The Gateway to Algorithmic and Automated Trading

Barclays Mix'n'Match

Published in Automated Trader Magazine Issue 16 Q1 2010

Toby Bayliss and Andy White, at Sanford C. Bernstein, illustrate the use of multiple complementary algorithms to work a substantial hypothetical buy order in Barclays in the immediate aftermath of the US president’s announcement of his vision for the banking industry.

Toby Bayliss

Toby Bayliss

Andy White

Andy White

The Scenario: It is January 22nd 2010 and a trader at an agency brokerage has recently been handed a substantial Barclays buy order from a large long-only European fund manager to work. The fund manager is taking a two year view on the bank and despite its recent weakness feels that its earnings prospects for the medium term are positive.

However, recent news has not favoured the stock's immediate outlook; banks in general have been weak and Barclays in particular is already under a cloud as a result of comment from a major investment bank earlier in the week. The investment bank in question announced that it was reducing its price target for Barclays to 350p from 400p. Its analysts stated that the target price reduction was largely made in order to reflect the impact of government levies and the likelihood of Barclays finding itself with a sizeable capital deficit as a result of new Basel Committee requirements regarding the bank's Tier 1 ratio.

However, a much more significant event likely to affect the stock during the current session is President Obama's announcement the previous day of his intended introduction of "Glass Steagall II",which would once again separate commercial and investment banking activities. Barclays was identified by some analysts as the UK bank most vulnerable to these proposed changes, with some 9% of its total revenue being potentially at risk.

The trader has considerable discretion with the fund manager's buy order and given the recent weak market for Barclays is hoping to improve on the allotted limit price of 272p.

The Asset: Barclays PLC Ordinary Shares 25p

The Challenge: To buy 15 million shares of Barclays with a limit price of 272p during the trading session of Friday January 22nd.

The Algos: The trader is expecting to use a combination of three algorithms, depending upon the circumstances prevailing at the time. These algos are:

An adaptive inline algo that scales participation rates in line with the flow of the markets. The algo becomes more active as the stock price become more favourable (in a non-linear manner) and aims to beat the arrival price.

The algo is best suited for use when the user is sensitive to price movement and willing to deviate from a rigid participation rate. Its aggression level enables the user to define the rate of change in participation as price becomes more or less favourable. With higher aggression, a greater percentage of the order is released to matching pools.

A liquidity aggregation and capture algorithm, which (in addition to active liquidity aggregation) also attracts liquidity by passively posting iceberg orders in the visible order book. The aggression level of the algorithm is controlled by its trading style parameter. In its most aggressive form it will seek to capture in excess of 75% of visible market volume while continually sweeping all available dark venues and seeking hidden liquidity inter spread. If hidden or dark liquidity is discovered at the right price level, the algorithm will seek to take full advantage with the aim of completing the order. In its less aggressive form, the algo's interaction with lit markets becomes more opportunistic.

A dark and hidden liquidity algorithm that aims to minimise market footprint; it intelligently represents orders in dark pools and hunts for hidden liquidity between the spread. The algo's venue selection is optimised to minimise the opportunity cost and risk of being gamed. Orders are split across multiple venues to increase the probability of a match, with additional liquidity being directed to the most successful, but with the full order size never exposed to any single dark pool. The size released to individual venues is continually adjusted, based on short term alpha models and controls, that limit the risk of being gamed. Venues and individual executions are monitored in real time for any sign of gaming.

The Trader: The trader, who is based in London works for an agency only broker. His typical clients are large long-only buyside institutions. In addition to in-house research, he uses short term technical indicators to highlight likely exhaustion and turning points as an aid to real time algorithm selection and deployment.

7.30am: All the indications are that Barclays will open gap down and probably continue to slide. On that basis, the trader opts to set up the inline algorithm for deployment in the opening period with a price limit of 272p and with minimum/maximum volume participation thresholds of 5% and 25% respectively.

8.00am: As predicted, Barclays gaps down on the open some 3% (see red rectangle highlight in Figure 1) but initially remains above the limit price on the inline algorithm. Volume completed - 0.

Figure 1

Figure 1

Source: CQG, inc. © 2010. All rights reserved worldwide. www.cqg.com

8.01am: The stock has fallen below the limit price of the inline algorithm, which is now active. Volume completed - 13,235.

8.14am: Stock has moved in and out of limit, the in line algorithm has passively picked up liquidity. Volume completed - 145,236.

