concept

Quantitative trading refers to the use of advanced mathematical models instead of artificial subjective judgments, and the use of computer technology to select a variety of "high probability" events that can bring excess returns to formulate strategies, which greatly reduces the impact of investors' emotional fluctuations and avoids making irrational investment decisions under extremely enthusiastic or pessimistic market conditions.

trait

Quantitative investment and traditional qualitative investment are essentially the same, both of which are based on the theory of inefficient or weakly efficient market. The difference between the two is that quantitative investment management is a "quantitative application of qualitative thinking", with more emphasis on data. Quantitative trading has the following characteristics:

1, discipline Make decisions according to the running results of the model, not by feeling. Discipline can not only restrain human weaknesses such as greed, fear and luck, but also overcome cognitive bias and can be tracked.

2. systematize. The specific performance is "three more". First, a multi-level model, including asset allocation, industry selection and specific asset selection; Second, from multiple perspectives, the core idea of quantitative investment includes macro-cycle, market structure, valuation, growth, profit quality, analyst's profit forecast, market sentiment and so on; The third is multi-data, that is, the processing of massive data.

3. Arbitrage thought. Quantitative investment captures the opportunities brought by mispricing and mispricing through comprehensive and systematic scanning, so as to find out the valuation depression and make profits by buying undervalued assets and selling overvalued assets.

4. Probability wins. First, quantitative investment constantly digs out the expected repetitive laws from historical data and uses them; The second is to win by combining assets, not by a single asset.

Applied editing

Quantitative investment technology includes many specific methods, which are widely used in investment variety selection, investment opportunity selection, stock index futures arbitrage, commodity futures arbitrage, statistical arbitrage, algorithmic trading and other fields. Here, statistical arbitrage and algorithmic trading are taken as examples to illustrate.

1, statistical arbitrage

Statistical arbitrage is a kind of risk arbitrage by using the historical statistical law of asset prices, and its risk lies in whether this historical statistical law will continue to exist in the future.

The main idea of statistical arbitrage is to find out several pairs of investment products with the best correlation, and then find out the long-term equilibrium relationship (cointegration relationship) of each pair of investment products. When the price difference (residual of cointegration equation) of a pair of varieties deviates to a certain extent, they start to open positions, buy relatively undervalued varieties and short relatively overvalued varieties, and take profits after the price difference returns to equilibrium. Hedging of stock index futures is a long-term operation strategy of statistical arbitrage, that is, using the index correlation of different countries, regions or industries to buy and sell a pair of index futures at the same time. Under the condition of economic globalization, the correlation of stock indexes in various countries, regions and industries is getting stronger and stronger, which easily leads to systematic risks of stock indexes. Therefore, the statistical arbitrage between hedge indexes is a low-risk and high-yield trading method.

2. algorithmic trading.

Algorithm trading, also known as automatic trading, black box trading or machine trading, refers to the method of issuing trading instructions by designing algorithms and using computer programs. In trading, the scope that the program can decide includes the choice of trading time, the price of trading, and even the number of assets that need to be traded finally.

The main types of algorithmic trading are: (1) passive algorithmic trading, also known as structural algorithm trading. In addition to estimating the key parameters of the trading model by using historical data, the trading algorithm does not actively choose trading opportunities and trading times according to market conditions, but trades according to established trading policies. The core of this strategy is to reduce the sliding price (the difference between the target price and the actual average transaction price). Passive algorithm trading is the most mature and widely used, such as transaction weighted average price (VWAP) and time weighted average price (TWAP) which are the most widely used in the international market. (2) Active algorithmic trading, also called opportunistic algorithmic trading. This trading algorithm makes real-time decisions according to market conditions, and judges whether to trade, the number of transactions, the transaction price, etc. In addition to efforts to reduce price slippage, active trading algorithm gradually shifts its focus to price trend prediction. (3) Comprehensive algorithm trading is a combination of the first two. The common way of this kind of algorithm is to disassemble the trading instructions and allocate them to several time periods. How to trade in each time period is judged by the active trading algorithm. The combination of the two can achieve the effect that simple algorithm can not achieve.

There are three trading strategies for algorithmic trading: one is to reduce transaction costs. Large orders are usually divided into several small orders and gradually enter the market. The success of this strategy can be measured by comparing the average purchase price and the weighted average price of transaction volume in the same period. The second is arbitrage. A typical arbitrage strategy usually includes three or four kinds of financial assets. For example, according to the exchange rate parity in the foreign exchange market, there will be some correlation between the price of domestic bonds, the price of bonds denominated in foreign currencies, the spot exchange rate and the price of exchange rate forward contracts. If the market price deviates greatly from the price implied by the theory and exceeds its transaction cost, four transactions can be used to ensure risk-free profits. The term arbitrage of stock index futures can also be completed by algorithmic trading. The third is to make the market. Market making includes hanging a limit order above the current market price or hanging a limit order below the current price in order to profit from the bid-ask spread. In addition, there are more complex strategies, such as the "benchmark" algorithm used by traders to simulate index returns, while the "sniffer" algorithm is used to find the most volatile or unstable market. Any type of pattern recognition or prediction model can be used to start algorithmic trading.

Potential risk

Quantitative transactions are generally tested by massive data simulation test and simulation operation, and positions and funds are allocated according to certain risk management algorithms to minimize risks and maximize benefits, but there are often some potential risks, including:

1. Integrity of historical data. Incomplete market data may lead to mismatch between model and market data. The style conversion of market data itself may also lead to model failure, such as trading liquidity, price fluctuation range and price fluctuation frequency. This is difficult to overcome in quantitative trading at present.

2. The position and fund allocation are not considered in the model design, and there is no safe risk assessment and preventive measures, which may lead to the mismatch of funds, positions and models and the phenomenon of warehouse explosion.

3. Network interruption and hardware failure may also have an impact on quantitative trading.

4. Homogeneous model will bring risks caused by competitive transactions.

5. Unpredictable risks brought by a single investment variety.

In order to avoid or reduce the potential risks of quantitative trading, the following strategies can be adopted: ensuring the integrity of historical data; Adjust model parameters online; Select the model type online; Online risk monitoring and avoidance, etc.