By Ginger Szala
The 2015 New Year surprise happened fast: the Swiss National Bank announced on Jan. 15 that the Swiss Franc would no longer be capped by the Euro and it jumped 41% in minutes. This black swan event happened just before a U.S. holiday when traders were already winding down for a three-day weekend. Those left trading included a combination of retail players, banks and some prop firms. And the market went into a frenzy.
The losses were still being calculated into the next week and beyond. Several retail forex firms – from London to New Zealand – went bust. Alpari was an early victim. FXCM was on life support until it received a $300 million rescue from Leucadia National Corp., owner of Jefferies Group. Deutsche Bank, Citicorp, and Barclays each suffered losses in the $150 million zone. Clearing firms took hits — Interactive Brokers’ CEO Tom Peterffy admitted the following day they lost $120 million. Early hits in the hedge fund community included Everest, which closed seven of its eight funds. Its one remaining fund, with less than $500 million, represents less than 20% of the assets they had months before, according to Reuters. BlueCrest Capital Management, a multi-billion dollar hedge fund based in London, reportedly shut the trading book of one senior currency traders. Retail losses were widespread and hefty. The maximum leverage permitted for retail cash forex trades in the United States is 50:1. Unfortunately for Europe, the higher leverage permitted there exacerbated losses.
Within a week of this move came another seismic shock in a smaller marketplace. Although most economists had predicted no rate cut, the Bank of Canada caught many traders off-guard on Jan. 21 with a surprise cut in its overnight rate. The resultant rally in front-month Canadian Bankers Acceptance futures, BAs, was 35 basis points. A one standard deviation move in this product equaled 1.5 basis points at the time. The market therefore experienced a 23 standard deviation move. The Swiss Franc move was nearly identically rare from a statistical modeling standpoint.
Each of these events would occur only once in billions of years if this data were distributed “normally.” That we have witnessed so many “black swan” events in our lifetimes highlights an important issue: SPAN and VAR analysis based upon historical data and normal distributions often underestimate REAL risk.
Fortunately, the FX outright futures at CME utilizes FIFO (first in/first out) and not Pro Rata for its fill distribution methodology. If the FX market had used pro rata, the carnage from the Swiss National Bank event could have been even more disastrous.
The fixed income markets at CME utilize pro rata. Had the unexpected rate cut effectuated by the Bank of Canada taken place in the United States by the Federal Reserve Board, the dislocation and market chaos created could have been even more extreme. This is because many traders in pro rata markets enter bids and offers far in excess of the quantity they actually expect to buy or sell. They may bid for 1000, hoping to buy a 20 lot when the bid is hit for 100 and the matching algorithm splits up the sale quantity partly based upon a pro rata methodology.
Consider if we had an unexpected rate move similar to the recent event in Canada. On a typical day, the offered-side in front month Eurodollar futures could be tens or even 100,000 contracts or more. An unexpected move repricing the market by 25 basis points could lead to significant dislocations if traders inadvertently buy or sell multiples of the quantity they actually desire to trade. The trader bidding for 1000 in the hopes of buying 20 while the market remains bid will be considerably disrupted if his bid is filled and the market is nearly instantly 25 basis points lower. If a bid for 100,000 was instantly filled and the market dropped 25 basis points, that filled trade is collectively a $62,500,000 loser. Liquidity in the immediate aftermath of such an event could be disastrously meager as market maker liquidity providers who were just filled beyond their desires struggle to cope with their unwanted long position in a market where bids may be scarce and fleeting. This market disruption is something nobody wants and yet pro rata distribution sets the table for this to occur.
Market participants are fortunate the Swiss Franc and Canadian market events did not occur in marketplaces with a pro rata distribution methodology. There remains a question of whether it is prudent to use this methodology at all. As long as we maintain this distribution process it seems only a matter of time before some form of significant dislocation will occur. What happened in currencies on Jan. 15 was the sort of perfect storm that highlights the potential chink in the pro rata armor.
