The cryptocurrency market is rife with fraud and manipulation. The ICO craze spawned a wave of internet pump-and-dumps that made Jordan Belfort, the “Wolf of Wall Street,” look like a friendly Jim Cramer. Apart from the obvious, we have witnessed numerous fraud cases and countless instances of price action indicative of other forms of market manipulation.
In September 2017, while Bitcoin rose to prominence, JP Morgan Chase CEO, Jamie Dimon, was one of its most vocal critics. Dimon warned that Bitcoin was a “fraud” and threatened to “fire in a second” any JPM trader who recklessly facilitated the volatile Ponzi. “If you’re stupid enough to buy it, you’ll pay the price for it one day,” cautioned the investment bank chief.
In the aftermath of Dimon’s comments, the digital currency promptly plummeted 32%. The day of crypto-reckoning restored our collective faith in the otherwise healthy and wholesome financial markets.
Three days after Dimon’s 2017 comments, JP Morgan Securities Ltd. and Morgan Stanley bought roughly €3 million worth of XBT, an ETN that tracks the Bitcoin price. The coincidental timing led Blockswater, a London-based liquidity provider, to file a market abuse report in Sweden against the JP Morgan CEO. Blockswater alleged that Dimon violated Article 12 of the European Union’s Market Abuse Regulation by making a misleading statement in order to drive down the price of Bitcoin. However, JPM made these transactions on behalf of clients. In fairness, it is unlikely that saving a couple of million dollars for a client was of any material significance to a bank like JP Morgan, although it may raise broader questions about potential conflicts of interest concerning sell-side recommendations.
Furthermore, in April of 2021, JP Morgan announced the launch of an actively managed bitcoin fund in collaboration with NYDIG. They began pitching the fund in August. It is admirable how the tolerant chief will now allow employees to trade bitcoin without firing them. In a surprising turn of events, JP Morgan and Morgan Stanley will be the two biggest banks first to offer direct Bitcoin access to clients. Other banks such as Goldman Sachs and Bank of America do not facilitate any direct cryptocurrency investments.
INSPECTING THE ENTRAILS
Dimon’s comments were not the only reason why cryptocurrency investors raised concerns about JP Morgan’s public position on the new market. Over the past few years, numerous JP Morgan researchers have made controversial statements in relation to Bitcoin pricing. For instance, JPM’s crypto quant Nick Panigirtzoglou, author of their Flows and Liquidity newsletter, has employed a confusingly varied assortment of methods for valuing the cryptocurrency. Panigirtzoglou has seemingly changed his position abruptly on multiple occasions. Although changing course is part-and-parcel of predicting volatile price action, when publishing sell-side recommendations, it naturally raises questions about potential conflicts of interest.
In January of this year, Panigirtzoglou put forward a technical-based assessment of why Bitcoin would fall from its highs. He argued that trend-followers such as CTA funds would no longer want to pursue it while it was losing momentum. However, the price soon rose to over $60,000 and triggered a short squeeze in the process.
Any clients who shorted Bitcoin on the back of that forecast or any other inaccurate prediction were probably less than pleased.
In November 2020, the JPM Flows & Liquidity team used technical analysis to deduce that Bitcoin was overbought. Panigirtzoglou predicted a mean-reversion. It was then trading at $14,000. By the end of November, it shot up to over $20,000.
Three weeks following his less-than-stellar call, Panigirtzoglou admitted he was wrong and then proposed a revised thesis for bitcoin based on a sort of gold-parity methodology. The elaborate thesis, unlike conventional cashflow or residual income-based valuation models, posited that if Bitcoin were to reach an equal value to gold, relative to the supply of fiat money in circulation, the price of bitcoin should increase to $650,000. The JPM long-term target was $146,000. The price was then sitting at $24,000. Despite the unusual valuation model, the directional guess was accurate. The price moved up to over $30,000 by 2021.