8.15am: A strong rally has seen Barclays back significantly away from our limit. Inline algorithm out of limit and inactive.

9.04am: Market has come off significantly in the past fifteen minutes from a high of over 278, briefly coming back in range but then starts to rally once more. Volume completed - 154,138.

9.08am: Barclays is now back in limit for the inline algorithm.

9.55am: For most of the past forty minutes the market has been fluctuating to and fro across the limit prices of the inline algorithm. Volume completed - 953,204.

Figure 2

Figure 2

Source: CQG, inc. © 2010. All rights reserved worldwide. www.cqg.com

10.45am: The trader's view from monitoring short term indicators is flagging the probability of a significant further decline (see red rectangle highlight in Figure 2). The trader responds by setting up the liquidity aggregation and capture algorithm. He layers the order book with orders from this algorithm in 2p increments starting at 266p. On the basis that bank stocks can slide rapidly on bad news and generally negative sentiment (such as the weak December retail figures announced earlier) he extends these layers all the way down to 250p. Volume completed - 1,651,666.

10.51am: Share price is drifting lower and has just broken previous low for the day. Volume completed - 1,771,892.

10.56am: Price has broken the 266p level and the first of the liquidity aggregation and capture algorithm orders has been hit capturing the vast majority of dark, hidden and lit liquidity. Volume completed - 2,585,352.

11.22am: Frustratingly, Barclays has remained above the 266p level for most of the past half hour, so no further action from the liquidity aggregation and capture algorithm. However, the inline algorithm has continued to accumulate stock. Volume completed - 2,825,645.

11.38am: The market is now below 264p; second layer of liquidity aggregation and capture algorithm orders have been hit. Volume completed - 4,465,879.

11.40am: Seeing the high dark fill rates achieved in the inline strategy the trader turns on the dark/hidden algorithm with a limit of 267p and turns off the inline algorithm hoping to reduce impact. Volume completed - 4,662,365.

11.50am: Market has dipped below 260p. The trader's short term indicators highlight that the current decline may be over-extended and likely to reverse (see red rectangle highlight in Figure 3). Volume completed - 8,221,879.

Figure 3

Figure 3

Source: CQG, inc. © 2010. All rights reserved worldwide. www.cqg.com

11.56am: The market appears to have made a recent bottom at around 255p and is now rallying strongly. The liquidity aggregation and capture algorithm's orders down to 256p have been hit. Trader reloads algorithm with order at 256. Volume completed - 9,532,852.

12.21pm: The market has continued to rally hard and is approaching the 268p level. Volume completed - 10,988,233.

12.51pm: A significant fall has taken place over the past half hour, market is now just above 262p. Volume completed - 11,497,233.

1.00pm: The short term indicators suggest that a rally is imminent (see red rectangle highlight in Figure 4). Volume completed - 11,637,565.

Figure 4

Figure 4

Source: CQG, inc. © 2010. All rights reserved worldwide. www.cqg.com

2:10pm: A rally is underway and the stock is now above 266p. Nevertheless, the dark/hidden algorithm has been highly productive in the past thirty minutes and all the liquidity aggregation and capture algorithm's orders have been filled. Volume completed - 12,198,258.

2.22pm: Price has just risen above the dark/hidden algorithm's limit; the trader nevertheless opts to leave it running in case of further pullback. However, he also allocates some volume to the inline algorithm with a 270p limit. Volume completed - 12,983,654.

2.33pm: Price now out of limit on inline algorithm. Volume completed - 13,103,951.

2.40pm: The mean reversion model suggests that the current strong rally is over-extended and a pull back is likely. Volume completed - 13,103,951.

2.44pm: Rally continuing, approaching 273p. Volume completed - 13,103,951.

3.05pm: Reversal underway and market approaching dark/hidden algorithm limit at 267p ; inline algorithm already active again. Volume completed - 13,103,951.

3.22pm: Dark/hidden algorithm highly active during past fifteen minutes. Volume completed - 14,112,956.

3.27pm: Continued activity by the dark/hidden strategy sees order completed at an average price of 265.82p. 38% of the volume was traded by the dark/hidden algorithm with the total proportion of dark liquidity sourced reaching 68% of the total order. Figure 5 illustrates the trade's progress throughout the session, while Figure 6 provides an overall perspective of trading activity in Barclays on January 22nd.

Figure 5

Figure 5

Figure 6

Figure 6