The Gateway to Algorithmic and Automated Trading

Combining Algorithmic Trading with SOR

Published in Automated Trader Magazine Issue 15 Q4 2009

Smart Order Routing (SOR) has been the buzz phrase in European equity trading over the last 18 months. Deregulation of trading venues allowed Multilateral Trading Facilities (MTFs) to trade stocks listed on other exchanges. Consequently trading service providers started offering SOR products to allow clients to trade the same stock on MTFs as well as the original exchanges where stocks are listed (the primary exchanges). However, SORs are limited in how much trading fee reduction and price improvement they can obtain. Dr Usman Malik, partner of algorithmic trading specialists PE Lynch, outlines how combining SOR with algorithmic trading strategies offers a method to overcome the limitations of standalone SOR.


MTFs allow a trader to use more than one venue to trade the same stock. There are two major benefits to trading on MTFs:

1.Trading Fee Reduction

The new MTFs are charging a much smaller trading fee compared to the primary venues and in many cases offer a maker taker model to establish the transaction fee. The maker taker model is one where the person offering liquidity to the orderbook (a passive order) gets a rebate paid by the person taking liquidity from the orderbook (an aggressive order). Also the aggressive fee on MTFs is generally lower than on the primary exchange.

2. Tick Size Reduction

When they first launched, MTFs offered smaller tick sizes compared to primary venues. This has led to competition on tick sizes between venues including the primary exchanges. In many cases the effective spread of the stock has been reduced. Consequently the price clients execute at has become closer to the stock mid-price, a benefit that clients have received irrespective of whether they have access to trade on MTFs.

Since passive and aggressive orders incur different trading fees it is worth defining how they can be achieved:

An aggressive order takes liquidity from the orderbook by executing either as a market order or as a limit order that pays the spread. A limit order that pays the spread would be a buy order where the limit price is the best ask price or a sell order where the limit price is the best bid price.

A passive order adds liquidity to the orderbook by setting a limit price so that the order does not pay the spread and only executes once an aggressive counter party order matches the limit price.

Buy Order Type Spread Captured Label on diagram
Primary Passive Order Primary Ask - Primary Bid A
MTF Passive Order Primary Ask - MTF Bid B
MTF Aggressive Order Primary Ask - MTF Ask C
Primary Aggressive Order 0 -

Price Improvement via MTFs (Spread Captured)

Multiple trading venues potentially allow improvement in execution price when compared to the primary exchange. SOR providers define spread captured in order to show how trading on an MTF can sometimes lead to price improvement. Spread captured is the difference between the execution price of the order and the primary crossing price. The table below shows the spread captured for different buy order types.
It is interesting to note that a trader can capture spread without using an MTF, by simply using a passive limit order routed to the primary exchange. However, the reason for routing passively to an MTF as opposed to the primary exchange would be that the probability of getting filled passively may sometimes be higher on an MTF. Higher passive fill rates on an MTF are due to the lower trading fee and smaller tick sizes attracting liquidity. One problem with spread captured is that it is not clearly defined for passive orders. For aggressive orders we execute immediately so spread captured is relative to the arrival primary crossing price.

Primary Best Ask....

However, for passive limit orders it is possible to wait for several minutes before getting filled at the limit price during which time the primary crossing price may change. Some SOR vendors quote spread captured for passive orders relative to the crossing price at the execution time of the order. The spread captured for passive orders should be measured relative to the arrival crossing price, as this is the execution opportunity being rejected when the decision is made to place a passive limit order.

There are two questions that every SOR provider should be able to answer:

1. How much are the trading fees reduced by using the SOR instead of trading aggressively on the primary exchange?

2. How much is the execution price improving by using the SOR instead of trading aggressively on the primary exchange?

The answer to 1) is if an SOR trades aggressively then the trading fee saving is just the difference between the aggressive fees on the primary exchange and on the venue where execution occurred. If the SOR trades passively then the trading fee saving is just the difference between the aggressive fee on the primary exchange and the passive fee/rebate on the venue where execution occurred. However, for passive orders the passive fill rate (partially determined by the user limit price) will determine how often the SOR can achieve a trading fee saving.

The answer to 2) is the price improvement is just the spread captured for both aggressive and passive orders. However, for passive orders the passive fill rate will determine how often the SOR is capable of capturing spread.


Vendors currently offer stand-alone SOR facilities where a client can submit an order which is routed to the best trading venue or split between venues. There are various strategies and parameters available for different vendor SORs; however, all trades will execute as either aggressive or passive orders.

SORs are good for aggressive orders i.e. those that cross the spread immediately. Given an order to execute immediately on any trading venue, the SOR logic is well defined: go to each venue in order of best price and then in order of liquidity available at that price. For aggressive orders, execution is immediate as long as liquidity exists.

SORs have a more complicated task for passive orders for two reasons. Firstly, a passive SOR order will often come with a user limit price. The ability for the order to get filled passively will then be determined by the user's skill in selecting the limit price. The second reason is that the best venue is not immediately obvious. For passive execution the SOR requires knowledge of passive fill probability on each venue (a parameter which varies with time on each venue) in order to route the order.

For passive orders execution is not immediate or guaranteed, so it would be useful to consider the passive fill rate (the number of orders filled passively relative to the number of passive orders submitted to the SOR).

The passive fill rate for an SOR is determined by the user limit price and the SORs ability to measure passive fill probability on each venue. It is difficult for SOR providers to quote their passive fill rates across multiple venues, since it is partially determined by the actual user. It is an inability to achieve high passive fill rates across multiple venues (and generate rebates) that appears to be the biggest criticism of SORs.