However, shortly after that, in January 2021, the Panigirtzoglou et al. sentiment became bearish once again. They published a report questioning whether bitcoin had “equalized with gold already.” No, bitcoin didn’t surge to $650,000 or even to $146,000. Panigirtzoglou and his colleagues were now instead referring to a ratio between the volatility of gold and the volatility of the Grayscale Bitcoin Trust, which they asserted was of material significance.
Additionally, the JPM soothsayers also used the cost of Bitcoin mining to read the tea leaves even further. They suggested that market participants consider the cost of mining as a fundamental determinant of Bitcoin’s value. However, the Bitcoin elasticity of supply is not exactly akin to switching on an oil rig. If that were the case, the recent supply shock from China banning bitcoin mining should have edged us toward that elusive $146k.
What evidently does drive the price of Bitcoin is sentiment, whether induced by Elon Musk’s Twitter fingers or Wall Street sell-side analysis.”
When JPM issued that report on January 4 2021, the price of Bitcoin stood at over $33,000. On January 5, it fell to $31,431.61, before accelerating to $40,500 over the next four days.
Dimon has made it crystal clear that while he may disapprove of Bitcoin, he is a believer in blockchain technology. JP Morgan sold its enterprise blockchain platform, Quorum, to ConsenSys in 2020. Quorum is an open-source protocol for building applications on Ethereum. JP Morgan is also working on its Onyx blockchain division. The bank announced the Onyx project in October 2020 to reflect their “commitment towards innovation” and “building a more inclusive financial system.”
As for today’s tune, on July 20, Bloomberg released an interview with JP Morgan Asset & Wealth Management CEO Mary Callahan Erdoes, during which she claimed clients now view Bitcoin as an asset class — and an asset class which they want to own.
However, Callahan Erdoes walked back her comments within the very same interview, clarifying that JPM does not view Bitcoin as an asset class “per se.” However, she did forecast that Bitcoin will, in time, prove to be a store of value.
A few weeks after Erdoes’ bullish sentiment, Panigirtzoglou conceded that the rebound “caught most investors by surprise.” It’s also surprising that JP Morgan decided to shut the accounts of a Delaware Bitcoin mining firm on the same day it registered the NYDIG fund with the SEC. It is understandable how that may be seen as a somewhat contradictory stance.
SHAKEN, NOT STIRRED
Most recently, on September 1, Panigirtzoglou was back to bearish sentiment, criticising the recent surge in altcoins. He and his team called it an indicator of “froth and retail mania.” Notably, the spike was led by Solana (SOL), the world’s fastest blockchain, along with other viable, scalable payment solutions such as Ripple (XRP). Ethereum (ETH) has also been head of the pack, as investors speculate on the future of payments and smart contracts. JPM offers its own payment ecosystem via its Chase Pay app.
Panigirtzoglou’s comment about market froth was not exactly unjustified. It remains to be seen what was behind the recent Bitcoin plunge. The coincidence of the fall with El Salvador’s adoption of Bitcoin was curious.
What exactly is “froth” anyway? Well, Oxford defines “frothy” as the state of being “full of or covered with a mass of small bubbles.” The cryptocurrency market is undoubtedly bubbly. Has the additional layer of derivatives such as CME futures and options helped the market to mature and stabilize? Or is it an added layer of sugar, cream, froth and maybe even a drop of whiskey?
Since the inception of Cboe and CME Bitcoin futures in December 2017, there have been allegations of foul play. The introduction of futures coincided with a plunge into a deep, prolonged bear market. Many investors alleged that institutions were afforded access to hedge against the price of Bitcoin, ensuring it would crash.
A potential manipulation technique widely cited in the financial media was the opportunity to “bang the close.” Banging the close is an illegal strategy whereby manipulators seek to influence the price as a futures contract is approaching expiration.
The above chart depicts the expiration dates since the inception of the CME Bitcoin futures. For example, suppose you were to short 10,000 futures that expire on September 24. These contracts are valued at the CME CF Bitcoin Reference Rate (BRR). The BRR aggregates trade flows from major exchanges over a one-hour window between 3 p.m. to 4 p.m. London time, on the last Friday of each month.
