CDO Reload

Katya Wachtel, Business Insider, THEY’RE BA-ACK: CDOs And CLOs Are Popular Again As Investors Go On A Chase For Yield, here.  I read last week that the head of CS Structured Products was charged with mismarking his book. He won’t be the last structured guy to go down. The London Whale reloaded on CDX tranches, what like a year ago? So it is about time to rack them up again. This is what folks know how to do to manufacture risk and yield, so not that surprising. Maybe you can wrap them (CDOs and CLOs) in an ETF so they have a different name and trade on an exchange. High Frequency Repackaged Exchange Traded Corporate Bond/Loan Collateralized Debt Products, say it three times fast and the Themis Trading boys will just get their own show on CNBC Tee Vee.

“If you’re willing to go out more into more illiquid, structured or complex trades, there’s more opportunity, and potentially mid-teen returns,” said BlueMountain co-founder Stephen Siderow.

BlueMountain is even venturing back into CDOs, the much-maligned investment product that became synonymous with the housing bust and the financial crisis. In March, the hedge fund, purchased a portfolio of synthetic collateralized debt obligations, from French bank Credit Agricole.

The firm’s flagship $4.8 billion BlueMountain Credit Alternative fund rose about 12 percent in the first eight months of this year.

The search for yield is understandable with the benchmark 10-year U.S. Treasury at 1.61 percent and government guaranteed mortgage debt – the securities the Fed is purchasing – yielding just 1.50 percent. Even corporate junk bonds aren’t so high-yield these days, with those securities yielding 6.36 percent on Friday, after hitting a record low of 5.98 percent last week.

bps or pennies, Kid Dynamite, MacBook, IP Maven, and ETFs

Salmon, Wall Street’s preference for low-priced stocks, here. Plotting what you already kind of knew. Why do folks like BAC so much? Look at the graph across the link.

Three weeks ago, Alex Tabarrok found an intriguing post by high-frequency trader Chris Stuccio. The idea is very elegant: if you want to stop high-frequency traders extracting rents from the market, there’s an easy way to do so — you just allow stocks to trade in increments of much less than a penny. Matt Levine puts it well: right now, he says, “because you can’t be outbid by another bidder within the same penny increment, you get free money by just getting there first”. If high-frequency traders could compete on price rather than just on speed, then a lot of the silly arms-race stuff would be replaced by better prices for investors.

Interesting how much play Stuccio gets.

Kid Dynamite,  On UBS, Facebook, Nasdaq, and Erroneous Orders: A Story From The Good Ol’ Days, here.

So my boss’s trading desk goes to sell a basket of stocks while they are simultaneously buying futures on the phone with a broker in Chicago, trying to capture the spread.   They hit the SELL button for the basket.   No confirm.   So the trader hits SELL again.   Nothing.  SELL. SELL. SELL.   “Why the f*ck isn’t this order going through?”

Then just as the market begins to drop (under the weight of these waves of sell orders that were in fact going through), someone yells out “THE PRINTER IS OUT OF PAPER!“   It wasn’t that the orders weren’t going through, it was that the acknowledgements weren’t coming back!  *gulp*.

Pogue, NYT,  MacBook, a Point Shy of Perfect, here.

Superfast. Superthin. Superlight. Superlong battery life. Immense storage. Enough memory to keep lots of programs open at once. Stunning screen, comfortable keyboard, terrific sound. Fast start-up, rugged body, gorgeous looks.

Carlin had a bit that sounded like this, here.

Ron Coleman, Likelihood of Confusion, Jonathan Rogers: Is YouTube “Monetizing Piracy”? here. Pointers from Ron Coleman.

As for me, I’m Ron Coleman, a commercial litigator, business attorney and, some say, “IP maven” with a special interest in copyright and trademark infringement involving the Internet–including advising clients how to avoid them. I am also a writer and general counsel of thenotionalMedia Bloggers Association.

