Results of TSE Market Impact View!Trading Simulation

Size: px

Start display at page:

Download "Results of TSE Market Impact View!Trading Simulation"

Phillip Gordon
6 years ago
Views:

1 HOW DO ORDER BOOK CHANGES AFFECT INTRA-DAY PRICE MOVEMENTS? Results of TSE Market Impact View!Trading Simulation CMD Laboratory Inc.

CHAPTER 1 Trading Simulation Using TMIV Data Contents 1. The Impact of Order Book Changes on Price Movements Trading'Simula8on' Using'Changes'in'Order' Book'Balance% 2. Are Profits Stable? 3.

2 CHAPTER 1 Trading Simulation Using TMIV Data Contents 1. The Impact of Order Book Changes on Price Movements Trading'Simula8on' Using'Changes'in'Order' Book'Balance% 2. Are Profits Stable? 3. Relationship Between Thresholds and Profitability 10:30:00 10:30:05 10:30:10 10:30:15 10:30:20 10:30:25 10:30:30 10:30:35 10:30:40 10:30:45 OB)Balance )3.2 )0.1 Over%2.5 )2.1% )1.8% 1.2% Over% % No'trade'made'since' sign'is'different'from' last'ob'change% Over%2.5 Time% )3.7% )5.5% Over%2.5 )6.8% Fig. 1.1 Trading Simulation Using Changes in OB Balance as a Signal 1.1.!The Impact of Order Book Changes on Price Movements In the previously published report entitled Market Analysis Using TSE Market Impact View, we showed that changes in order book depth (i.e. changes in Order Book Balance [OB Balance]) had a greater impact on subsequent price fluctuations than book depth itself. Furthermore, results from a simulation that used OB Balance changes as a trading signal showed that a trader could earn positive returns using this method for a majority of stocks (80 stocks out of the 101 for which TMIV information is available). This report builds on that study by expanding 2

the data timeframe used in the simulation to the period from January to August, 2012. The specific process is outlined below: Defining the Trading Signal!

3 the data timeframe used in the simulation to the period from January to August, The specific process is outlined below: Defining the Trading Signal! In principle, a trade is made whenever a change in OB Balance exceeds a given threshold (e.g. 2.5 points) in a positive direction (buy) or negative direction (sell). However, trades will not be made if the sign of the change in OB Balance directly before the threshold-breaking change is different than the sign of the threshold-breaking change, in order to eliminate order book noise created by things like algorithmic trading Are Profits Stable? Figs. 1.2 to 1.5 show the results of a series of trading simulations using TMIV data from January to August of The averages given here refer to the daily average per stock. Trade Timing! After confirming the presence of a trading signal as defined above, the stock price 5 seconds after the OB balance change is used as the base price for the trade in order to account for the time required for order and execution. Calculating Returns Fig. 1.2 Number of Winning/Losing Stocks and Return Rates Per Month for 30 Sec. Threshold! The rate of change in price from the base price determined in above to a given point sometime in the future (30 seconds or 1, 2, or 3 minutes) is recorded as a profit or loss. Fig. 1.1 provides an illustration of the above process. Note: The focus of this simulation is to illustrate the relationship between changes in OB Balance and stock price fluctuations, and thus closing trades are assumed to have been made at current (i.e. midpoint) prices after the given time interval has elapsed. Fig. 1.3 Number of Winning/Losing Stocks and Return Rates Per Month for 1 Min. Threshold 3

Threshold In order to explore the relationship with actual market conditions, we look at daily movements in the TOPIX index in Fig. 1.6.

4 the number of profit-making stocks exceeded the number of lossmaking stocks for the entire period (per month). In addition, average returns were positive over the entire period as well, regardless of the time interval. Fig. 1.4 Number of Winning/Losing Stocks and Return Rates Per Month for 2 Min. Threshold In order to explore the relationship with actual market conditions, we look at daily movements in the TOPIX index in Fig Since this simulation investigates the relationship between fluctuations in OB Balance and intraday stock price movements, we did not anticipate any direct connection with market movements on a day-today basis. Interestingly enough, however, a comparison of Figs. 1.2 to 1.5 with Fig. 1.6 indicates that the rate of return tends to increase during uptrends in the overall market. Having said this, eight months of data alone is inadequate to draw any definite conclusions, and further analysis is required in the future. Fig. 1.5 Number of Winning/Losing Stocks and Return Rates Per Month for 3 Min. Threshold Figs. 1.2 to 1.5 show the total number of profit-making stocks (blue bars), loss-making stocks (red bars), and the average return rate per stock for the entire group (green line) for each month. These results indicate that regardless of the time interval between trades, Fig. 1.6 Daily Fluctuations in TOPIX Index (Jan. 4, 2012 to Aug. 31, 2012) 4

1.3. Relationship Between Thresholds and Profitability The differing threshold levels in Fig. 1.7 result in widely diverging trading frequencies, as shown in Fig. 1.8.

Although a high threshold implies that trades are only made when there is a large change in OB Balance, Fig. 1.7 shows that trades made at a threshold of 2.