Upside of pro rata
As mentioned, there are generally two forms of trade matching algorithms used by global exchanges: 1) Pro Rata, and 2) FIFO, also referred to as Time/Price. Basically, pro rata is distributing trades at a certain price level by size ratio, that is if there are three bids at a specific price, one for 2000 contracts, the second for 200 contracts and the last at 20 contracts, the fills would be in a ratio of 200:20:2 respectively. Although this is an oversimplified explanation of algorithms that can get considerably complex, this is the basic idea.
FIFO is exactly what it describes. Bids/offers at a certain price are ranked by time, not size. Most financial futures and options contracts use a version of the pro rata method for trade matching algorithms. The Intercontinental Exchange (ICE) has it divided simply: pro rata for interest rates and FIFO for energies. Eurex uses a time pro rata for its interest rates. CME Group is nuanced, with many versions of pro rata as well as a mix of pro rata and FIFO. FIFO is used largely in the commodity products and stock indices. However, for the interest rates, pro rata is the main matching algorithm for the electronic products. As Bryan Durkin, CME Group’s Chief Commercial Officer, outlined in a letter to the Commodity Futures Trading Commission (CFTC) in 2013 responding to the regulator’s Concept Release on Risks Controls and System Safeguard for Automated Systems:
“Our Pro-Rata allocation matches fills based on the resting order’s proportional size of the cumulative bid or offer. This is especially valuable in markets with low price volatility where a FIFO allocation could result in a large limit order at the top of the book to maintain a priority position for longer periods of time, effectively blocking later entrants from receiving any fill volume. The Pro-Rata allocation enables all traders to join the queue at a particular price level and have an opportunity to compete for a fill, independent of their order’s relative time priority. Further, since time is not factored into the allocation, like-sized orders at the best bid or offer are at equal risk of being allocated a fill from an aggressor order.
In certain markets, we employ a split FIFO/Pro-Rata allocation. The FIFO component motivates traders to narrow the bid/ask spread and rewards traders who are among the first to enter orders at the top of the queue. The Pro-Rata component leads to greater participation and depth because orders other than those near the top of the time priority queue still have an opportunity to be allocated a portion of the fill. The combination of the two match algorithms therefore helps to foster tight bid/ask spreads and broad participation in the market. It is important for the Commission to note, however, that there are a variety of factors that a marketplace must consider when instituting a particular type of matching algorithm in a market, some of which include, but are not limited to, market type, bid/ask spread, minimum tick sizes, liquidity, and price volatility.”
Downside of pro rata
In itself, pro rata is not dangerous, and in fact, it was developed to be fairer to smaller traders, as Durkin explains above. However, a study by Karel Janecek and Martin Kabrhel (Matching Algorithms of International Exchanges) found: “that the Time Pro-Rata algorithm substantially complicates decision making, and, more importantly, induces individually rational trader’s behavior that is inconsistent with the general market efficiency.” Although this paper focused on splitting of orders, a practice many exchanges have banned since it was written, a newer practice is when traders, trying to get a number of contracts filled, will “pad the book,” that is, if they want 200 contracts filled at a certain price, they bid for 2000, and once they get the fill, they cancel the rest.
In the Federal Reserve of Chicago 2014 paper, “Recommendations for Equitable Allocation of Trades in High Frequency Trading Environments,” author John McPartland states: “If there is a criticism of the Pro Rata trade allocation logic, it is that many market participants are constantly bidding or offering unrealistically large quantities, often far greater than they could likely absorb.”
Illustrating his assertion is a London-based quant trader who noted in her blog, Math Trading, how FIFO allowed faster firms to get the bids/offers first as they were in line earlier, and if the market was unbalanced, she might get her order filled at that price. She states that “there are other situations when order matching algo in use and trades execution in general can become as important as the strategies/trade ideas themselves,” such as pro rata used in Eurodollar futures. She states: “If you really want a fill of X lots, you could just send an order that is somewhat bigger than X – with the extra amount being dictated by how aggressive you want/need to be – and once filled try to cancel the remaining lots. (DISCLAIMER: of course by doing this you are actively risking of being filled in all the lots, so just don’t take my word on this being a good practice and do it at your own risk.)”