ETF Industry Association, here. Odd that I have not bumped into more ETF stories recently. Must be missing good information somewhere.

Some of the key highlights from the May 2012 ETF Data report include:

• Assets in US listed Exchange Traded Funds (ETF) and Exchange Traded Notes (ETN) totaled approximately $1.14 trillion at May 2012 month-end, an increase of 2% over May 2011 month-end, when assets totaled $1.11 trillion.
• ETF/ETN net cash inflows totaled approximately $4.2 billion for the month of May 2012, bringing year-to-date 2012 net cash inflows to $63.1 billion.
• At May 2012 month-end, there were 1,465 U.S. listed products, an increase of 17% compared to 1,254 U.S. listed products at the same time last year.
• Fixed Income led all categories for May with $8.9 billion in net inflows, bringing the YTD total to over $30.2 billion.

London Whale deconstructed

DealBreaker, Matt Levine, You Say “Voldemort” Like That’s A Bad Thing, here. Levine plays who’s the narc:

What is going on here? Like, for one thing: who narc’ed on him? And why? The most sensible account as always comes from Lisa Pollack; her take is basically that (1) a bunch of hedge funds are betting that the skew between spreads on the individual names in the CDX.IG.NA.9 (which names they are long) and spreads on the actual index (which they are short) will converge, (2) Iksil recently got massively long the index, blowing out that skew and losing them money on a mark-to-market basis, and (3) the hedge funds are mad and sad and going to the press to embarrass and/or regulate JPMorgan out of this market? This seems fine except that except it’s hard to see the hedge funds making money on an actual skew trade; Markit shows a -12bps skew and my sense is that after bid/ask you just can’t make a living on 12bps of convergence.

There’s a part of me that wants the narc to be JPMorgan itself, calling attention to its brilliant risk management, spooky nicknames, and ability to move markets with one flick of a London-based Frenchman. Also perhaps to provide a platform for its anti-regulatory case. 

Levine seems to get a handle on the moment and the Lisa Pollack reference seems valuable long term.

But realistically, the press has been bad, with Bloomberg going so far as to say “Neither Iksil nor JPMorgan have been accused of wrongdoing,” which, ouch! So maybe it’s other banks, jealous of how good JPMorgan’s hedging is, calling attention to the Very Important Issue of how un-Volckery and maybe-market-manipulative it is?

If so I feel like they’re … doing kind of a weak job? I will be surprised if anyone gets worked up about the market manipulation angle given that (1) the losers are eeeeevil hedge funds and (2) it’s having a fairly small effect on the market for one off-the-run CDX index. And for the Volcker Rule angle … I am serenely untroubled by JPMorgan risking $100bn on US investment grade credit, and everyone else is similarly untroubled given that there’s no real evidence that this trade (arguably a hedge, arguably long-term, etc.) would actually violate the Volcker Rule. Regardless of how you get there, though, if your model of bank regulation prohibits JPMorgan from risking $100bn on diversifid US investment grade credit, your model is wrong.

If I were writing the anti-JPMorgan PR campaign here I might come at it differently. If you take these reports at face value – and you can’t entirely; I don’t believe that JPM is long all this credit risk unhedged and neither does anyone who talked to the Journal or Bloomberg – then JPMorgan has invested $100bn of its huuuuge but finite balance sheet in US corporate credit via this trade. Unlike its loans, this extension of credit is unfunded, but still – JPMorgan is not exactly short of cash, as they’d be the first to tell you, and they can always get more if they need it. So it’s reasonable to think that JPMorgan’s ability to extend credit is finite and that is due to capital, not funding. But that $100bn of credit risk has been extended not to the 121 actual businesses in the CDX.IG.NA.9 index, many of whom probably also have too much cash but some of whom could presumably use the money to like Build A Factory or Hire Some Workers or Buy An Instagram or whatever. Instead it’s being extended to … well, indirectly, to eeeeevil hedge funds who are short the credits and churlish enough to complain about it to the press. If you’re a regulator or politician whose complaint about banks is that they aren’t doing enough lending to support the real economy, news that 5% of JPMorgan’s balance sheet is in the form of synthetic corporate lending that doesn’t actually go to those corporates might be enough to get you mad.

