Trading Tasks Vwap

The trading tasks on August 8, 2011 consisted of two sessions. The first session involved selling 100,000 shares of BTA on behave of the client, while the second session was to purchase 50,000 shares of LYC within a trading day.

In session one, the order was partitioned into six parcels. The volume of shares allocated to each parcel was based on the recent 5-days of BTA trading data. The objective of the trading session was to execute a sell order of 100,000 shares of BTA by the end of the day, while at the same time achieving WVAP result close to or better than the market WVAP. Executing parsels of orders was very straight-forwarded. We made sure that at each hour the allocated parsel of shares was executed initially at limit price, slightly higher than the market ask price. Near the end of the hour the limit prices were readjusted to lower values than that of market price to guarante execution. This procedure was repeated for every hour throughout the whole trading session to ensure the whole package was executed before the trading day expires. At the end of the day, the calculated WVAP was as follows:

Market VWAP (VWAPM) = $2.2344

Trader VWAP(VWAPResult) = $2.2358

Execution cost(EC) of a sell order

= (VWAPM) - (VWAPResult) = 2.2344 - 2.2358 = -0.0041

By the end of the trading session, we were able to fully execute an sell order of 100,000 shares of BTA per client's request; at the same time, our result showed that we beat the market VWAP benchmark with profits from selling BTA shares at $0.0041 more per share.

Chart 1: Traded volume by Hour

Chart 2: Historical Volume by Parsels

Short comings of the strategy in session One: Although we were able to achieve the goal of executing the order completely by the end of the trading day while achieving VWAP result better than the market VWAP, it was not easy to execute the strategy completely according to the plan. From Chart1, the actual volume executed for parsel 5 was much higher than the pre-determined volume, by a difference of 20,000 shares. The reason was that since it was close to the trading day, more traders were placing orders near the bid-ask market price, resulting in stagnant market price. Thus much of our limit orders failed to execute timely as the market price failed to rise to the target prices. Manual intervention was required to adjust limit prices to one or two ticks above market price before the end of the trading day. On the other hand, while adjusting prices there were cases that some of the limit order were placed too close to the bid-ask spread, resulting in swift execution as market price increased unexpectly, and orders were executed ahead of schedule. Figure 2 demonstrated the weakness of using historical volume statistics to plan strategy execution. The VWAP strategy for this session was based on 5-days of historical volume statistics, however, the parsels for the 5-days period varied significantly. For example, the same parsel 2 for day 2 and day 5 differed by as much as 25%! Although we used the average value across 5 days to determine the proportion of shares to place in each hour, the actual volume of shares that was traded across the market for each hour differed significantly from the historical volume, which eand we had to make adjustment to the order placed inorder to keep up with the schedule.

Possible improvements from Session One: We can indeed make improvements to the VWAP strategy utilized in executing this order. Konishi (2002, p205) suggests using small time increments and smaller parsels to avoid execution cost, especially the component of the cost related to market impact. The idea is to excute the orders with as little effect to the changes to market price as possible. Instead of focusing on executing one parsel per trading hour, we can carry out the order in 30-minute period, with half as much volume relative to the origin volume traded in an hour-span period, or at 15-minute period, with one-fourth the amount of the original one-hour volume, etc. This should improve consistency of manual order execution at market price while still achieving the goal of obtaining the market VWAP at the end of the trading session.

Similar to previous session, session 2 was to execute an buy order of 50,000 LYC shares by the end of the day. A combination of volume statistic strategy and limit order strategy was also applied in session 2. In the first half of the trading day, volume statistic strategy was carried out similar to session 1. However, we soon realised that we were incurring a high execution cost when trade according to the percentages of volume scheduled for each hour. The reason was the high liquidity of LYC stock with relative low price volatility, large volume of orders queued up near market price around the small bid-ask spread and our limit orders stagnated even in the beginning of the day when the stock was supposed to be most active. As a result, in order to execute predetermined parcels at each hour, we had to bid for the amount of shares we need at the best market ask price, resulting in slippages that deviated our VWAP from market VWAP.

In the second half of the trading session, we switched to a passive trading strategy by locking in limit orders that were one and two ticks lower than the best bid price trading at that time ($0.550) for the remaining parcels and wait for them to be executed. The main argument behind this approach was suggested through a study by Abhyankar, Ghosh, Levin and Limmack (1997, p344), whose research result showed that average bid-ask spread exhibits an U-shaped pattern, with highest spread in the beginning and closing of a trading session, while falling relatively flat during mid-trading session. Thus holding onto limit orders and not executing them during the middle of the day when the stock was most inactive should prevents further slippage. Even if not all orders are filled toward the end of the trading day, executing the remaining unfilled volumes with market orders in the closing hour should have higher chance of achieving or beating market VWAP since there would be enough volatility in price. Of course, this passive strategy is not risk free. By holding onto limit orders, we exposed ourselves to the risk of execution uncertainty (Larry, H 2003), where the price may move away from our orders, but through observating LYC's volatility in the first half of the trading session, this is unlike to happen.

By the end of trading session, although not