High-Frequency Trading Comes to Cryptocurrency

Courtesy of Marissa Cantu and Lee Reiners

Amidst last December’s stock market volatility, Treasury Secretary Steve Mnuchin argued “high-frequency traders combined with the Volcker Rule” were to blame.  Secretary Mnuchin then announced he intended to ask the Financial Stability Oversight Council to review how market structure, in conjunction with high-frequency trading, contributes to stock market volatility.[1] This was but the latest attempt to blame the vagaries of the stock market on high-frequency trading, which entered public consciousness in 2014 upon publication of Michael Lewis’s Flash Boys.

High-frequency trading (HFT) is an automated trading strategy that uses decision making algorithms, supercomputing power, and low-latency trading technology to exploit market pricing inefficiencies for profit.[2]  HFT strategies require investors to trade in high volumes and are most profitable in volatile markets,[3] making HFT a convenient scapegoat for market instability. During 2017’s period of relative market tranquility – the VIX, or so called “fear gauge,” reached a historic low in November 2017[4] – HFT firms suffered, with aggregate revenue from trading US equities dropping below $1 billion for the first time since the financial crisis.[5]

While all was calm in the equities market in 2017, the cryptocurrency market was experiencing opposite conditions. When Bitcoin shot from $900 at the beginning of the year to $20,000 by years-end, HFT firms and other institutional investors took notice. Recognizing an opportunity, several cryptocurrency exchanges began rolling out specific products and platforms to cater to HFT firms, including: colocation services, application programming interfaces, sliding fee scales, and institutional accounts. While these services have been present in traditional securities markets for over a decade, their roll-out in the cryptocurrency space has largely gone unnoticed. This is partially due to the relatively small size of the cryptocurrency market, and the fact that cryptocurrency exchanges are unregulated at the federal level in the U.S.[6]

This post will shed fresh insights on the role of HFT in the cryptocurrency market and highlight the associated risks. High-frequency trading strategies can cause negative market externalities, including price variation, high volatility, and illiquidity.  Specifically, high volume trading and institutional-sized investing could pose a danger to the cryptocurrency market because of the market’s smaller size and corresponding sensitivity.

HFT Meets Cryptocurrency

 HFT strategies can be classified into four general categories: market making, momentum or event trading, liquidity detection, and arbitrage.[7]  In market making, high-frequency traders take advantage of latency to buy and sell securities in microseconds, thereby profiting from the spread between the bid and ask prices.[8]  Momentum strategies focus on identifying short-term price movements and acting upon predicted market responses.  Liquidity detection strategies focus primarily on identifying the market engagements of other traders, usually institutional investors, and then trading upon that information.

Arbitrage trading, the most common HFT strategy, identifies misalignments in the pricing of two similar assets and exploits the price differential for profit.  These misalignments are often the result of low latency, and HFTs can engage in latency arbitrage.  Latency arbitrage involves exploiting a time disparity between a price movement in the market and the time it takes individual exchanges to update the price.  Latency arbitrage occurs in a matter of milliseconds or even microseconds.  In order to effectively implement a latency arbitrage strategy, firms must have supercomputing capabilities and trading algorithms that quickly identify and act on market price variations.

The cryptocurrency market is particularly attractive to automated traders because of the market’s high volatility.  In fact, cryptocurrency is the most volatile liquid asset on the market;[9] the historical volatility of equities is approximately 13.4%, while the historical volatility of Bitcoin is 70%.[10]

Moreover, like the forex market – where HFT is particularly active – the cryptocurrency market is operational 24 hours per day.  Additionally, the cryptocurrency market offers investors trading variability.  Cryptocurrencies can be bought, sold, and traded using fiat currency or another cryptocurrency, and with over 1700 cryptocurrencies and 180 fiat currencies, investors have plenty of opportunities to execute strategic arbitrage trades.

Cryptocurrency Exchanges React

In their early days, cryptocurrency exchanges catered to investors buying or selling in low volumes. When institutional investors took notice and began entering the market in 2017, most cryptocurrency exchanges were under-prepared, leading to long trading halts, delayed operations, and frequent crashes.[11]

To accommodate the new demand from institutional investors, and to entice HFT firms, which presented a lucrative new revenue source, several cryptocurrency exchanges introduced low latency capabilities and adaptable trading platforms equipped to handle rapid, high-volume orders.