5 1.3. Relationship Between Thresholds and Profitability The differing threshold levels in Fig. 1.7 result in widely diverging trading frequencies, as shown in Fig How often a given level of change in OB Balance results in positive returns depends on the state of the market. Fig. 1.7 displays the average return per transaction by threshold per month. Although a high threshold implies that trades are only made when there is a large change in OB Balance, Fig. 1.7 shows that trades made at a threshold of 2.5 points result in the highest returns throughout the entire period. Throughout the entire simulation period, the only months with a negative average rate of return was May, 2012 for the 5 point threshold and June, 2012 for the 10 point threshold. Except for these two months, returns (i.e. profitability) increase as threshold values decrease. Fig. 1.8 Number of Transactions Per Month by Threshold Level Fig. 1.8 shows that the difference in transaction numbers for each threshold is roughly proportional to the size of the difference in threshold for each month (for example, the number of transactions for the 5 point threshold is roughly twice the number of transactions for the 2.5 point threshold). Furthermore, the fact that the number of transactions in May, 2012 is the highest across the entire period (i.e. the change in OB Balance was most volatile) illustrates that while normal volume was decreasing (see Fig. 1.6), changes in the order book were occurring with increasing intensity. Fig. 1.7 Average Returns Per Month by Threshold Level 5

6 CHAPTER 2 Analysis of Trading Simulation Data on a Daily Basis Contents 1. Trading Simulation on a Daily Basis 2. Simulation Results by Threshold Value Topix 780" 770" 760" 750" 740" 730" 720" 710" 700" 690" No.%of%Winning/Losing%Stocks%by%Day% Compared%With%TOPIX% No."of"Winning"Stocks"(Le<"Axis)" No."of"Losing"Stocks"(Le<"Axis)" Topix" 801" 802" 803" 806" 807" 808" 809" 810" 813" 814" 815" 816" 817" 820" 821" 822" 823" 824" 827" 828" 829" 830" 831" 200" 180" 160" 140" 120" 100" 80" 60" 40" 20" 0" No.of%Stocks Fig. 2.1 Trading Simulation Results on a Daily Basis (August, 2012: 2.5 Point Threshold) 2.1. Trading Simulation on Daily Basis Fig. 2.1 breaks down the results of the trading simulation outlined in the previous section (given on a monthly basis) into individual days for the month of August, The blue bars at the bottom of the chart indicate stocks that recorded a profit for the given day while the red bars indicate stocks that recorded a loss thus when the blue bars are higher than the red bars, there was an opportunity to make a profit on that specific day. Stocks for which no trades were made, however, are not included, and therefore days that saw volatile changes in OB balance (i.e. changes in the order book) are denoted by a large combined number of stocks in the red and blue bars. 6

While the level of profit opportunity for each day is illustrated by the balance between the red bars and the blue bars in Fig. 2.1, Fig. 2.2 contrasts this with each day s average rate of return.

Furthermore, when we juxtapose this with Fig. 2.1, we can observe that the tendency for profit opportunities to increase during an upward trend in the market (e.g. 8/1 to 8/8 and 8/16 to 8/20) as described in the previous section apply to the daily trading simulation as well.

4 as the threshold increases, the number of transactions decreases. Although this simulation does not take trading costs into account, these are affected by the number of transactions made. Fig. 2.

7 While the level of profit opportunity for each day is illustrated by the balance between the red bars and the blue bars in Fig. 2.1, Fig. 2.2 contrasts this with each day s average rate of return. A quick look at Fig. 2.2 confirms that the more profit making opportunities there are for any given day (i.e. days for which the blue bars are high), the higher the rate of return. Furthermore, when we juxtapose this with Fig. 2.1, we can observe that the tendency for profit opportunities to increase during an upward trend in the market (e.g. 8/1 to 8/8 and 8/16 to 8/20) as described in the previous section apply to the daily trading simulation as well. threshold size does not necessarily always determine the rate of return. The average number of transactions made each day by threshold is given in Fig. 2.4 as the threshold increases, the number of transactions decreases. Although this simulation does not take trading costs into account, these are affected by the number of transactions made. Fig. 2.3 Average Rate of Return by Threshold (August, 2012) Fig. 2.2 Profit Opportunities and Average Rate of Return (August, 2012, 2.5 Point Threshold) 2.2. Simulation Results by Threshold While Fig. 2.2 shows the results of the trading simulation with a trading threshold of 2.5 points (i.e. trading based on small changes in the order book), Fig. 2.3 shows the rate of return when the same simulation is performed with a threshold of 5 and 10 points. While returns tend to be larger for smaller thresholds, Fig. 2.4 Number of Transactions by Threshold (August, 2012) 7

8 DISCLAIMER The information provided in this report is for informational purposes only and is not intended as investment solicitation or recommendation of any particular securities company. TSE Market Impact View is an information product developed by Tokyo Stock Exchange, Inc. in collaboration with CMD Lab, Inc. Tokyo Stock Exchange Inc. and CMD Lab, Inc. shall bear no responsibility whatsoever in the unlikely event that some loss is incurred based upon the content of this report. While the utmost effort is exerted in ensuring the completeness of this report, it makes no guarantee with regards to such. While this report contains information/ commentary on TSE Market Impact View, it does not encompass the entire contents of the product. The information contained in this report may be altered without prior notification. The information contained in this report indicates past results and is no indication of future performance. CMD Lab, Inc. retains all rights with respect to all parts of this report and prohibits duplication, transmission, etc. for any purpose and through any method (electronic or mechanical). i

CHARTS. Bar Line Candlestick Charts are the basis of technical analysis They are a graphic display of price action. Notes:

CHARTS. Bar Line Candlestick Charts are the basis of technical analysis They are a graphic display of price action. Notes: TECHNICAL ANALYSIS CHARTS Bar Line Candlestick Charts are the basis of technical analysis They are a graphic display of price action Candlestick charts have become the industry standard for technical traders.