Normally ‘padding the book’ doesn’t blatantly harm the market, like spoofing or quote stuffing, but it should cause concern for risk managers, brokers, and exchanges in cases of extreme moves. Of course, it is the responsibility of the trader’s firm and clearing broker to set pre-trade risk parameters. In their 2012 paper How do proprietary trading firms control the risks of high speed trading? authors Carol Clark and Rajeev Ranjan of the Chicago Federal Reserve bank found :
“Each proprietary trading firm interviewed has pre trade risk controls on the trading platform that are applied at one or more of the following levels: strategy, account and/or gateway. However, no firm evenly applies each risk check to every trading strategy. Most firms apply fewer pre trade risk checks to some strategies to reduce latency (delays). “
Despite potential holes, the exchanges appear confident their systems work. Yet, CME Group, though insisting the trading firm/clearing firm that sets the trader’s/client’s risk limit, also understand it “takes a village.” Durkin’s letter to the CFTC noted:
“CME Group supports allowing exchange clearing members to provide direct market access to their customers, provided the clearing member has appropriately vetted the client and implemented appropriate risk management controls, including exchange mandatory pre-trade credit control functionality, and the client has satisfied the system conformance testing requirements of the exchange… We feel that each level of the “electronic trading ‘supply chain’” (trading firms, clearing firms, and exchanges) must share in the effort to preserve market integrity through the implementation of effective risk controls, no matter if that participant has direct market access or is routing to the exchange via its clearing member firm.”
So is there a problem?
Many traders and brokers interviewed didn’t see a problem especially with the multiple risk checks. Logic concludes a trading or clearing firm isn’t going to relax risk parameters that could mean its own demise if the trader blows up. But anecdotally, that’s not exactly the truth. As the Fed 2012 found, sometimes things go awry, especially in today’s high-speed markets. Even the CFTC Technology Advisory Committee noted in its recommendations on pre-trade practices, clearing firms have to trust that their clients/trading firms with direct access to the exchange are following proper risk management set between them and the clearing firm because the trading firm “will generally not want clearing firm personnel examining their proprietary code.” Also, some of these risk parameters are set when the client first comes on board and checked only periodically afterward.
Joe Mazurek, president of Straits Financial, and a veteran of brokerage wars, decided with his business partner to avoid the hassle of that kind of the high-speed business. “The real impediment was the lack of risk controls with front-end boxes or manual inputs when credit controls were imposed,” he notes. “Any pre-trade check of position limits or margin simply slowed down every system we saw to a point where the HFT lost their edge. Everyone coming to us swore up and down that their programs would never allow excessive trading but being ex-risk guys, we knew that was just not true. We were and still are unwilling to allow any trading through our pipelines where we do not own the risk controls.”
A clearing firm risk manager saw the problem mainly in Eurodollars, the largest and most liquid futures market that is made up of trades typically part of much larger strategies that encompass layers of related contracts, spread months and cash. Having started in the business in the pit and now on a desk, he has seen that some internal risk people will allow a trader to put on larger than allowed size. This is “because they know in the situation that they’re in, most of [the traders] aren’t doing this on an outright basis, most are trying to spread against another Eurodollar or maybe some equivalent…and they’ll sell what’s correlated. So the [the risk manager] will allow them to do much more so the machines front end system accepts the order…that way they get a bigger portion [assigned through the pro rata trading matching algorithm].”
What worries him is in a market that has seen little to no volatility in recent years, a spike similar to what happened in the Swiss franc could be devastating with larger sized bids/offers being filled without the capital to back them up.
Why would a clearing firm risk manager allow this? Because the client can go elsewhere, and often does. Even the CFTC’s pre-trade recommendations acknowledged competition for business could force some clearing firms to try to reduce latency, and one may lose business by acting more responsibly than another.