Trading environment seems a little more toxic than usual. Levine gives it all a gritty early 70s Popeye Doyle, French Connection feel.

Ft.com/alphaville, Lisa Pollack, Hedge funds and the Whale, credit index edition, here.  Lisa Pollack is publishing this reasonably early, 6 Apr. Look at the charts in “A graphical investigation” toward the end of the piece to get some sense of how the IG9 market has moved and on what volumes.

Zerohedge, Behind ‘The Iksil Trade’ – IG9 Tranches Explained, here.

So what was once a 3%-7% tranche is now roughly a 2.4% – 6.4% tranche.

So if you sell protection on this tranche, you need further cumulative defaults of 2.4% before you make any payments, and then you make payments until 6.4% of the notional has had losses.  If there is a 0% recovery on each default, you could have 3 defaults before having to make any payment (each name is 1/125 or 0.8%).  If recovery was 40% then you have no payments until the 6th default.

The big question is, what do you get paid on this tranche?  20 points up-front and 500 bps running.  So if you sell $1 billion of this tranche, you receive $200 million up front and $50 million per annum.  In a relatively tight credit spread environment, this is a lot of money.  If you use the upfront payment to “defease” losses, the $1 billion of exposure has a maximum loss of $800 million, and would require  4 defaults at 0% recovery before actually having a loss, and more realistically, would only take a loss on the 8th default with a 40% recovery.  Suddenly the trade seems less scary, as least to me.

But how do people come up with a number of a “100 billion”?  That comes down to “deltas”.  The delta on this tranche is about 7.5 times.  So if someone wanted to take this risk, without delta (just sell the tranche and not have a “correlation” bet), every $1 billion would create $7.5 billion of index trading.

You could sell this “no delta” and the buyer would pay you for the tranche, but then have to go and sell 7.5 times that amount of index out to the market so they could manage their “correlation” risk – a giant model based book.  Some dealers are very good at tranches, but are weak at trading the underlying index.  In those cases, you might sell the tranche “with delta” and sell the index position yourself because you can get better execution that way.  So you sell the tranche and buy 7.5 times the index from the correlation desk (the with delta trade).  Then you sell the straight index into the market.  It would explain why you are seen as a seller of index when the real trade is actually being a seller of the tranche.

Morgan Stanley, 2012 Handbook of Credit Derivatives and Structured Credit Strategies, here. 250+ page doc on credit derivatives via Levine at DealBreaker. I’ll take a look at it today.

Ft.com/alphaville, Lisa Pollack, The mystery of Morgan Stanley’s footnote unravels Part 1, here. Part 2, here. MS reduced exposure to Italy by $3.4bn while benefiting from a positive hit to net revenue of $600m. How did that work? Alternative Termination Event clauses – just like they teach in the CVA courses.

ETP Project

Setup

Assume time is discrete t1, t2, … tn and a single exchange e1 for quoting and trading to start. Domestic stocks u1, u2, u3, … ui and ETP trade on exchange e1. All currency transactions are in USD. All quotes and trades are in lots of 100. The trading desk maintains a Book of domestic equities and ETP as well as a cash account. On any given business day the Book holds no unhedged domestic equity or ETP positions at the open of trading or at the close of trading. At the end of day (EOD) every long trade is matched with a corresponding short trade and every short trade is matched with a corresponding long trade (Matched Book). The minimum non-zero absolute difference between two quotes or prices is a penny (0.01 USD). Assume all quotes and prices are strictly positive.

Define the ask and bid quotes a() and b(), respectively:

a(e1, ui, tj) ask price for party B to sell lot of 100 shares of ui at time tj at exchange e1.

b(e1, ui, tj) bid price for party B to buy lot of 100 shares of ui at time tj at exchange e1.