Colocation Services

In the past 24 months, several large cryptocurrency exchanges have introduced new features designed to reduce trading latency.  Last May, Coinbase, the largest U.S.-based cryptocurrency exchange, announced low-latency colocation services would soon be available to institutional investors.[12]

Colocation allows investors to house their privately-owned servers inside an exchange’s data center[13]; colocation is particularly valuable to HFT because it substantially decreases the amount of time required to execute trades.[14]  Decreasing a transaction’s latency allows traders to outpace competitors, leading to more trades filled at favorable prices.[15]

Colocation is widely used in liquid asset markets to achieve the fastest possible data processing speeds.[16]  In 2010, the NYSE built a 400,000 square foot data center that offered colocation services to traders.[17]  Because of the high demand for colocation services, the facility sold out of server space before it even opened.[18]

Colocation also provides HFTs with the capabilities to engage in parasitic trading practices.[19]  With faster access to market information, HFTs may be able to engage in a predatory trading practice known as “front-running” which involves using the knowledge of a large, impending order to trade ahead of that order so that the HFT firm benefits from the subsequent price movement.

Front-running can erode investor confidence in an exchange and market. Considering that the cryptocurrency market already suffers from a credibility problem – according to the Wall Street Journal, since 2011, there have been fifty-six virtual currency exchange cyberattacks, resulting in $1.63 billion in losses[20] –  the presence of HFT firms may make average investors even more reticent to engage with cryptocurrency.

Application Programming Interfaces

An application programming interface (API) is a critical software programming tool that controls how an application’s functionary components interact with a server.[21]  APIs are software packages containing routines, protocols, and tools that control how new programs and applications will communicate with a server.[22]  APIs permit application developers to customize functions and capabilities within a proven design framework.[23]  Software programmers must incorporate a server’s API packages when developing new applications that interact with the API’s parent server.

When operating on a cryptocurrency exchange, an automated trader will need to incorporate an exchange’s specific APIs into any automated trading programs.  APIs permit new programs to perform preset functions and interactions with a server by allowing a program to call back to an exchange’s server in a uniform manner.  The APIs are adaptable and can be used when programming algorithmic or automated trading functions.  Furthermore, APIs allow traders to create faster and more complex trading programs.[24]

API access is essential for institutional investors running trading algorithms on specific cryptocurrency exchanges.  Currently, the 15 largest cryptocurrency exchanges all provide APIs as freeware.[25]  Traders can download an exchange’s APIs and build upon the existing API framework to run customizable trading programs.  Some cryptocurrency exchanges have developed APIs specifically designed to work with institutional trading strategies.[26]

Institutional Accounts

Many prominent cryptocurrency exchanges now offer institutional or professional accounts.  Depending on the exchange, these upgraded accounts may allow customers to execute trades in higher volumes or implement certain software capabilities designed specifically for institutional investing.

Coinbase recently launched Coinbase Prime, a professional trading platform designed for institutional traders.[27]  Coinbase Prime users will have access to exclusive, institutionally-focused, trading services and onboarding training.[28]  Binance, another large cryptocurrency exchange, recently announced a unique account structure designed to fit the platform needs of institutional investors.[29]

Sliding Fee Scales

To accommodate high volume traders, some cryptocurrency exchanges have implemented sliding fee schedules.  These schedules can be based on specific market activities or trading volume.  A sliding fee schedule provides behavioral incentives for traders through lower transaction costs.

Some exchanges have implemented fee schedules based on trading volume. These exchanges charge high volume traders smaller fees as their overall trading volume increases.[30] This can increase a high-volume investors overall profitability and provide incentives to consolidate trading activities on a single platform.

Additionally, some platforms separate their fee schedules based on market-making activity and market-taking activity.[31]  Market-makers provide liquidity and price stabilization to the marketplace by placing limit orders.[32]  Market-makers are rewarded for the benefits they provide to the marketplace through lower transaction fees.[33]  Market-takers create market orders that are immediately filled and are on-par with current market values.[34]  In a sliding fee system based upon market making and taking activities, market takers are generally charged higher transaction fees.[35]  Because automated trade algorithms operate using limit orders, the market-making fee discount is particularly attractive to HFT.

