- Price discrepancies between exchanges can emerge for a variety of reasons, including market manipulation, exchange downtime, and trader error. These dislocations are aggravated by market inefficiencies that may prevent arbitrage.
- While market dislocations are particularly common for small exchanges and illiquid assets, even liquid markets on major exchanges are impacted reasonably often. The presence of market dislocations makes relying on price feeds from a single exchange unreliable for portfolio valuation and contract settlement.
- Creating reference rates that are robust to market dislocations is surprisingly complex. Coin Metrics offers Hourly and Real-Time Reference Rates for many of the assets covered by the CM Market Data Feed.
Liquidity in cryptocurrency markets is fragmented across a handful of major exchanges and scores of minor ones. Due to market inefficiencies, manipulation, and trader error, prices on these exchanges often diverge, with at least one venue mispricing the asset and failing to reflect the global market price. This issue is especially acute for illiquid and smaller-capitalization assets, which may have weaker settlement assurances and are more prone to manipulation.
Beyond creating arbitrage opportunities, this lack of a robust cardinal market price leads to difficulties in portfolio valuation and contract settlement. Derivatives like futures and options require a price against which to settle, necessitating the use of a reference rate that accurately reflects the conditions across markets.
To address the growing need for cardinal prices, Coin Metrics has developed Hourly and Real-Time Reference Rates for many of the assets covered by the CM Market Data Feed. These rates calculate the market price of an asset against a lookback period of one hour and one second, respectively, combining data from several marketplaces to create a price feed that is robust against market inefficiencies. Our live reference rates are available as part of our free community data.
Markets are typically modeled as efficient, reflecting in their prices all known information. In an efficient market, mispricings tend to be short-lived, since any price discrepancies are closed through arbitrage. These markets are said to obey the law of one price, which argues that identical goods should be sold for the same price across marketplaces.
In the presence of transaction costs and operational risks, however, even rational markets may not behave efficiently. In an inefficient market, price divergences may be sustained so long as friction persists.
The most substantial sustained price dislocation in the cryptoasset market has been between spot prices on Bitfinex and on other exchanges. Due to concerns over the exchange’s solvency, Bitcoin on Bitfinex has frequently traded at a premium to the rest of the market, most prominently during late 2018 and early 2019.
Since the spread was first observed by Coin Metrics in April of 2017, Bitfinex’s BTC/USD market has not been factored into Coin Metrics’ Hourly or Real-Time Reference Rates. A snapshot from a typical trading day in late 2018 shows a spread of about 1.3% between the Bitfinex BTC/USD market and the markets used to compute these rates.
In addition to concerns over a counterparty’s cash flows, dislocations may be sustained due to concerns over the settlement assurances of the asset being traded.
The primary function of a public blockchain is to provide a settlement layer for the transfer of assets, and the proper execution of this function requires that transactions be probabilistically irreversible given a sufficient number of confirmations. This immutability can be compromised in several ways, most infamously through 51% attacks, in which an attacker with control of the majority of a network’s hashpower reorders the blockchain.
Recipients of poorly-secured assets may therefore require a large number of confirmations in order to recognize a transfer as valid. Exchange operators must be particularly cautious, due to the volume of deposits they receive and therefore stand to lose in the event of a reorganization. This has led exchanges to raise the wait time and number of confirmations required to deposit some assets. Increased wait times, in turn, increase the amount of risk taken on by traders seeking to profit from market inefficiencies in these assets, aggravating existing illiquidity and potentially leading to sustained market dislocations.
The most notable incident of this type occurred on April 29, 2020, when Coinbase’s Ethereum Classic (ETC) markets diverged significantly from those on other exchanges. The dislocation was wide and lasted several hours, in part due to the large number of confirmations required by exchanges for ETC deposits following several 51% attacks on the chain.
Complicating this dislocation is the fact that Coinbase is the primary marketplace on which ETC is traded, reducing clarity on which price should be considered the market price and highlighting the need for transparently calculated reference rates.
Capital controls are another source of market friction, introducing barriers in foreign exchange markets that have echoes in cryptocurrency markets. These barriers were largely responsible for the so-called “Kimchi premium” between spot markets quoted in Korean won and those quoted in other fiat currencies.
Capital controls fall into the broader category of restrictions on fiat transfers that impact liquidity in cryptocurrency markets. Delays in fiat deposits or withdrawals caused by exchange downtime or strained relationships with banking partners are another, related source of friction.