Define mid price m() as a function of a() and b():

m(e1, ui, tj) = b(e1, ui, tj) + (a(e1, ui, tj) – b(e1, ui, tj))/2 if a() >= b()  and

m(e1, ui, tj) = b(e1, ui, tj) otherwise.

Define the executed trade price p() from the perspective of EOD P&L accounting at tn (tj < tn), where C() is change to the cash account at execution time tj including cash, fees, and rebates attributed to the specific trade; and E() represents change to Book positions and the cash account due to EOD P&L reconciliations and amendments at tn and reallocated back to the executed trades including trade rejections and trading desk charges accrued at EOD:

p(L, e1, ui, tj) – the executed price on exchange e1 to go long 100 shares of ui at time tj.  The function P has the side effect of adding 100 shares of ui to trading desk’s Book and subtracting p() from the trading desk’s cash account.

p(L, e1, ui, tj) = a(e1, ui, tj) + C(L, e1, ui, tj) + E(L, e1, ui, tj)

p(S, e1, ui, tj) – the executed price on  exchange e1 to go short 100 shares of ui at time tj. The function P has the side effect of subtracting 100 shares of ui from the trading desk’s Book and adding p() to the trading desk’s cash account.

p(S, e1, ui, tj) = b(e1, ui, tj) + C(S, e1, ui, tj) + E(S, e1, ui, tj)

Single stock EOD P&L conditions

Since we run a matched book at EOD P&L each matched trade falls into one of two categories. Either the long was booked first or the short was booked first.

1. Long booked first EOD P&L charge:

p(L, e1, ui, tj1) – p(S, e1, ui, tj1+k1) k1>0, j1+k1 <n

2. Short booked first EOD P&L charge

p(S, e1, ui, tj2) – p(L, e1, ui, tj2+k2) k2>0 j2+k2 <n

The EOD P&L of the matched book is the sum of the above formulas selected for each pair of matched trade executions. By the definition of p() in the single stock case there must have existed a corresponding quote a() or b() in exchange e1 to derive an EOD P&L.

ETP EOD P&L conditions

For a single ETP long and short positions in a matched book we run a P&L EOD process substantially similar to the single stock P&L conditions above. The P&L process is slightly more involved if we allow (wlg) a long ETP position to be matched with a portfolio of short domestic stock positions and cash.

Assume the ETP is composed of three domestic equity stocks u1, u2, and u3 as well as a cash account c. Then

p(L, e1, ETP, tj) = m(e1, u1, tj) + m(e1, u2, tj) + m(e1,u3, tj) + c

and

p(L, e1, ETP, tj) = a(e1, ETP, tj) + C(L, e1, ETP, tj) + E(L, e1, ETP, tj)

where c is the residual cash account adjustment required to make the equality hold.

Couple of observations:

1. Trading ETP allows the desk to source the underlying long or short at mid atomically at a known cost. The desk cannot do that atomically in the single stock and single exchange case. Moreover, the desk can source underlying domestic equity at mid atomically even when the underlying spread is  0.01 USD.
2. If the spread of the ETP is tight then the spread of the underlying must tighten in proportion to the number of underlying stocks to eliminate arbitrage (think SPY).  Presumably the underlying spread tightening process is dynamic as opposed to static unless domestic stock spreads go to zero as a result of ETP spreads.

Exchange Traded Product Project

Here is how I think this project is going to go for now:

ETP Products

Run through a more detailed review of UIT, ETF, and ETN  assuming 95% of what you need to know will follow from specific examples SPY, XLF, and TZA via prospectus.

UIT – SPY,  SPY Prospectus: https://www.spdrs.com/library-content/public/SPY%20Prospectus.pdf

ETF – XLF, XLF Prospectus: http://www.sectorspdr.com/shared/pdf/prospectus.pdf

ETN – TZA, TZA Prospectus: http://direxionshares.onlineprospectus.net/DirexionShares/MOB_library/MOB_data/LIB_SummaryProspectus/DirexionETFstatPro/DirexionETFstatPro.pdf

Domestic Equity ETP/ Equity Hedges

We sort though the specific results of hedging UIT, ETF and ETN securities with one another as well as with the underlying stocks.