While some exchanges only implement one of the fee schedules, many exchange platforms have adopted some combination of the two. Binance, Coinbase, and CEX.IO, three of the largest cryptocurrency exchanges, each offer fee discounts to high volume traders and charge a fee premium for market taking activity.[36]

Limitations of Existing Cryptocurrency Exchanges

While the previously mentioned products and services were designed with HFT in mind, their overall effectiveness may be limited due to some unique features of the cryptocurrency market. Many exchanges have daily limits on a customer’s trading activities, which reduces the available liquidity pool for HFT to trade against. Furthermore, institutional traders operating on a cryptocurrency platform do not have the same hardware advantages as institutional traders operating in traditional markets.

Use Limits

Most prominent cryptocurrency exchanges impose investor limits,[37]  be they limits on the amount of fiat currency investors can withdraw in a given time period or caps on the volume of trades investors can conduct on the platform within a time period.[38]  For example, Binance, one of the exchanges with the highest daily trading volume, limits users to tradingno more than 100 Bitcoin per day.[39] Bittrex, a trading platform designed for programmers and institutional investors, currently limits users to 500 open orders and 200,000 orders per day.[40]

While these limits may make sense for an illiquid retail market, HFT firms will bump up against these limits in a matter of minutes.[41] Recognizing this reality, last August, Coinbase raised the daily purchase limit from $25,000 per week to $25,000 per day.[42]

In addition to trading volume caps, many exchanges also implement activity limitations through their APIs.[43]  API restrictions limit the number of requests a trading algorithm can make to an exchange’s server in a given time period.  Some APIs limit the amount of call requests a trader can make in a day and some exchange APIs create call limitations in minute-long increments; both can severely impair a trader’s ability to function effectively.

For example, Binance’s APIs restricts each trading algorithm to making 1,200 requests to Binance’s server per minute.[44]  While 1200 calls per minute sounds like a lot to the average observer, HFT orders are usually executed within microseconds (there are 60 million microseconds in a minute.)

Coinbase, Binance, and Bittrex are just a few examples of cryptocurrency exchanges that enforce activity limitations on users.  However, these exchanges have also introduced products and features designed to attract institutional investors.  Bittrex offers users advanced APIs that are compatible with high volume, automated trading algorithms.[45]  Coinbase offers colocation to investors.  And Binance recently announced a list of products and services geared towards institutional investors.[46] Thus, it appears as though these exchanges are attempting to maintain the speed bumps necessary to prevent predatory and manipulative behavior on their platforms while at the same time attempting to capture the lucrative fee revenue that comes from institutional – specifically HFT – trading. Whether this balance can be maintained long-term is highly questionable.

Field-Programmable Gate Arrays

Cryptocurrency traders also lack some of the hardware advantages HFTs implement in traditional markets, such as field-programmable gate arrays (FPGAs).  FPGAs permit computing systems to perform customizable tasks at a faster rate than standard computer systems.[47]  In the same manner that APIs permit trading programmers to customize functionality, FPGAs permit functionality customization directly through a computer’s preexisting or specially designed hardware.[48]  However, APIs require programs to call to a server before processing a request and are therefore slower than FPGAs.  For customizable programming, FPGAs permit computers to parallel process programs which could allow systems to execute an algorithm simultaneously instead of in a sequence of operations.[49]

Traditional asset exchanges, like NASDAQ, already use FPGAs for faster information processing and lower latency[50] and FPGA’s are also used by HFT in traditional trading markets.[51]  Algorithmic trading requires programs to execute a series of tasks in order to function properly and an FPGA can execute a trading algorithm 1,000 times faster than traditional hardware running the same program.[52]

Because of their speed and capabilities, FPGAs have long been used for cryptocurrency mining.[53]  Last fall, Squirrels Research Labs, a cryptocurrency hardware company, announced a partnership with a computing hardware company to develop FPGAs designed to specifically work with cryptocurrency.[54]  While these new FPGAs are currently being developed for cryptocurrency mining, future FPGAs could be created specifically for cryptocurrency trading.