MTFs give the most benefits when passive execution occurs leading to a bigger trading fee saving (often a rebate) and a better execution price. Currently many SOR products are not capable of achieving a high passive fill rate and are limited in providing the benefits of MTFs to the client. In order to pass more of the benefits of MTFs to clients a method must be found to improve passive fill rates.
Algorithms with SOR
The orders being sent to an SOR are usually child orders, i.e. they are smaller parts of a bigger parent order which the trader is working over a period of time. Algorithms assist the trader by using historical statistics and live market data to automate the process of splitting up the parent order.

SOR systems are currently used to route orders to the exchanges and MTFs. For each order generated by a trader, a decision is made on which is the most effective destination for the order. An order can also be split between destinations. SOR systems can be applied in the same way to child orders generated by an algorithm. However, in the case of algorithmically driven orders, there are quite dramatic additional benefits to be had by incorporating the routing logic into the algorithm itself.

Combining SOR technology with algorithms leads to a very high passive fill rate. Algorithms are dynamic
i.e. they manage the orders on market on every price tick and update limit prices and venues to improve the passive fill rate. Additionally the combined algorithm and SOR can control the opportunity risk associated with a passive order. In other words, a standalone SOR cannot determine how long an order should wait passively before becoming more aggressive. Opportunity risk can become very important when the client wants the parent order completed by a certain time. If the two systems are combined the order can be scheduled and priced so that the opportunity risk of passive orders can be controlled by the algorithm. Essentially the algorithm combined with the SOR can try to maximize spread captured whilst limiting opportunity risk.

It is worth giving a very simple example to show the advantages of a combined algorithm and SOR compared to standalone SOR. Consider a parent order with two passive child portions on market, one portion on the primary and the other on a MTF. If the algorithm decides to cross one portion (become aggressive in order to reduce opportunity risk) it can look at which of the two portions is most likely to get filled passively along with the relative exchange fees for becoming aggressive on all venues. Consequently the algorithm can cross the portion which it statistically believes is less likely to get filled passively on the cheapest venue. Alternatively, if these two portions were just sent to a standalone SOR there would be no comparison done between the two portions as the SOR does not know they belong to the same parent order. Thus a standalone SOR system, however well tuned, always trades at a disadvantage when compared to a system that is aware of the parent order instructions, and hence has knowledge of the combined execution responsibilities of multiple child orders.

Overleaf are two examples of actual orders which used a VWAP algorithm combined with SOR leading to high passive fill rates.

Figure 1 shows the price movements of the primary venue (Italy) and Chi-X for a buy order in Mediaset Spa between 14:00 and 16:20. The tick sizes on both venues are the same. On Figure 1 the fills are also displayed for the order. It can be seen that algorithm fills 92% of the orders passively (23 fills out of 25 were passive).Also, this order demonstrates how the algorithm tries to be passive at the beginning of the order, the 2 aggressive fills occurred very close to the client end time.

Figure 2 shows the price movements of the primary venue (Virt-X) and Chi-X for a buy order in CSGN between 15:35 and 16:20. The tick sizes on both venues are different. It is possible to see when each venue is offering a better price. On Figure 2 the fills are also displayed for the order. It can be seen that algorithm fills the order 78% passively (7 fills were passive out of 9 total fills). It can also be seen that this order never filled aggressively on the primary venue taking advantage of better prices and lower trading fees on the MTF.

Figure 1: Mediaset VWAP buy order

Figure 1: Mediaset VWAP buy order

Figure 2: CSGN VWAP buy order

Figure 2: CSGN VWAP buy order

Algorithms with SOR - improving trading fee

A good VWAP algorithm would expect to get filled passively at least 70% of the time on a primary exchange. So a VWAP strategy is a natural way to generate rebates when combined with SOR logic. In a system which is specifically tuned to seek passive fill opportunities on the MTFs, passive fill rates in the region of 90% are achievable.

Algorithms with SOR - improving execution price

The question is how much price improvement can algorithms achieve by having access to MTF liquidity? It is important to note that this analysis cannot be done on a single order basis. For example we might have a parent order where the overall execution price was 100, with 50% executing on the primary exchange at 100.01 and 50% executing on an MTF at 99.99. It is not possible to conclude that a further price improvement in the overall price could have been achieved if more of the order had been routed to the MTF as we may cause more impact on the MTF.

So what is the best way to measure price improvement an algorithm can achieve by having access to MTF liquidity? The answer to this comes from comparing the average VWAP performance for algorithmic orders where everything was executed on the primary venue vs. algorithmic orders where SOR was used. It is important to only compare orders on stocks that are tradable on MTFs otherwise a biased result may be obtained.

In the current market conditions, we observe a typical improvement of 20% on VWAP for MTF enabled orders compared with a similar sample of non MTF-enabled orders. So if the average VWAP performance was -1 bp for orders on the primary then it would be -0.8 bp with MTF-enabled orders. It is worth noting that the rise of the MTFs has provided the additional benefit of downward pressure on tick sizes. This has facilitated a further improvement of around 10% on VWAP which is observed by all orders, regardless of whether MTF execution is available. Recently the different trading venues appear to have come to an agreement on standardizing tick sizes, so it will be interesting to see if further price improvement occurs.


MTFs give traders the opportunity to reduced transaction fees and trade at better prices. SOR technology allows some of the MTFs benefits to be realized. Combining SOR with algorithms allows more of the benefits of MTFs to be passed onto the clients, notably increasing passive fill rates which generate rebates and a better average execution price.