In CME Group’s 50-page response to the CFTC’s conceptual release, which asks if risk controls should be different for systems and firms that engage in HFT from those that apply to automated trading systems in general, CME was adamant about controls being equal across all groups, adding that:
“Each market participant should be obligated to have risk systems necessary and adequate for the type of trading they employ. The risk systems at HFT firms, for example, would have to have sufficient capacity to handle the order flow generated by the firm. While the capacities of the systems may vary, the controls themselves may be identical between an HFT firm and a firm that utilizes non-HFT systems. “
That said, CME keeps a closer eye these days on games traders play as illustrated with its new Rule 575 banning disruptive trade practices. When asked directly about “padding the book,” the exchange responded, “We monitor all of the trading characteristics of our products.” Looking at recent disciplinary infractions where going over individual limits has been fined by the exchange, it seems at least CME Group doesn’t have a lazy eye and is keeping up on iterations of trader games.
Yet another problem caused by “padding the book,” (as well as spoofing, layering, quote stuffing, etc.) is the pure amount of data that must be processed, despite much of it being cancelled. Scott Caudell, CTO of Internet Infrastructure at Interactive Data, noted that when an order is placed, at least the FIFO matching algorithm has a minimal resting time. As an expert in the internal workings of the business, he believes traders should put on only the size that they want.
“If you’re going to put something out there, you’ve got to be exposed for some period of time so you actually think twice about it being out there” he says. “And you can’t keep doing these ‘cancel replaces’ putting the burden on the rest of us because we get all this extra data because of that, which is constantly growing, and we’re constantly dealing with it, which is a constant expense. The amount is a collective expense across the industry as a whole, [all] because [someone] does 100,000 cancel replaces every minute. No one ever adds that up from a cost perspective but a lot of people directly bear that expense.”
One former exchange clearing house manager agreed, stating that several years ago, the average number of Eurodollar futures contracts per trade submitted to CME clearing house was 41. Today it’s between 1 and 2, meaning the number of transactions that have to be processed is more than 35 times than it was 15 years ago. “That means there are 35 times as many trades that have to go to a quotation system, not to mention a 1000 times more quotes that have to go to the quotation systems. Bandwidth and mainframe processing scale is a really big deal,” he said.
One FCM CEO attributed the extra data to many aspects of today’s markets, not just “padding the book.” But the amount of data flowing through the system today led one manager to say “market surveillance people don’t stand half a chance. There’s too much data. It’s like trying to drink from a fire hose.”
Is there a solution?
The Fed’s paper by McPartland studies the NYSE/LIFFE Time Pro Rata algorithm and concludes that pro rata algorithms would be better using a cardinal ranking (rather than an ordinal one) on resting bids and offers “based on the actual length of time that bids and offers have been resting in the order book relative to the time that all of the other orders have been resting in the order book.” Durkin states in CME letter to the CFTC that it doesn’t use Time-weighted Pro Rata, and it doesn’t appear it plans to change that anytime soon.
To be fair, most exchanges use certain combinations of the two methods. But in markets like Eurodollars, which uses the pro rata algorithm, and could see a spike similar to the Swiss Franc with macro economic changes coming into play, problems could ensue. Still, it seems CME Group has taken precautions, providing software to aid trading firms and brokers in pre risk management, requiring kill switches and Message Volume Controls.
Further, there’s no doubt most trading firms understand the risks and have strict procedures, some fairly Draconian. One trading firm in the Clark/Ranjan paper said it considers “operational risk as its greatest threat” and among other restrictions, it holds the head of trading responsible for the trade execution and is financially responsible for any violation of risk. If there was a breach, the head trader would be 100% financially responsible for the mistake and would have to make the firm whole. Of course that was only in the case of a loss; if the violation proved profitable, the trader could not share in the proceeds. Tough love, no doubt, but may not be standard across the business.
This piece appeared originally in InsideAdvantage, a quarterly publication of Advantage Futures.
Ginger Szala is the former editor-in-chief and publisher of Futures Magazine Group. She has reported on and written about the global derivatives and managed funds business for the past 32 years. Today she is a freelance journalist, business writer and media consultant, writing for AllAboutAlpha.com, CTA Intelligence, ThinkAdvisor.com and InsideAdvantage. You can follow her on Twitter @gingerszalaink or e-mail her at: firstname.lastname@example.org