The Gateway to Algorithmic and Automated Trading

Just Showing Up "On Time" Isn't Enough

Published in Automated Trader Magazine Issue 22 Q3 2011

Recently, many of our client conversations have focused on high-frequency latency arbitrage. The concern: many order routers do not compensate for differences in the distance to the various market venues. Thus, orders to several market venues sent at the "same time" arrive at their destinations at different times. These discrepancies may provide an opportunity for high-frequency algorithms that make markets simultaneously on several exchanges to exploit timing differences by cancelling some orders after the first one is filled.

Recently, many of our client conversations have focused on high-frequency latency arbitrage. The concern: many order routers do not compensate for differences in the distance to the various market venues. Thus, orders to several market venues sent at the "same time" arrive at their destinations at different times. These discrepancies may provide an opportunity for high-frequency algorithms that make markets simultaneously on several exchanges to exploit timing differences by cancelling some orders after the first one is filled. For example, an order router that sends 4 X 1,000 share orders to four different exchanges may only get filled on 1,000 shares because the HFT, once executed on 1,000, cancels the other three orders. The resulting "Liquidity Fade" manifests itself in lower fill rates (in this example 25%, 1,000 out of 4,000 shares), thereby introducing opportunity cost and slippage in average prices.

Bloomberg Tradebook clients do not need to be concerned with this order synchronization issue. Tradebook maintains its own high-speed network of connections to each venue from our SOR proximity host. Instead of simply firing all orders simultaneously, our order routers seek to have orders arrive at the venues simultaneously. Moreover, when we "miss" on the synchronization, we measure the differential of order arrivals in micro-seconds. The result: Tradebook has median fill rates of 100% and averages 95-97% per execution venue.

In the aggregate, between 65-75% of the market's liquidity is traded at an exchange. Order routers that mainly manage order arrival are a limited solution. They are only truly effective in cases where order size is less than or equal to the known display size, and they do not find and extract the significant amount of liquidity that is contained in the exchange's dark order types (hidden orders, Reserve, Discretion). We believe that there is more to order routing than arrival time. True smart order routers also address the challenges associated with smart quoting passive orders and when routers are aggressive (cross the spread) they intelligently seek to find and extract liquidity that is hidden in dark order types, Reserve, and Discretion.

Figure 1
Figure 1

The True Picture of Market Fragmentation

The U.S. marketplace is a complex landscape, with 14 exchanges and about 44 dark pools. In addition, no liquidity pool commands more than 20% aggregate market share. When the marketplace is broken down by listing exchange (NASDAQ OMX and NYSE Euronext), market share rises but no centre can be said to be dominant. Physically, liquidity is evenly distributed across three geographical locations in four data centres, each with about one-third of the liquidity (Figure 1).

The "liquidity poles" are the NYSE Euronext group in the North, NASDAQ OMX Group in the South and Direct Edge and BATS are in the middle (although we are focusing on the exchange-based liquidity, it is important to note that a majority of dark pool liquidity is also located in the middle). Each exchange venue has both displayed liquidity and a significant amount of non-displayed liquidity. Rosenblatt Securities analyst Justin Schack estimates that about 4% of the market's aggregate executed liquidity is held in exchange dark order types such as "hidden" orders. Bloomberg Tradebook estimates that anywhere from 10-to-20% of the market's liquidity is held in Reserve (Iceberg).

This market configuration, has led some brokers, like Bloomberg Tradebook and some HFT players, to develop a "proximity hosting" solution - place the SOR at a location equidistant from the different data centres rather than co-located at each one to aggregate the market's liquidity.

Routing 101: Order Arrival Is Limited

An order router can manage an order size that is either: (1) less than or equal to the display or (2) greater than the display. If the order size is less than or equal to display, then the first round of orders to the different market venues should complete the order. However, larger orders can only be completed with an initial volley and then subsequent re-fires. For example (Figure 2), if the total size of the order is 3,000 shares and the trader is price sensitive (e.g., wanting to trade only at the BBO by choosing a Bloomberg Tradebook "PRICE" sweep style), then the first round of sell orders will only be sent at 10.51 for 1,000 shares. Note, Bloomberg Tradebook has put in place the significant infrastructure needed to send ISO- Reg NMS intermarket sweep orders-that enable to the smart order router to send orders across multiple price levels in the first volley. In this case, traders leveraging Tradebook would be able to send "LIQUIDITY" sweep orders to a 10.50 limit and attack all the liquidity in the price band simultaneously. Not all order routers have this capability.