Blockchain-focused FPGAs could eliminate blockchain server overloads, preventing trading delays on exchange platforms, and permit faster execution of trades.[55]

Liquidity Problems

Many argue that HFT provides market liquidity through an increase in trading volume and trading executions .[56]  However, by the time an individual trader notices an opportunity resulting from the added liquidity, the opportunity will vanish.  This is known as a “liquidity mirage,” because individual traders lack the computing speed and technological capabilities necessary to benefit from the liquidity provided by HFT.[57]

Furthermore, because automated traders rapidly react to market movements, any market shock, no matter how slight, could cause liquidity to evaporate. This could happen if automated traders instantaneously halt buy orders and freeze operations before human traders have an opportunity to intervene.[58]

 Even more concerning, is the possibility that stressed market conditions could lead to a rapid sell-off by automated traders. Different trading algorithms may share the same market triggers, and should these triggers go off, there could be a mass sell-off in the market in less than a second. A rapid price decline caused by automated trading is known as a “flash crash.”

Flash crashes have become more common in recent years.  Since 2006, over 20,000 small flash crashes have been recorded, or an average of 12 flash crashes per day.[59]  These flash crashes are triggered by an algorithm’s rapid response to market conditions and are largely unpreventable without altering the algorithm’s function.

While most flash crashes are small in scale, large flash crashes can create extreme market stress and volatility.  In 2010, a single instance of extreme market volatility caused by a malfunctioning HFT algorithm, led $1 trillion in value to disappear from the U.S. equities market in a matter of minutes. Considering that the entire cryptocurrency market has a market capitalization of $176 billion, a single flash crash could conceivably wipeout the entire market before anyone notices.

Conclusion

The cryptocurrency market may simply be too small to safely accommodate HFT. To illustrate, the forex market has an average daily trading volume of $5 trillion, while the average daily trading volume of cryptocurrencies is $7.1 billion (0.0014% of the forex market).[60]  The average daily trading volume of cryptocurrency is also significantly smaller than the $1.6 trillion average daily trading volume for U.S. equities.[61]

The scale of traditional markets is what protects them from the risks posed by HFT. It would be extraordinarily difficulty for a single trader, or algorithm, to destabilize the forex or equity markets. The same cannot be said for the cryptocurrency market, where manipulation is rampant and relatively small trades can meaningfully move the market.

The cryptocurrency market’s limitations have not stopped several high-profile exchanges from charging ahead with their efforts to cater to HFT firms and other institutional traders. These new products and services provide additional revenue to cryptocurrency exchanges, many of whom have been struggling with a decline in retail trading volume that corresponds with the bear market in cryptocurrency.

While opening up to HFT may make sense from an individual exchange’s standpoint, it could potentially destabilize the entire cryptocurrency market. Because cryptocurrency spot markets are essentially unregulated, there is no government agency pushing back on these exchanges and warning them of the potential risks to the broader market. As institutional investors take up a larger share of the market, one bad algorithm could sink the entire cryptocurrency experiment for good.

 

 

[1] Saleha Moshin et al., Mnuchin Blames Volker Rule, High-Speed Trading for Volatility, Bloomberg (Dec. 18, 2018), https://www.bloomberg.com/news/articles/2018-12-18/mnuchin-blames-volcker-rule-high-speed-trading-for-volatility?srnd=premium.

[2] Rina S. Miller and Gary Shorter, High Frequency Trading: Overview of Recent Developments, Congressional Research Service (Apr. 4, 2016), https://fas.org/sgp/crs/misc/R44443.pdf.

[3] Alexander Osipovich, High-Frequency Traders Fall on Hard Times, The Wall Street Journal (Mar. 21, 2017), https://www.wsj.com/articles/high-frequency-traders-fall-on-hard-times-1490092200.

[4] Preeti Varathan, 2017 was the Least Volatile Year in Decades, Quartz (Dec. 22, 2017), https://qz.com/1154499/the-vix-2017-was-the-least-volatile-year-in-decades/.

[5] Gregory Meyer et al., How High-frequency Trading Hit a Speed Bump, Financial Time (Jan. 1, 2018), https://www.ft.com/content/d81f96ea-d43c-11e7-a303-9060cb1e5f44.

[6] Bitcoin first came under the Commodity Future Trading Commission’s remit in 2015, when the agency determined, in an order against Coinflip, Inc., that bitcoin—and by extension all other virtual currencies—met the definition of a commodity under the Commodity Exchange Act (CEA). The CFTC has jurisdiction over commodity derivatives, but they do not oversee commodity “spot” or cash markets, except in instances of fraud or manipulation. As a result, no federal agency supervises virtual currency exchanges in the U.S.

[7] Matt Prewitt, High-Frequency Trading: Should Regulators Do More?, 19 Mich. Telecomm. & Tech. L. Rev. 131, 134-37 (2012).

