The Matching Mechanisms in the U-Mart Experiment
4.3.1 The Shapes and Performances of the Market Mechanism
When referring to the Tokyo Stock Exchange, we use the terms “Itayose” or “Zaraba”, rather than “ask/bid auction” or “double auction”, because the latter
| Ask | Price | Bid |
| 82 | 1,798 | |
| 73 | 1,793 | |
| 38 | 1,779 | |
| 44 | 1,778 | |
| 91 | 1,776 | |
| 1,770 | 67 |
Note: The restricted range given by the exchange
Table 4.6 Order book sample
terms have slightly different meanings, depending on context.
The double auction mechanism is designed to achieve the following priorities for trading:1. Market order priority
2. Higher price priority (Buy); Lower price priority (Sell)
3. Preceding offer first priority (time preference)
Zaraba is an example of an institutional linked market mechanism. The trading priorities must be associated with the following institutional devices:
1. Price movement limit
2. Special bid and ask price
3. Updating the price movement limit
4. Stop high/stop low
These institutional devices can guarantee smooth price formation in the stock exchange system. The performance of the trading mechanism depends on its own institutional settings.
The limits exhibited are not always successful in accepting bids and asks within the desired range. Either special bids or asking prices are necessary if the prices exceed the limit.
The exchange system must simultaneously update the price movement limit. In Table 4.6, there is no bid within the given price movement limit. The stock exchange announces a special bid and updates the limit to accept it. Without this kind of iteration based on an institutional arrangement, quotations would not be feasible. The specification of the limit may be multiple, if there is a long tail to the bid/ask distribution. The result of the contract may depend on a particular institution and rule, so the exchange system needs a refined skill to set out and manage the price movement limit.Even contract guidance could not necessarily guarantee a normal trade if there is a rapid fluctuation of ask/bid. In this situation, the exchange system must employ the rule of Stop High/Stop Low to interrupt current trades. This breaks the circuit of the exchange.
| Ask | Price | Bid |
| 82 | 1,798 | |
| 73 | 1,793 | |
| 38 | 1,779 | |
| 44 | 1,778 | |
| 91 | 1,776 | |
| 1,770 | 67 |
Note: An updating session by the exchange
| Ask | Price | Bid |
| 246 | 1,770 | special |
Note: Special bidding in an updated session
Table 4.7 Updating the limit
Table 4.8 A special bidding
4.3.1.1 Institutional Customs to Set the Stepwise Ranges
for Stop High/Stop Low
Price movement limits on the bid and asked prices, i.e., a range in prices, is set out by custom on the Tokyo Stock Exchange.
In Table 4.7, the prices fluctuate around 1,500-2,000yen, so the price movement limit is taken within 20 yen, from 1,773 to 1,793. This is used to set the price contracted at the end of the previous day as the marker price for each new day. Based on the marker price, the rule of stop high and low regulates any violent price variation by stopping the trade (Table 4.8).In 2012, TSE revised the stepwise ranges as shown in Table 4.9.
4.3.1.2 Visualizing the Differences in Matching Mechanisms
See Fig. 4.8; Tables 4.10 and 4.11.
4.3.2 Zero-Intelligence Tests in the U-Mart System
The U-Mart system is mainly designed for futures market trading. In this trading, either machine or human agents join to create a sequence of successive futures prices, with reference to given real spot market prices arbitrarily chosen. The settlement must be done at the final delivery date by employing the real spot price. This provides some entertaining gambles. By removing regulation from futures market trading, we can imitate a trade for the spot market. In other words, the spot market can be derived as a special case of the U-Mart system.