The BRR is calculated as the equally-weighted average of the volume-weighted median price for each 5-minute window during the benchmarking hour. As of writing, the exchanges used for the calculation are; itBit, Bitstamp, Coinbase, Kraken, and Gemini. Say you and your crew of conspirators were to instigate intense selling of Bitcoin on these exchanges during the benchmark period. Such heavy selling would cause a temporary imbalance, pushing down the price at which your futures contracts are valued. Furthermore, if your name is Satoshi Nakamoto, you may be happy to cash out your old Bitcoin USB stick profits while simultaneously profiting on the futures shorts.
As the WSJ noted back in 2017, such a scheme would be difficult for U.S. authorities to catch, particularly due to the relative anonymity and lack of regulations for cryptocurrencies. Nevertheless, they accepted it would be extremely difficult to execute without substantial capital (or by swaying public opinion).
However, the WSJ and others failed to mention the fundamental issue with futures trading, which is not specific to Bitcoin. The CME contracts can be settled in cash rather than delivery of Bitcoin. Therefore, derivatives have enabled the creation of “paper” or “digital” Bitcoins (double digital?), i.e. Bitcoin contracts unconstrained by limited supply. The “artificial” volume effectively dilutes the number of Bitcoins in circulation. Effectively, there is no limit to the amount investors can bet against it.
In 2019, Arcane Research pointed to precisely that manipulation. The researchers analyzed Bitcoin price action from January 2018 to August 2019, concluding that Bitcoin fell 75% of the time before CME settlement dates.
We set out to examine the price action of Bitcoin around expiration dates to see if the market exhibited the same anomalous behaviour since 2019. Some of our findings are published below.
We used CME CF Bitcoin Real Time Index (BRTI) data at a 1-second interval, resampled to 1-hour bars for the purposes of this article. CME’s BRTI is an aggregation of order books from the same exchanges used for the BRR. We also looked at some of the constituent exchanges and both CoinMarketCap and Nomics overall market data.
The below histogram depicts the distribution of hourly Bitcoin returns since 2018.
GUATEMALA TO THE LEFT OF ME
As expected, the series is characterized by positive skew and extreme outliers. No doubt the El Salvadoran president is painfully aware of that after yesterday’s rout, during which the price sank by 12% within an hour. The distribution has a kurtosis of over 36 and a skew of almost 0.3.
The next histogram and sequential bar chart show the distribution of hourly returns during the 3-4 p.m. benchmark window on the last Friday of each month. These are the returns from buying at the 2 p.m. candle close and selling at the 3 p.m. candle close.
It should come as no surprise that volatility increases substantially during the benchmark window. In other asset classes, while some asset managers and dealers trade “at the fix” to manipulate prices, there are also those seeking to manage risk. For instance, if a fund is benchmarked by its tracking error relative to the London WMR fixing rate, the manager may want to trade as close to the fix as possible to minimize their tracking error. Although there are relatively few cryptocurrency funds, volatility and increased volume at expiration do not necessarily indicate manipulation.
However, what is striking is the pattern of price movements. It may surprise many cryptocurrency investors that contrary to the widely held perception that price is being “banged” downward each month, it is, in fact, the complete opposite during the benchmark hour. The returns during the last hour of trading for the past 44 months have actually been 29 times higher than the mean hourly returns over the entire period. The skew of this distribution is 2.88. We performed t-tests on the log returns and found the probability of observing such a distribution from a random sample of 44 hours to be about 9.73%.
Despite the adequate sample sizes, we further ran some nonparametric Mann-Whitney U rank tests of the null hypotheses that the distributions underlying the days before and after expiration are the same as the distribution for the entire period. The p-values are compiled in the below table.