[8] Id. at 135.

[9] Arthur Iinuma, Why is the Cryptocurrency Market so Volatile: Expert Take, Cointelegraph (Feb. 27, 2018), https://cointelegraph.com/news/why-is-the-cryptocurrency-market-so-volatile-expert-take.

[10] So You Want to Trade Crypto – Volatility, Hackernoon (May 25, 2018), https://hackernoon.com/so-you-want-to-trade-crypto-volatility-part-2-4dc5ca23ba15.

[11] Modulus: Global Crypto Exchange with Low Latency, High Frequency Trading Launches, Bitcoin Exchange Guide (Jul. 9, 2018) https://bitcoinexchangeguide.com/modulus-global-crypto-exchange-with-low-latency-high-frequency-trading-launches/.

[12] Rich Miller, Coinbase Plans Low-Latency Colocation for Bitcoin Trading, Data Center Frontier (May 15, 2018), https://datacenterfrontier.com/coinbase-plans-low-latency-colocation-for-cryptocurrency-trading/.

[13] What is Colocation, Interoute, https://www.interoute.com/what-colocation (last visited Dec. 11, 2018).

[14] Sureh Aswani, Leveraging Co-location for Competitive Advantage in HFT, HPC Wire (May 31, 2016), https://www.hpcwire.com/solution_content/hpe/financial-services/leveraging-co-location-competitive-advantage-hft/.

[15] Latency War: Why is Low Latency Important?, Quant Insti (Aug. 25, 2015), https://www.quantinsti.com/blog/latency-war-why-is-low-latency-important.

[16] Id.

[17] Rich Miller, NYSE Opens Mahwah Data Center, Data Center Knowledge (Aug. 9, 2010), https://www.datacenterknowledge.com/archives/2010/08/09/nyse-opens-mahwah-data-center.

[18] Id.

[19] Jacob Adrian, Informational Inequality: How High Frequency Traders Use Premier Access to Information to Prey on Institutional Investors, 14 Duke L. & Tech. Rev. 256, 258 (2015).

[20] Steven Russolillo & Eun-Young Jeong, Cryptocurrency Exchanges are Getting Hacked Because it’s Easy, Wall St. J. (July 16, 2018), https://www.wsj.com/articles/why-cryptocurrency-exchange-hacks-keep-happening-1531656000.

[21] Petr Gazarov, What is API? In English, Please, Free Code Camp (Aug. 13, 2016), https://medium.freecodecamp.org/what-is-an-api-in-english-please-b880a3214a82.

[22] Vangie Beal, API – Application Programming Interface, Webopedia, https://www.webopedia.com/TERM/A/API.html (last visited Dec. 14, 2018).

[23] Steve Davis, 3 Ways APIs Help Developers Deliver Software Faster, Tech Beacon, https://techbeacon.com/3-ways-apis-help-developers-deliver-software-faster (last visited Dec. 14, 2018).

[24] Mike Sedzielewski, How to Use API-First Platforms to Build Your Websites Faster, Free Code Camp (Jan. 4, 2018), https://medium.freecodecamp.org/how-to-use-api-first-platforms-to-build-your-websites-faster-e917e8318ee.

[25] Sinisa Radovcic, Market Structure: Cryptocurrencies vs. Stock Markets, Medium (Dec. 18, 2017), https://medium.com/algosforcryptos/trade-market-structure-stocks-vs-bitcoin-f81e04aa9d94.

[26] Developer’s Guide – API, Bittrex (Jan. 2, 2018), https://support.bittrex.com/hc/en-us/articles/115003723911-Developer-s-Guide-API.

[27] Coinbase Prime, https://prime.coinbase.com/ (last visited Dec. 14, 2018).

[28] Id.

[29] John Njui, Binance Opens its Doors to Institutional Investors, Ethereum World News (Nov. 10, 2018), https://ethereumworldnews.com/binance-opens-its-doors-to-institutional-investors/.

[30] Alex Munkachy, The Cryptocurrency Exchanges in the USA with the Lowest Fees, Coin IQ (Oct. 4, 2018), https://coiniq.com/crypto-exchanges-lowest-fees/#Flat_rate_vs_sliding_scale_commissions.