Table 4.9 Partial revision of daily price limits and special renewal price intervals
| Price (JPY) | Daily price limits | Renewal price intervals | ||
| Current | Revised | Current | Revised | |
| ~100 | 30 | 30 | 5 | 5 |
| 100-200 | 50 | 50 | 5 | 5 |
| 200-500 | 80 | 80 | 5 | 5 |
| 500-700 | 100 | 100 | 10 | 10 |
| 700-1,000 | 100 | 150 | 10 | 15 |
| 1,000-1,500 | 200 | 300 | 20 | 30 |
| 1,500-2,000 | 300 | 400 | 30 | 40 |
| 2,000-3,000 | 400 | 500 | 40 | 50 |
| 3,000-5,000 | 500 | 700 | 50 | 70 |
| 5,000-7,000 | 1,000 | 1,000 | 100 | 100 |
| 7,000-10,000 | 1,000 | 1,500 | 100 | |
| 10,000-15,000 | 2,000 | 3,000 | 200 | 300 |
| 15,000-20,000 | 2,000 | 4,000 | 200 | 400 |
| 20,000-30,000 | 3,000 | 5,000 | 300 | 500 |
| 30,000-50,000 | 4,000 | 7, 000 | 400 | |
| 50,000-70,000 | 5,000 | 10,000 | 500 | 1,000 |
| 70,000-100,000 | 10,000 | 15, 000 | 1,000 | |
| 100,000-150,000 | 20,000 | 30, 000 | 2,000 | 3,000 |
| 150,000-200,000 | 30,000 | 40, 000 | 3,000 | 4,000 |
| 300,000-500,000 | 50,000 | 70, 000 | 5,000 | 7,000 |
| 500,000-700,000 | 100,000 | 100,000 | 10,000 | 10,000 |
| 700,000-1,000,000 | 100,000 | 150,000 | 10,000 | 15,000 |
Note: The highlighted cells indicate newly revised prices
| Ask | Price | Bid |
| MO | 1,000 | |
| 5,000 | 1,702 | |
| 8,000 | ] 1,701 | |
| 1,700 | 4,000 | |
| 1,699 | 7,000 | |
| 1,698 | 20,000 |
| Ask | Price | Bid |
| MO | 1,000 | |
| 5, 000 | 1,702 | |
| 7, 000 | ] 1,701 | |
| 1,700 | 4,000 | |
| 1,699 | 7,000 | |
| 6,000 | 1,698 | 20,000 |
Table 4.10 The first scene
Table 4.11 The second
scene
In the U-Mart system, default loaded agents with technical analyses are[53]:
TrendStrategy: Set price1=last futures price, and price2= second last futures price.
If price1 is higher than price2 then the agent orders buying.
If price1 is lower than price2, then the agent orders sell. The amount of order is randomly decided. AntiTrendStrategy: price1=last futures price, and price2= second last futures price.If price1 is lower than price2 then the agent orders buying. If price1 is higher than price2 then the agent orders sell. The amount of order is randomly decided. RandomStrategy: The agent buys or sells randomly. The limited price on order is set randomly around the latest futures price, and quantity of the order is set randomly within a prescribed range. Position of the agent is also considered in decision making.
SRandomStrategy[54]: The agent buys or sells randomly. The limited price on order is set randomly around the latest “spot price”, and quantity of the order is set randomly within a prescribed range. The position of the agent is also considered in decision making.
RsiStrategy (RSI: Relative Strength Index): The agent buys or sells according to the Relative Strength Index (RSI) o futures price. RSI is one of major technical analysis methods. The limited price is given randomly around the latest futures price, and quantity of the order is given randomly within a prescribed range. The position of the agent is also considered in decision making.
SRsiStrategy: The agent buys or sells based on Relative Strength Index (RSI) of spot price. RSI is one of major technical analysis methods. The limited price is set randomly around the latest spot price, and quantity of the order is set randomly within a prescribed range. The position of the agent is also considered in decision making.
MovingAverageStrategy: The agent sends an order when the short term moving average line of futures price crosses over the medium term moving average line. If the trend of the short term moving average line is up, the agent gives buy order and when it is down, he gives sell order.
SMovingAverageStrategy: The agent sends an order when the short term moving average line of the spot price crosses over the medium term moving average line.
If the trend of the short term moving average line is up, the agent gives buy order and when it is down, he gives sell order.SFSpreadStrategy: The agent orders when a spread between the spot and then future price is wider than threshold. If the future price is higher than the spot price, the agent gives buy order, and if the futures price is lower than the spot price, he gives sell order.
DayTradeStrategy: This is the kind of day trading strategy. The agent gives sell and buy orders simultaneously. The limit price of sell order is slightly higher than latest futures price. The limit price of buy order is lower than latest futures price.
In the latest version of the U-Mart system called U-Mart ver.4, the default agent set has been slightly expanded to incorporate MarketRandomStrategy. It is also noted that the names of agents based on the future prices are changed from TrendStrategy into UTrendStrategy, AntiTrendStrategy into UAntiTrendStrategy, RSiStrategy into URsiStrategy, Moving AverageStrategy into UMovingAverageStrategy.
Needless to say, a mixture of several of these strategies may be feasible.
One of the default settings is the random strategy agent. This means employing simultaneous random moves on mode choice (sell or buy) and limit order (price and quantity). The choice entirely depends on pseudo-random number generation by a computer. We can therefore use these as zero-intelligence agents, and conduct two kinds of zero-intelligence simulations: one for the futures market, the other for the spot market.
Over the long history of the U-Mart experiment, we have become familiar with the workings of the random strategy in the futures market. We have developed an empirical rule that the random strategy is not defeated by many other strategies, and it may even be a winning strategy when all other agents are similar. This experiment is quite easily run by a standalone simulator (Fig. 4.6), and I recommend trying it to expand your repertoire of market simulation and discover new heuristics in market trade properties. We use this strategy as a test for new machines.
4.4