The Scenario: It is after market hours on 3rd October 2011 and a Hong Kong based manager is concluding a review of the portfolio of a small/mid cap fund for which he is responsible. The fund acquired a stake of 400,000 shares in China Resources Gas Group in early March. While the stock has been resilient since then in comparison with the market as a whole, the manager is inclined to offload the position. The company's city gas distribution business, which includes piped natural or petroleum gas distribution as well as operating compressed natural gas filling stations in mainland China, makes it a relatively defensive stock. Nevertheless, the manager is taking the view that a cyclical turning point is approaching (despite the recent market turmoil) and that the time is right to switch to more growth oriented stocks.
Two stocks that the manager feels fall into this category are China Zhongwang Holdings and Country Garden Holdings. China Zhongwang manufactures aluminium extrusion products for the construction and industrial sectors. Country Garden and its subsidiaries are involved in various businesses associated with property development, including construction, fitting and decoration, property management and hotel operation. The manager's view is that both stocks have also become oversold over the past three months (especially Country Garden) and that the current febrile market conditions could be a good buying opportunity with a six to twelve month outlook.
The manager therefore instructs his execution desk to complete the necessary orders by the close of the Hong Kong trading session of the following day, 4th October 2011.
The Assets: China Zhongwang Holdings, China Resources Gas Group and Country Garden Holdings.
The Challenge: To sell 400,000 shares in China Resources Gas Group at an average price of 10.3. To buy 1,800,000 shares in China Zhongwang Holdings at an average price of 2.55 and to buy 4,000,000 shares in Country Garden Holdings at an average price of 2.
The Trader: The trader who has been assigned the three orders specialises in small/mid cap names and works closely with the execution desk of the broker providing the execution algorithms the trader uses.
The Algos: The Trader will be using two algorithms to execute the three trades. Algorithm A is an opportunistic strategy designed to aggressively capture liquidity while using stealth to minimise market impact and signalling risk. It takes advantage of dark, non-displayed liquidity by intelligently routing to multiple alternative trading systems. The strategy also looks for opportunities where aggressive trading without unnecessarily pushing the price is possible. As overall liquidity improves, the strategy will aggressively cross the spread to maximise fill rates.
Users of Algorithm A can specify their execution style as one of three modes - patient, normal and aggressive:
• Patient will send the order to dark venues only. Traders can choose minimum block size (to limit others' ability to pick them in the dark), block price (to set a limit), as well as maximum block size (for the overall block quantity). Example: Post 500,000 shares in the dark as a block (to be replenished if fully executed) and do not execute against orders that are less than 10,000 shares - use the parent order as the limit for the dark block.
• Normal will also send IOC orders to lit venues to improve fill rates. If a stock lacks liquidity, the algorithm will wait for a volume shock (see "Volume shock" below) before triggering in lit books to minimise the possibility of detection while participating in dark books. As such, the strategy will be switching its behaviour between passive and aggressive styles by continuously re-evaluating volume patterns.
• Aggressive will relax the conditions required to trigger trading aggressively at the far touch, and will therefore be more active than Normal mode. The algorithm aggressively jumps the spread within its fair price limitations in the visible venues, while always participating in the broker's alternative trading system.
Algorithm A depends heavily upon a fair value model to adjust its participation rates and to minimise the risk of executing at a price which would seem unreasonable in the context of prevailing market conditions. The model calculates a fair value in real time for each security and places upper and lower bands a certain percentage above and below this. If price is unattractive to a certain degree (for example, for a buy order if the offer exceeds the upper band), the algorithm completely stops executing for a short, random period of time. On the other hand, if the price is extremely attractive (possibly due to a fat finger error or other positive event), it aggressively takes liquidity and may for a limited period walk the book in order to do so.
Algorithm B is a variant of a popular tactic used to target an implementation shortfall benchmark - instead of aiming to beat the arrival price, the primary goal is to execute the trade at or better than the volume weighted average price (VWAP) since arrival time. Users can also target their desired participation weighted price benchmark as an alternative to VWAP - this behaviour can be changed by the trader on the fly.
The algorithm can work an order to be a dynamic percentage of the traded volume, based upon the attractiveness of the current stock price versus the VWAP since arrival time (VSAT). It is designed to trade faster when the price is better and trade slower when the price is worse than the VSAT - this behaviour can be switched by the Trader on the fly.
The algorithm's baseline participation rates will be determined by the aggressiveness of the order, the price volatility of the stock and the attractiveness of the price. Users of the algo can input their own participation minimum and maximum to moderate the order's participation and execution performance.
Algorithm B also uses the same fair value model as Algorithm A, but only to decide if it should disengage temporarily during adverse price dislocations. Additionally, similar to Algorithm A, one can float a dark block while trading using the same parameters (minimum block size, block price, maximum block size).
