined by
Where:
Vd = a market depth volume
Vi = a volume at the order-book price level i
ap = a market depth price level at an order book
The market depth price level ap is defined by a0·(1 — d)/100 and a0·(1 + d)/100, respectively, at the bid- and ask-side order book. a0 is the mid price of the order book. Alternatively, a1 (i.e., the best bid or ask price), is used instead of a0 (i.e., the mid price) — a1·(1 — d)/100 and a1·(1 + d)/100.
F.1 Multiple market depths at the bid-side order book
F.2 Multiple market depths at the ask-side order book
API and Database Access
ExxaBlock supports real time and historic APIs and Database to support programmable access to daily BRR reports as well as raw and analytic data. Please contact ExxaBlock Support (jwiehl@exxablock.com) if interested.