ETP/ETP hedges: Perfect Hedge UIT/UIT, ETF/ETF, ETN/ETN

ETP/Equity hedges

UIT/ETF hedges

UIT/ETN hedges

ETF/ETN hedges

Exchange Traded Product

An Exchange Traded Product (ETP) is an investment vehicle similar to an equity index mutual fund but trades like an individual stock on an exchange in that intraday trading, short selling, and margin financing (leveraging) are allowed. Through ETPs investors can gain exposure to almost every asset class (stocks, bonds, real estate, commodities), geographic region (U.S., international), investment style (growth or value, small cap or large), and industry sector (technology, health care, financials and more). Canada and Europe are the market leaders for ETP security structure innovation while the US lags subject to a stricter regulatory environment. This summary will focus on US equity ETPs.

NYSE lists stocks for approximately 2800 companies. The universe of domestic equity ETPs is more or less defined by the indexes or portfolios drawn from this pool of listed stocks. There are perhaps 500 listed domestic equity ETPs variously branded as: State Street’s SPDRs, BlackRock’s iShares, Invesco’s Qubes (QQQ), and Vanguard’s Vipers. State Street Global Advisors, Black Rock, Vanguard, Invesco, ProShares, and Van Eck are the largest ETP issuers. The table below displays the five top domestic equity exchange traded products from etfdb.com (here). For comparison, the average daily volume for NYSE:BAC is 306 million shares out of the US Equities daily volume of 5-10 billion shares.

Symbol	Name	Avg. Daily Volume (3M) shares	AUM (USD)
SPY	SPDR S&P 500	162.3 M	94B
XLF	Financial Select Sector SPDR	73.2M	6.7B
IWM	Russell 2000 Index	49.4M	16B
QQQ	NASDAQ-100	46.5M	33B
TZA	Daily Small Cap Bear 3x Shares	23.7M	0.8B

Note there are several listed ETPs for the same nominal index. In addition to the popular SPY ETP from State Street (but not shown in the table), there are IVV, iShares S&P 5000 Index Fund, from Black Rock which trades about 3.7 million shares daily and VOO, Vanguard S&P 500 ETP, from Vanguard which trades about 560 thousand shares daily.  The nominal underlying index is the same but the actual portfolio composition is slightly different between the various “S&P 500” ETPs. The ETPs are said generally to track a specific index rather than to necessarily replicate the index. Many of the differences between branded ETPs can be quantified by the fees and tracking error. There are several other important distinctions between various ETPs worth tracking. Most ETPs are Exchange Traded Funds (ETFs), a variant on the traditional Investment Company Act of 1940 open-end mutual fund with “exemptive relief” from some structures applied to mutual funds by the Securities and Exchange Commission. State Street Global Advisor’s XLF is an ETF as is the Black Rock iShare IWM. Exchange Traded Notes (ETNs) are not Funds but Notes issued by an investment bank committed to deliver a specified set of cashflows to the holder of the note. ETNs are typically issuer obligations pari passu with unsecured, unsubordinated debt as opposed to a direct sale and ownership of the underlying collateral. TZA in the table above is an ETN. Credit Suisse temporarily suspended further issuance of VelocityShares Daily 2x VIX Short-Term ETN (NYSEArca:TVIX) in (Feb 2012) due to internal limits on the size of the exchange traded note. SPY and QQQ are neither ETFs nor ETNs but are Unit Investment Trusts (UITs), an alternate product structure under the 1940 Act and so, for example, has a maturity date when the product can be redeemed or cancelled unlike ETFs. UITs

1. Typically engage in full replication rather than optimized sampling,

2. Cannot reinvest received dividends, and

3. Cannot engage in securities lending.

ETFs traditionally have been index funds, but in 2008 the U.S. Securities and Exchange Commission began to authorize the creation of actively managed ETFs.