Finally, both algorithms allow a trader to enter an "I Would" price at which they are happy to complete the order by overwriting all volume restrictions. If the price moves back away from the I Would price prior to the completion of the order, the algorithm will abandon the completion mode and trade the order normally.
For illiquid names, such as those being traded here, both algorithms also use a concept called volume shock that affects their behaviour. If the stock has been more active recently than it was earlier in the trading session, the algorithm is triggered to trade aggressively as it can hide its trail. Otherwise it will assume that the liquidity is thin and will elect to trade passively.
8:30 (all times are HKT) - The trader is expecting very substantial volatility both at the open and throughout the trading session in view of a 200+ point sell off in the Dow the previous day on the back of Eurozone concerns.
9:00 - The trader sets up the three trades, initially electing to use Algorithm A for all of them. As he is keen to get in and out of these names, he specifies no limit price. The trader also sets minimum block sizes at 25,000 shares for all three stocks, while using the parent order price and size for block price and maximum block price. He sets an I Would price of 2.52 for China Zhongwang at which he is happy to complete the order, as he is particularly worried about completing this order given its size relative to Average Daily Volume (ADV).
Based on a market microstructure study provided by the quantitative strategy team at the broker, the trader believes that on average Asian markets tend to overreact to US market returns at the open but revert/correct later in the session. Given that the US market was down considerably, the trader chooses to join the opening auction for buy orders to take advantage of potentially low prices while he chooses to forgo the auction volume for the sell order.
Figure 1: Buy 1333 HK - China Zhongwang Holdings ltd
9:30:00 - China Zhongwang opens at 2.56, the algorithm gets 16,000 shares done at the opening print, while Country Garden opens at 1.92 and the algorithm captures 28,000 shares. The trader was correct in his anticipation: China Zhongwang and Country Garden are rallying rapidly after a low opening print.
9:34:00 - Continued participation. Volumes traded: China Zhongwang 30,400, China Resources 6,000, Country Garden 140,000.
9:35:00 - Country Garden has broken above the upper band of the algorithm's fair value model (see Figure 3 ). The trader is automatically alerted via electronic message and also by a call from the broker's execution desk. The algorithm responds by going into sleep mode on this order while constantly watching for prices and volume shocks.
9:35:55 - A major fat finger error by another market participant also pushes China Zhongwang's offer well beyond the upper band of the algorithm's fair value model (see Figure 1 ). The algorithm goes into sleep mode on this order - again, an automated electronic message alert is generated to inform the trader of an adverse fair value breach - the algorithm stops executing, while constantly observing price and volume information as well as order book imbalances. The price move narrowly misses the broker-defined advanced order protection circuit breaker at the HKD2.7 price level (if breached the order would be automatically paused and would require manual intervention by the broker's execution desk to restart it).
9:36:23 - China Zhongwang's offer price falls quickly back below the fair value model's upper band. Algorithm A responds by starting to participate at the low rate. As the decline continues, the offer falls below the central fair value and the algorithm increases its participation drastically.
9:36:27 - With the low volumes in the stock going through the market, the volume shock trigger for China Zhongwang is turned off. While constantly watching order book and trading activity, the algorithm is able execute a block in the broker's dark pool for 35,200 shares. Volumes traded: China Zhongwang 87,600, China Resources 10,000, Country Garden 146,000.
9:36:54 - The continued market sell-off in the aftermath of the fat finger error sees China Zhongwang's offer drop for about a second below the fair value model's lower band. The trader is alerted electronically and by phone by the broker. Meanwhile, the algorithm responds by capturing the entire 20,400 shares currently sitting on the offer (see Figure 1 ). Volumes traded: China Zhongwang 108,000, China Resources 18,000, Country Garden 147,000.
9:40:00 - China Resources' offer has been fluctuating to and fro across the central fair value level since the auction closed and Algorithm A has been gradually unloading stock. Volumes traded: China Zhongwang 114,000, China Resources 24,000, Country Garden 160,000.
9:42:00 - Country Garden hits what proves to be its high of the day 2.07. This not only takes it above the upper fair value band, but also triggers the broker's advanced order protection functionality, which automatically suspends the algo completely (unless manually restarted by the broker's sales desk), sends an automated alert to the trader and results in a call from the broker's sales desk asking if the suspension should be made permanent. The risk management engine that utilises the circuit breaker for all algorithms traded sets the new circuit breaker level for the order (see Figure 3 ).
In view of overall market conditions and after a discussion with the fund manager, the trader takes the view that the rally is an anomaly that will not persist and expects the stock to fall back and (at best) trade sideways for the rest of the session. He therefore notifies the sales desk that the order will be reactivated once the Country Garden offer falls back below the fair value upper band, but elects to switch execution algorithm to Algorithm B with an I Would price of 1.93 (at which he is happy to complete the order).