Because supply and demand are not guaranteed to be homogeneous across exchanges, prices can diverge in an inefficient but rational market. In reality, market participants are prone to error and irrationality, introducing further sources of misalignment.
In one common type of error, known as a fat-finger error, a trader mistakenly submits an incorrectly typed trade. These errors may result in flash crashes, or rapid downward market movements that are quickly corrected.
On May 17, 2019, the Bitcoin market experienced a flash crash caused by a single large sell order that may have been placed in error. The effects of the crash were felt particularly strongly on Bitstamp, the exchange where it originated.
As a result of their long computation window, Coin Metrics’ Hourly Reference Rates were unaffected. Coin Metrics’ Real-Time Reference Rates correctly tracked the market price, excluding the additional downward movement on Bitstamp.
It’s difficult to discern a user’s intent in submitting a trade, and flash crashes that appear to be caused by an incorrectly placed trade may in fact have resulted from intentional manipulation. In spot markets, there are several ways in which a seemingly irrational activity could be profitable, including as a way of tampering with the index used by a derivatives market.
Bitstamp’s markets are used in the calculation of BitMEX’s settlement prices, leading to speculation that the Bitstamp flash crash may have been induced by someone seeking to profit from a trade on the derivative exchange. Because BitMEX’s Bitcoin price index is computed as a weighted mean, it is not robust to outliers in its constituent markets. As a result, a malicious actor could have profited from opening a leveraged short on BitMEX and manipulating the underlying index.
Manipulation is especially common in illiquid markets, where the amount of capital required to move the market is much lower. Coin Metrics’ Hourly and Real-Time Reference Rates managed to exclude one possible instance of manipulation on Bittrex’s TRON markets from their calculations by giving lower weight to markets with unexpectedly high variance.
It’s difficult for observers to determine market participants’ intent, and this incident may have resulted from user error rather than manipulation.
Another common source of dislocation in crypto markets is operator downtime. While exchanges operating in traditional markets typically conduct scheduled maintenance outside of trading hours, the round-the-clock nature of crypto markets means that any maintenance requiring downtime will necessarily be disruptive to users.
Exchange downtime can lead to stale pricing data if not properly accounted for by index providers. These disruptions are common and underscore the need for reference rates that collect data from diverse sources and are robust to anomalies in one or more of these sources.
One instance of exchange downtime on Kraken was preceded by an outlying trade, and immediately followed by a flurry of stale transactions once trading resumed. This disruption also happened to coincide with anomalous trading on bitFlyer, but was successfully navigated by Coin Metrics’ Hourly and Real-Time Reference Rates.
Spontaneous downtime can be even more disruptive to traders than scheduled maintenance. Unscheduled trading disruptions are often caused by overloaded servers, which tend to act up during periods of high market volatility, potentially leading to loss of funds by users.
The most infamous incident of this type occurred during the March 12th market crash, when BitMEX was brought down. The mechanisms behind this crash were covered in depth in State of the Network Issue 43, and while exchange downtime did not lead to a spot market dislocation in this instance, it left lasting marks on the market that were discussed in State of the Network Issue 47.
Robust Reference Rates
Regardless of their cause, pricing anomalies are an inevitability in crypto markets. Reference rate designers, particularly derivative exchange operators, must prepare for these eventualities by ensuring that their prices are robust to such events. The manner in which a reference rate is calculated can dramatically impact its response to market dislocations and other anomalies.
The Bitstamp flash crash highlights that even liquid markets on major exchanges can experience anomalies. With this considered, reference rate operators should design their methodologies to be robust against failures on these markets. One effective technique for doing so is to use a median price across exchanges rather than a mean price, since medians are more robust to outliers.
Reference rate designers should also establish a weighting penalty against markets that experience an unexpected amount of variance during the calculation period, since this is likely to indicate a dislocation from the rest of the market through a fat-finger error or market manipulation. Since illiquid exchanges are both less critical to global market conditions and more prone to manipulation, reference rates should also factor trading volumes into their weightings.
The reputation of a marketplace should be also considered in deciding whether or not to whitelist its markets, since reported prices and volumes must be trusted. While this process cannot be fully automated, the implementation of a listing framework can dramatically reduce the number of cases requiring operator discretion.
Maintaining robust reference rates is crucial for effectively handling pricing anomalies, which occur frequently in the fragmented and inefficient crypto market. This is particularly significant for derivative exchanges, whose products require a settlement price and may be vulnerable to manipulation in the underlying markets if this index is not well-designed. For more information about Coin Metrics’ reference rates, check out our reference rates product page and live reference rates chart.