The equity stock market is $22 trillion/ann. on 1.2 to 2.5 trillion shares per year. $10 trillion/ann. is traded in equity indexes of which about $6 trillion/ann. is US public traded indexes. US ETP trading accounted for a little over $1 trillion in 2011. The New York Stock Exchange (NYSE) and Arca-NYSE have extensive ETP trading sections and firms making markets in ETPs. The Chicago Board of Options (CBSX) and NASDAQ also list and trade ETPs. Other US Equity trading venues include: Bats, BEX: Boston Equity Exchange, CSXZ: Chicago Stock Exchange, DRCTEDGE: Direct Edge (Jersey City, NJ), ISE: International Security Exchange, Lava: Citigroup, NSX: National Stock Exchange (Chicago), and TrackECN: Track ECN.

All ETPs require SEC approval prior to public sale. The SEC filing determines how the ETP will operate and report results.  There are regulations covering the permissible types of underlying securities, dividend accumulation and distribution, and the taxation of those distributions. The Investment Company Act of 1940 regulates the operations of open-end mutual funds which includes Unit Investment Trusts and Register Investment Companies (Open-end ETFs and Vanguard ETF Structure). The Securities Act of 1933 regulates initial public offering of a company’s stock on the primary market to assure securities pass government standards and provide suitable investor reporting. The Securities Act covers ETNs as well as Registered Trusts (Grantor and Investment Trusts).

ETFs have been available in the US since 1993 and in Europe since 1999.

In 1998, State Street Global Advisors introduced the “Sector Spiders”, which follow the nine sectors of the S&P 500. Also in 1998, the “Dow Diamonds” (NYSE: DIA) were introduced, tracking the Dow Jones Industrial Average. In 1999, the “cubes” (NASDAQ: QQQQ) were launched attempting to replicate the movement of the NASDAQ-100.

In 2000 Barclays Global Investors put a significant effort behind the ETF marketplace, with a strong emphasis on education and distribution to reach long-term investors. The iShares line was launched in early 2000. Within 5 years iShares had surpassed the assets of any other ETF competitor in the U.S. and Europe. Barclays Global Investors was sold to BlackRock in 2009. The Vanguard Group entered the market in 2001.

Goldman Sachs, Susquehana, Getco, Citadel, Timber Hill, Knight Trading, and Interactive Brokers all cover ETP markets and secondary trading through their trading and sales desks.

ETP Algorithmic Trade Strategies include: Index and ETP arbitrage – Arbitrage ETP against underlying portfolio; Statistical Arbitrage; Market making – liquidity, volume provider; and Market Microstructure Price forecasting. For this overview let’s assume the trade strategy is ETF, ETN, and UIT high frequency simple arbitrage with the underlying index constituents.

HFT – No More Clocks from Moore’s Law

Since 2006, the clock cycles offered in new microprocessors remain constant due to fundamental power and heat dissipation constraints in silicon chip design and production (Olukotun); nevertheless Moore’s Law remains in effect and these hot power hungry microprocessors are effectively the universal compute platform of choice. Cray and Convex are history and Burton Smith works for Microsoft, game over. The good news is everyone, regardless of capitalization, computes with the “same commodity microprocessor.”

A Demon of our Own Design author Rick Bookstaber argues in his blog in 2009 that:

1. High Frequency trading is a capacity constrained trading strategy very sensitive to the number of active High Frequency traders and

2. Competition in High Frequency trading is aptly characterized as an arms race and hence a negative-sum game; the collective infrastructure spend confers no solid expectation of long-term advantage just more opportunity to spend on improved infrastructure (in The Arms Race in High Frequency Trading ).

