This article is by Gabriel Kan & Hideaki Sano.
- There are long queue in Asia: It takes 33 trades to sweep one queue in HK, but only 7 trades in U.S. And 4.7 trades in U.K.
- Price reversion is strongly correlated to queue size: Trade-to-queue-size ratio decreases by 1%, probability of price reversion increases by 4.25%.
- Anticipate price reversion: We should not only get to the front of the queue to get filled, but also stay in queue longer to anticipate price reversion
Mailing a letter, waiting for a bus, buying a burger… queuing is an essential part of everyday life to get what you want, and trade execution is no different. By posting on the same side of the order book, passive orders can improve execution by capturing the spread and minimizing market impact. In order to increase the chance of getting filled under the first-in-first-out rule, orders need to queue early and obtain front positions in the order book. It is particularly challenging when the market has a long queue as is the case in Asia. Considering the average trade size as a percentage of the average best bid-offer (BBO) quote size, Hong Kong’s is only 3% compared to 14% in the US and 21% in the UK (figure 1). This means that it takes 33 trades to sweep one queue in HK, while it only takes 7 trades in the US and 4.7 trades in the UK.
Being in the front of the queue is always preferred to increase fills, however controlling market participation is more challenging. When the front of the queue is executed, orders at the back may withdraw to anticipate the price pressure. Therefore, queue size is not always a good proxy for the expected market volume. Furthermore, the queue size is dynamically changing which makes it more difficult to anticipate. In this article, we analyze the queue dynamics in different scenarios so that traders can fully utilize the advantages of passive orders to improve their execution.
Figure 1: Average trade sizes as percentage of average quote by market
To focus our study in Asia, we look at the constituents of Hong Kong Hang Seng Index (HSI Index) in June 2016. We consider a queue as the list of orders lining up at the best bid or offer price. A queue can start in two different ways:
- Start at New Level ↗: The best bid (resp. offer) price moves up (resp. down) to start the queue, i.e. there were no buy orders (resp. sell orders) queued up at the current best bid price (resp. offer price).
- Start at Old Level ↘: The best bid (resp. offer) price moves down (resp. up) to start the queue, i.e. there were existing buy orders (resp. sell orders) already queued up at the current best bid price (resp. offer price).
For each starting type, a queue can end in two different ways:
- End with Next Queue at New Level ↗: The next queue starts at a new best bid (resp. offer) price level, i.e. there still are residual orders at the queue, but the best bid (resp. offer) price becomes higher (resp. lower).
- End with Next Queue at Old Level ↘: The next queue will start at a lower bid price (resp. higher offer price) where there are existing orders at that new price level, i.e. all originally queued orders are either executed or withdrawn.
Figure 2: Queue Categorization
Therefore, a queue can be classified in one of the four cases, depending on how it starts and ends. In our dataset, we found that the most common cases are 28% of New-Level-to-Old-Level (↗↘) and 36% of Old-level-to-New-level (↘↗). These two cases, 64% in total, represent price reversion: E.g. A queue starts at the best bid that is higher than the previous, and then the best bid goes down to the next queue.
Figure 3: queue formation
The probability of having a price reversion is strongly correlated to the queue size. Our data suggests that if the trade-to-queue-size ratio decreases by 1% (longer queue), the probability of having a price reversion increases by 4.25%. This implies that if a stock exhibits longer queues, we should not only get to the front of the queue to get filled, but also stay in the queue longer even if the queue is no longer at the best bid (resp. best offer), i.e. we should keep the order at the second price level. It is because the chance of price reversion is large.
Figure 4: probability of price reversion is correlated to the queue size