Order synchronization becomes less effective after the first volley of orders because notice of execution messages come back from the different exchanges at different times-Why the time differential? Each exchange has different internal processing times- the amount of time needed for them to take in an order, process it and send the notice of execution. Although you can manage when the orders arrive, what is out of your hands is when the exchange reports back to you on the timing or quantity of the execution. The busier the exchange is, the slower or more variability in the response time.

Round II: Slippage Management

In our example, the first volley of orders sought to execute 1,000 of our 3,000-share order. Let's assume that the first round of orders has been sent. Executions start to come back piecemeal. So, the order router is now faced with a difficult decision: pause or continue. To try to synchronize after each volley, the order router must fire, pause for all responses and then re-fire.

Aks Size Mmkr Total
10.51 200 NSDQ 200
10.51 100 BATS 300
10.51 300 NYSE 600
10.51 400 ARCX 1,000
10.50 600 EDGA 1,600
10.50 200 BATS 1,800
10.50 300 NSDQ 2,100
Figure 2

Thus, if the order router pauses for notice of all executions on its initial volley before re-firing, then the price slippage could be substantial. Note: Even though there may be slippage in execution quality, the fill rate benchmark can remain very high (high = 90+ percent).

The market, however, doesn't sit still while the router is waiting. At this point slippage is introduced into the strategy-while the order router has paused, waiting for all the responses to come back from the other venues, the market can move in the direction of the trade, since: (1) others in the market may be trading in the direction of your trade (it would be naïve to believe yours to be the only buy or sell order in the market); (2) the HFT algorithms that you just aggressively hit will most certainly and immediately react as they very well may get their notice of execution much faster than you do if they are closer to the exchange; and (3) other algos and market participants may discover your trades on the consolidated tape or through quote updates and react to your initial volley of orders before you do because the order router is pausing waiting for the slowest exchange to respond.

Tradebook, however, does not wait for all the notice of executions to come back from exchanges to re-fire. Instead, we do an intricate dance to maximize liquidity capture by trying to find hidden liquidity in the first volley and looking at the notices of execution and quote updates to determine an optimal re-fire strategy.

Where's Waldo? Finding the Liquidity Lurking Underneath

Focusing on minimizing liquidity fade and counteracting perceived HFT latency advantages is just a small part of meeting the total smart order routing challenge. Smart Order Routing (SOR) isn't just about optimizing the yield on displayed liquidity, it also encompasses efficiently finding and extracting all of the liquidity on an exchange. When your DMA order or algorithmic trading strategy decides to be aggressive and cross the spread, the SOR must efficiently find and extract all the liquidity on the exchange. Orders firing only at display sizes and showing up at the same time aren't enough. A SOR must, as Tradebook's system does, factor in the significant amount of liquidity held in exchange dark order types, Discretion and Reserve, which can be found and extracted by optimally sweeping, firing and re-firing subsequent orders.

Two facts about trading micro structure: (1) orders and notices of execution travel faster than quotes updates and (2) the execution provides significant information. Tradebook manages a dynamic matrix of "expected liquidity" by stock, venue and time of day based on history and real-time success rates. Thus, on the first volley of orders, Tradebook doesn't send only for display size, rather it sends for display size plus what our statistics suggest is lurking beneath the surface. An optimally oversized order is sent in the first wave; since execution speeds are faster than quote updates and prints to the consolidated tape, Tradebook's smart order router has a speed advantage because it knows whether or not more hidden liquidity can be found at the destination.