On the other hand, in 2007 Information Week, writer Richard Martin quotes the assertion, “A millisecond advantage in trading application can be worth $100MM a year to a brokerage firm.” before concluding with the quote, “ Once you’ve got a half dozen systems that can all handle that kind of throughput, then you have to distinguish yourself somewhere else.” Protecting IP is hard, Teza Technology‘s recruiting of ex-Goldman High Frequency trading programmer Sergey Aleynikov and Tower’s recruitment of ex-Soc Gen High Frequency trading programmer Samarth Agrawal are presumably only representative known samples the actual velocity of code migration. It stands to reason that in reality the IP must move slightly faster than the code, indicating yet another way for this trade to get crowded. NYT quotes Andrew Lo in 2006 “Now it’s an arms race, everyone is building more sophisticated algorithms, and the more competition exists, the smaller the profits.”

Bookstaber, Lo, and Martin’s assessments were all made several years ago in a booming High Frequency trading market and all implicitly agree the space is destined to get crowded, implicitly sooner rather than later. Bookstaber correctly points out the figurative exits are small when the space gets crowded; Martin speculates where the exit is located (the exit is to go long, or to buy, improved computational latency).  Martin’s millisecond mark-to-market quote is pre Dodd-Frank Title VII implementation so it stands to reason that post Title VII implementation a millisecond will be worth much more than 100MM USD. So, euphemistically, how does one go long milliseconds when running Low Latency over a WAN running at or approaching known physical signal propagation limits? Moreover, given the expected crowding in the High Frequency/Low Latency business, it’s sort of important that the desk’s stack of milliseconds is larger than the competitor’s millisecond stack.

We have three variables to control in this game: Network Switch and Wire latency, Pre-trade Algorithms, and the Optimized Code/Compute Hardware. We have assumed away the wire latency at the outset of this survey; we have the lowest latency connection between NJ and Chicago by assumption. We do not know the details of the switch latency from Spread Networks, but the only reasonable question to ask is how much of a competitive advantage might accrue using this low latency WAN link?

Bandwidth does not appear to be competition driver. The High Frequency trading folks seem fairly consistent in claiming that the pre-trade decision algorithms complete (come to a decision) in microseconds. A scalar core can only touch about 64 megabits in one millisecond. Under the reasonable assumption that the algorithms are more or less scalar with respect to a given Low Latency arbitrage pair of securities/contracts, and there are only 100 or so liquid pairs, the data bandwidth requirements in a market microstructure based trading system seem modest. This limited bandwidth assumption could weaken with index arbitrage since each decision could require the time series of prices for 100s of underlying contracts – but lets stick with the limited bandwidth assumption. The only important networking factor for Low Latency trading is …wait for it … low latency, doh.

Latency benchmarks reported by Verizon show an US avg. latency of 42 ms to 44 ms going down to 33 ms to 34 ms for Private IP. AT&T reports NY to Chicago latencies of 21 ms with a nationwide average of 34 ms. The Barksdale Forbes article claims existing private lines between NY and Chicago running 16.3ms roundtrip latency.

Assume best possible case, the desk is on the Barksdale’s fiber and all competitors are running private connections with round trip latencies between 16.3ms and 21ms. So the desk’s one-way latency is 7 clocks versus the competitor’s 9 to 11 clocks. Any advantage should manifest itself in remote event notification such as: trade notification; order book updates; and exogenous trading system shutdown notice.

In SPY/SPY1D arbitrage the desk will see anywhere from 1 to 2 more time series points corresponding to executed orders from the remote opening market than the competition. In terms of event notification a competitor could run event driven rather than synchronously and recapture a trading system clock cycle of latency. Longer term though I think you have to assume the synchronous trading system design prevails over the event driven trading system design for similar reasons that synchronous circuits beat out asynchronous circuits due to complexity in handling signal race conditions. To the degree that the pre-trade algorithm execution time allows a higher trading system clock rate, the capacity of the event driven trading system to catch up to a synchronous system in event notification is proportionally decreased.