Interestingly, the Bitcoin price tends to markedly drop between 48 hours and 24 hours before expiration, before rebounding dramatically in the closing hour and the first day of the next contract period. Both the mean and median returns were statistically significantly greater than those of the entire series, at the p<0.1 level.
Notably, from September 2019 onwards, the statistical significance has been much weaker, despite somewhat similar dubious return patterns. Could the shift be due to the increased media scrutiny since Arcane and Forbes raised the issue in September 2019? Around the same time, a forensic study was published which alleged the ‘stablecoin’ Tether was being used to inflate Bitcoin prices. Perhaps the would-be conspirators have learned to be more subtle. On the other hand, it could indicate that the market is maturing and less prone to being pushed around by bad actors. That being said, returns in excess of average by 1,300%, significant at roughly the 15% level, shows a bit more arbitrage might be in order.
Below is a box plot of the hourly returns during the T-72 to T+72 hours on either side of the benchmark window. Note that the volatility of the ‘T-0’ box is not directly comparable to the others because it contains only 44 samples, while they each contain 1,056 samples. In fact, the hourly volatility for the entire period was 0.885%, and the days around settlement were largely similar. In contrast, the figure for the T-0 windows was about 1.13%.
Finally, we trained an artificial neural network to detect Bitcoin price anomalies. The model we used is a Long short-term memory cell (LSTM) Autoencoder. Autoencoders are an unsupervised machine learning architecture capable of discovering structure within noisy data. They’re often used in computer vision and image editing. Anomaly detection is based on the concept of modelling what is normal to discover what is not normal. Therefore, by training a neural network to encode the Bitcoin price, the movements that it fails to accurately predict beyond a certain threshold are considered abnormal or unexpected.
The model was trained on 95% of the data since 2018 and tested on the past month’s returns. The dots on the above chart represent anomalies that exceeded an absolute error threshold of 0.049. The model was only trained on hourly data rather than the microsecond frequency we use for our high-frequency currency trading strategies. Nevertheless, it was relatively accurate at modelling Bitcoin movements.
It may come as a surprise that the model anticipated the infamous price decline and rebound on September 7, aka “B-Day.” The people of El Salvador may consider that move an anomaly, but the neural network did not. Interestingly, the subsequent price action after the initial decline and rebound was flagged as anomalous.
WAGGING THE DOG
To those who say it is difficult to influence prices, let’s hear what some industry experts have to say. In March 2021, Cboe issued a rather telling proposal to the SEC, requesting to list and trade the VanEck Bitcoin Trust. In the submission, Cboe argued that there are no concerns that the Trust’s benchmark could be manipulated because it is calculated based on price data from the top 5 crypto exchanges, meaning it would require influencing the entire Bitcoin spot market — all well and good.
Notably, however, Cboe acknowledged that the spot market is now “led by the Bitcoin Futures market” in terms of price discovery. This comment confirms that the tail is now wagging the dog. Furthermore, Cboe reassured the SEC that a market participant could enter a market order for $10 million of bitcoin and “only” move the market by 0.5%. Does that mean an order or combination of sell orders totalling $200 million could tank the market by 10%?
There were additional concerns from investors due to a private meeting between Federal Reserve chairman Jerome Powell and Coinbase CEO Brian Armstrong — an unlikely duo. The secluded rendezvous, which took place in person as opposed to Powell’s typical Zoom preference, was followed by a weeklong Bitcoin plunge which brought the price from $57k to $33k before rebounding dramatically. On the other hand, the recent SEC threat to sue Coinbase for its proposed new lending feature raises questions about the political dynamics at play.
PWE Capital does not have exposure to Bitcoin or any other cryptocurrencies (nor do we have any exposure to JP Morgan as a prime broker).
What does the future hold for the mysterious cryptographic asset? Well, we better stay tuned for more JP Morgan analysis.
If you’re a sophisticated investor, a family office or an asset manager, we welcome you to contact us about our investment program. Our high-frequency strategies are unrelated to equities, fixed income or cryptocurrencies.