[31] Maker vs. Taker in Cryptocurrency, CryptoCurrency Facts, https://cryptocurrencyfacts.com/maker-vs-taker-cryptocurrency/ (last visited Dec. 11, 2018).

[32] Id.

[33] Id.

[34] Id.

[35] Id.

[36] Id.

[37] Ray King, Binance vs. Coinbase, BitDegree (Aug. 20, 2018), https://www.bitdegree.org/tutorials/binance-vs-coinbase/#A_Quick_Note_on_Trading_Limits.

[38] Binance Review: Fees, Deposits/Withdrawals, Purchase Limits, Token24 (Feb. 27, 2018), https://www.tokens24.com/cryptopedia/trading/binance-review-fees-depositswithdrawals-purchase-limits.

[39] BitDegree, Supra note 37.

[40] Bittrex, Supra note 26.

[41] Nick Baumann, Too Fast to Fail: How High-Speed Trading Fuels Wall Street Disasters, Mother Jones (Jan. 2013), https://www.motherjones.com/politics/2013/02/high-frequency-trading-danger-risk-wall-street/.

[42] Taylor Hatmaker, Coinbase Add Instant Trading and Increases Daily Limits, TechCrunch (Aug. 2018), https://techcrunch.com/2018/08/07/coinbase-adds-instant-trading-and-increases-daily-limits/.

[43] Jeff Vernon, Cryptocurrency Exchange Comparison – Features and Benefits, Haas Online Software, https://www.haasonline.com/cryptocurrency-exchange-comparison-features-and-benefits/ (last visited Dec. 11, 2018).

[44] Id.

[45] Bittrex, Supra note 26.

[46] Ethereum World News, Supra note 29.

[47] Miranda Hansen, What is the Deal with Cryptocurrency and FPGA?, Digilent (Feb. 21, 2018),  https://blog.digilentinc.com/what-is-the-deal-with-cryptocurrency-and-fpga/.

[48] Id.

[49] Nallapeni Manoj Kumar and Pradeep Kumar Mallick, Blockchain Technology for Security Issues and Challenges in IoT, 132 Procedia Computer Science 1815, 1818-20 (2018).

[50] NASDAQ TotalView-ITCH FPGA Version 4.1, NASDAQ, http://www.nasdaqtrader.com/content/technicalsupport/specifications/dataproducts/itch_fpga_4_1.pdf (last visited Dec. 11, 2018).

[51] In Pursuit of Ultra-Low Latency: FPGA in High-Frequency Trading, Velvetech (Sep. 20,2017), https://www.velvetech.com/blog/fpga-in-high-frequency-trading/.

[52] Id.

[53] Digilent, Supra note 47.

[54] Squirrel Research Labs and BittWare Launching new FPGA Crypto Mining Hardware, Crypto Ninjas (Sep. 7, 2018), https://www.cryptoninjas.net/2018/09/07/squirrels-research-labs-and-bittware-launching-new-fpga-crypto-mining-hardware/.

[55] FPGAs could help blockchain servers run more effectively and reduce or eliminate server overloads.  Many cryptocurrency exchanges stated that trading delays and trading halts were due to an overload of the blockchain’s servers.  By reducing server overload, crypto exchanges will be able to trade on a blockchain more efficiently.  Id.  at 1353.

[56] William Barker and Anna Pomeranets, The Growth of High-Frequency Trading: Implications for Financial Stability, Bank of Canada (Jun. 2011), https://www.bankofcanada.ca/wp-content/uploads/2011/12/fsr-0611-barker.pdf.

[57] Id.

[58] Id.

[59] Alexander Munk and Erhan Bayraktar, Opinion: The Stock Market has about 12 Mini Crashes per Day – and we Can’t Prevent Them, MarketWatch (Jul. 31, 2017), https://www.marketwatch.com/story/the-stock-market-has-about-12-mini-flash-crashes-a-day-and-we-cant-prevent-them-2017-07-31.

[60] Compare Cryptocurrency Market Daily Trading Volume to Forex Market Daily Volume, Medium (Nov. 9, 2017), https://medium.com/@ListICOio/compare-cryptocurrency-market-daily-trading-volume-to-forex-market-daily-volume-e4190773efbb.

[61] Market Statistics – Average Daily Trading Volume, Financial Industry Regulatory Authority, http://otce.finra.org/MSAvgDailyPriceVolume (last visited Dec. 14, 2018).