The updated remote order book, similar to the executed order notification, will run a couple of milliseconds ahead of a disadvantaged WAN competitor. Its not immediately clear that this advantage is as valuable or actionable as the trade execution notification. Perhaps there is some automated mechanism to determine if one market is leading another in price discovery for a given contract then if the updated remote order book is from the market leading price discovery there is some concrete advantage.

Shut down notification latency is always important but even more so if the desk starts to run strategies that are directional or not risk-neutral. Thorp described the scene at Princeton-Newport when a surprise merger was announced and the Stat arb trading system needed to shut down immediately; Low Latency prop trading has the same problem. Event driven communication is better than the synchronous model in this case. Long-term it might make sense to keep an alternate WAN channel around simply for asynchronous notifications for system exceptions.

Assuming the pre-trade algorithm is fixed then we do sort of know that code optimization and hardware selection are the keys to getting long milliseconds, reducing computational latency, and beating the competition. The key to code optimization now is the observation that contemporary microprocessors are little superscalar and superpiplined parallel machines. The job of the floating point programmer is to deconstruct the pre-trade analytics into an executable form that keeps the floating point units busy, the FP pipeline running with the fewest bubbles (idle pipeline stages) possible, and doesn’t miss in the cache too often. For typical Financial Engineering codes you can get good estimates of optimal core cycle executions times and drive the analytics performance reasonably close to the optimal core cycle count. On the hardware selection side, the job is to stay on the Moore’s law technology wave (could be any technology using the latest silicon fabrication generation: on-chip mp, FPGAs, or GPUs) without losing too much software support from optimizing compilers, vectorized math libraries, and multiprocessing code generation.

The fewer core clocks required by the pre-trade algorithm execution the faster we can run the trading system clock in the synchronous model.  As the trading system clock cycles get faster the ability to tolerate off chip communication for parallel computing will dissipate, so the main opportunity for parallelism will be on-chip.

If the pre-trade analytics is heavily dependent on IEEE 754 double precision floating point execution then the desk needs to use native compilers and native libraries for the chip running the pre-trade algorithm (see the Intel or IBM math library maps of operations to clock cycles and precision). The shipping 4.25 GHz, 8-way mp, 45nm IBM POWER7 with 4 double precision floating point units per core, running XLC code on top of the MASS vector libraries is probably slightly faster than even the upcoming 3.9 GHz, 4-way mp, 32nm Sandy Bridge Intel chips running ICC code on top of the MKL vector libraries for conventional double precision quantitative analytics commonly encountered on Wall Street. This is less of a vendor allegiance point (although it might sound otherwise) and more of just keep track of the smart people that know this particular stuff point. Long-term relative value between the Intel and IBM infrastructure floating point performance will hinge on the native compiler assist with on-chip parallel processing. Right now, IBM will throw you bigger caches, more independent floating point units, and a higher frequency clock than Intel. On the other hand, Intel’s microprocessor feature size runs almost one generation ahead of IBM and the Intel compiler folks and math libraries are quite competitive. If you look at SPECfp where vendors display the maximum floating point execution speed of their products, they will only quote native compiler executions (see SPEC CPU FP). There is no reasonable expectation I am aware of that a non-native compiler (or interpreter) is going to issue code that can run to speed after deconstructing the code and estimating the code cache footprint end-to-end. Moreover, the only optimized math libraries I have seen are for native compilers. On the other hand, if single precision is tolerable for pre-trade analytics there are several custom FPGAs  (see Wallach’s most recent startup Convey) and GPU computing options (NVIDIA Tesla) that could be potential competitive sources of compute power for pre-trade analytics, given sufficient compiler and math library support.

If the premium is on reducing computation latency then accounting precisely for core processor clocks in pre-trade algorithm code is important. It may be pragmatically reasonable to assume the computation execution is effectively scalar given that the parallel computation support even on the native compilers is still rather new and raw. The parallelization opportunities available currently are on-chip, with multicore execution (probably 2 to 4 way) moving to 8-way with newer silicon (see for example the IBM Power7).