9:51:00 - Country Garden now back below the upper fair value band and restarts trading with Algorithm B.
10:00:00 - All three stocks have been trading to and fro within the fair value bands. Country Garden has recently dipped below VSAT so Algorithm B increased participation, while China Resources has been trading favourably compared to fair value, so the algorithm has started to unload stock rapidly. Volumes traded: China Zhongwang 237,600, China Resources 100,000, Country Garden 267,000.
10:05:00 - After breaking the fair value upper limit and causing the algo to temporarily disengage, Country Garden has rallied above VSAT so Algorithm B has decreased its participation (see Figure 3 ).
10:35:11 - About a minute after a massive offer shows up and is lifted, a major price dislocation in favour of the trade in China Resources (which has been rallying strongly) occurs as the bidder follows and moves its price rapidly above the upper fair value band (see Figure 2 ). Algorithm A immediately moves to 100% participation and over the next 25 seconds manages to unload 64,000 shares before the bid drops back below the upper band.
Figure 2: SELL 1193 HK - China Resources
Figure 3: Buy 2007 HK - Country
11:00:00 - All three orders have been running normally for the past 25 minutes, apart from one brief flurry below the lower fair value band by China Resources - Algorithm A easily evades these temporary movements (see figure 2 ). China Zhongwang and China Resources have been trading worse than fair value; as such, the algorithm has limited its participation. Volumes traded: China Zhongwang 470,800, China Resources 332,000, Country Garden 862,000.
11:51:25 - A block of 27,000 China Zhongwang shares executes in the broker's dark pool. Volumes traded: China Zhongwang 556,000, China Resources 362,000, Country Garden 1,571,000.
12:00:00 - Market closes for lunch after an hour of gradual progress for all three orders. Volumes traded: China Zhongwang 559,200, China Resources 364,000, Country Garden 1,664,000.
13:30:00 - Trading resumes.
13:30:53 - Country Garden trades outside the fair value upper limit - Algorithm B temporarily stops trading (see Figure 3 ).
14:00:00 - No major developments, apart from a rally close to the upper value band by Country Garden. Volumes traded: China Zhongwang 753,200, China Resources 376,000, Country Garden 2,058,000.
14:42:00 - Country Garden falls below VSAT and algorithm increases its participation rate (see Figure 3 ).
15:00:00 - For the past hour, the ask price for China Zhongwang has remained above fair value, resulting in lower participation. Volumes traded: China Zhongwang 915,200, China Resources 386,000, Country Garden 3,072,000.
15:06:00 - China Resources order completes. As the algorithm was able to opportunistically pick its pricing points, it beat VWAP by 92 bps.
15:33:46 - Country Garden's offer price hits the trader's I Would limit (see Figure 3 ), the algorithm abandons volume restrictions and pounces to complete the order at or better than the I Would price - order completes within half a minute. By carefully avoiding adverse price discrepancies with the help of fair value limits and advanced order protection, and by utilising the algorithm's I Would functionality, the execution price beats VWAP for the trading period by 74 bps.
15:40:00 - The China Zhongwang order is giving the trader serious cause for concern. With around 850,000 shares still to acquire, order book volatility starts to increase considerably, with sizable bids and offers showing up. After a brief discussion with the broker's sales desk, the trader decides to change the trading style of the algorithm to Aggressive.
15:41:32 - The China Zhongwang offer goes outside the upper fair value band and the algorithm briefly suspends trading. Volumes traded: China Zhongwang 1,215,200, China Resources 400,000, Country Garden 4,000,000.
15: 46:50 - A massive offer in China Zhongwang of over 1,000,000 shares appears. The algorithm limits its participation based on the historic book size and only executes 60,000 shares, as it expects downward pressure on price while another market player lifts the entire remainder. Volumes traded: China Zhongwang 1,525,200, China Resources 400,000, Country Garden 4,000,000.
15:47: 26 - After a series of large offers shown in the market, price falls below fair value and the algorithm considerably increases its participation rate and in due course hits the trader's I Would Trigger. The algorithm starts to rapidly accumulate shares at or better than this level. Volumes traded: China Zhongwang 1,633,200, China Resources 400,000, Country Garden 4,000,000.
15:52:47- China Zhongwang continues to dip seriously with continued pressure, and I Would mode completes the order at 2.53011 or 5.2 bps worse than VWAP. Still good timing, as barely 8 minutes later the stock jumps by more than 10%.
Average execution prices achieved:
China Zhongwang: 2.544
China Resources: 10.48
Country Garden: 1.9877
Trading Characteristics for the Names on the Day
Price Range (High-Low)
Average Spread (bps)
Widest Spread (bps)
Figure 4: 1333 HK - China Zhongwang Completion
Figure 5: 1193 HK - China Resources Completion