Let's assume that against a display of 200 shares of a given stock at NDAQ, the dynamic statistical matrix tells us that at this time of day for this stock, on this exchange, usually 2,000 shares are not displayed. In this situation, Tradebook's smart order router would send 2,300 shares to the venue. If we execute less than 2,300 shares, then we know that we have taken all the liquidity at this level. Despite what the quote looks like, the opportunity cost of this transaction is decreased because we know immediately that nothing is left at that price level at that venue. In contrast, if all 2,300 are executed, then we know that more liquidity at that price level is present. At this point, the Reserve-HunterSM algorithm kicks in, and we immediately start to oversize re-fires back to NDAQ to optimize the capture of the rest of the non-displayed liquidity. In both cases, the added piece of information that is gained from how Tradebook engages the market enables the smart order router to optimize the re-fire.

Additionally, since approximately 4% or more liquidity is held in hidden orders at the exchanges, we have to consider that a venue could be holding a hidden or discretionary order. In the example above (Figure 2), at 10.51, no displayed liquidity is present at EDGA. So, the Liquidity-HunterSM algorithm will send some shares to EDGA. If we discover liquidity from the probe, Reserve-HunterSM then kicks in and starts to oversize to efficiently extract the remaining liquidity.

True SOR also includes Being Smart with Passive Orders

This note only scratches the surface of the complex issues and decisions that must be considered and made for true Smart Order Routing. For example, most participants view SOR as what happens when you cross the spread. However, only 60% of orders are aggressive - cross the spread; a significant 40% of orders are passive. Many times brokers will place passive orders in the venue that has the highest rebate. The challenge for passive orders is being traded around. The order is in one exchange when the seller is in another. So, a true Smart order router has to include smart posting algorithms in addition to smart aggregation.

For a more complete discussion, including the regulatory and technological challenges to sweeping multiple price levels at the same time for more aggressive liquidity capture, Liquidity-HunterSM, and Reserve-HunterSM, please get in touch with our Execution Consultants on the Tradebook Desk and go to our website and download "What Are the Components of True U.S. Equity Smart Order Routing?"

Bloomberg Tradebook is a global agency broker offering advanced trading algorithms and direct market access to over 60 global equity, futures, and options markets and 41 currency pairs in our Foreign Exchange marketplace. Many traders have created valuation, investment and trading strategy models in various applications and nourish them the Bloomberg Professional® service data API. Now, using the same connectivity as Bloomberg's data API, traders can integrate their strategies with Bloomberg Tradebook's high performance Order API and connect their strategies to the electronic execution highway.

Nothing in this document constitutes an offer or a solicitation of an offer to buy or sell any security or other financial instrument or constitutes any investment advice or recommendation of any security or other financial instrument. BLOOMBERG TRADEBOOK believes that the information herein was obtained from reliable sources but does not guarantee its accuracy.

Communicated by Bloomberg Tradebook Europe Limited, registered in England & Wales No. 3556095, authorized and regulated by the UK Financial Services Authority No. 187492. This communication is directed only at persons who have professional experience in the investments which may be traded over the systems and certain high net worth organizations. Available in all countries and is only offered where clearance has been obtained.

Bloomberg Tradebook LLC member of FINRA ( SIPC/NFA. Bloomberg Tradebook do Brasil is the representative of Bloomberg Tradebook LLC in Brazil registered with the BACEN. Bloomberg Tradebook Services LLC, Bloomberg Tradebook Australia PTY LTD ABN 36 091 542 077 ACN 091 542 077, Bloomberg Tradebook Do Brasil LTDA., Bloomberg Tradebook Canada Company Member of CIPF, Bloomberg Tradebook Limited, Bloomberg Tradebook Hong Kong Limited the first ATS authorized by the SFC-AFU 977, Bloomberg Tradebook Japan Limited member of JSDA/JIPF, Bloomberg Trading Services Japan LTD, Bloomberg Tradebook Singapore Pte Ltd Company No. 200104338R, Bloomberg Trading Services (Singapore) Pte Ltd Company No. 200101232G. Bloomberg Tradebook Bermuda LTD, licensed to conduct Investment Business by the Bermuda Monetary Authority.

service marks of Bloomberg Finance L.P.("BFLP"), a Delaware limited partnership, or its subsidiaries.
Bloomberg Tradebook is provided by a BLP subsidiary, Bloomberg Tradebook LLC, and its affiliates and is available on the BPS.