The proposed generator incorporates the information within order book microstructure and market news to provide directional predictions. The MM strategy in the second tier trades on the signals and prevents itself from profit loss led by market trending. From the empirical results, we find that 1 strategies with signals perform better than strategies without any signal in terms of average daily profit and loss PnL and sharpe ratio SR , and 2 correct predictions do help MM strategies readjust their quoting along with market trending, which avoids the strategies triggering stop loss procedure that further realizes the paper loss.
An intelligent market making strategy in algorithm ic trading. Abstract Market making MM strategies have played an. MM strategies without any forecasting power are not safe.
In this paper, we design and implement a two-. The proposed generator incorporates the infor-. The MM strategy in the sec-. Using half a year price tick data. Keywords algorithmic t rading, market making strategy , or-. Received August 24, ; accepted January 16, A market maker refers to a bank or brokerage company that.
In particular, the market making MM strat-. Due to the fast speed and high accuracy , the MM. The MM strategy is doing the job of specialists who. When the prices change within a. There are two main risks on MM. This risk is especially severe for.
As illustrated by [ 3], when market price is. Since the MM strategy is a fully automatic. There are many research papers in literature about MM. To formulate, the proposed strategi es could. The market condition p t refers to events extracted from price. And Formul a 1 is changed to. In this paper, we develop a trading signal generato r and. W e employ support vector machines SVMs ,. The second tier is the trad-. We back-tested this MM strategy with half a.
The rest of this paper is organized as follo ws. Section 2 is a. There are many works on MM strategy analysis. Mining signals from market prices and news articles have.
For the mining of. Similar result could be found. They applied SVM to. After comparing SVM with linear discriminant anal-. Market news, which is tradi tionally processed by human.
The basic motivation is to analyze the statistical relation-. AZFinT ext system, built by Schumaker and. Chen [13—16], was also able to give directional forecast of. One recent work shows that it would be better to use. Interested readers can refer to the references listed.
Xiaodong LI et al. An intelligent market mak ing strategy in algorithmic trading 3. However , due to the small number of pieces of news re-. For example, on December 28,. On the same day, there are hundreds of samples of order.
In order to fully use the trading opportunities, it would. Theoretical MM strategies have been proposed in many. Othman and Sandholm  ran an auto-.
GHPM which was designed and buil t to predict the opening. One theoretical analysis of market maker. In contrast, we use supervised machine learn-. The market rules, such as tick size i. The whole system is further back-. As sho wn in Fig. For example, market quote has many price.
In this way, the database no t only. It places orders dependi ng on the market conditions. Notice that there is a spare ma-. The pla tform has four components: As mentioned above, the main part of our MM strategy. In the following, we will show the de-. Order book information and market news articles are mod-. W e will refer to them as.
As shown in Fig. The series of the snapshots are then translated into order. Space Model  the processing detail is discussed in Sec-. All the samp les instances are then aligned with. Since we are g oing to pre-. W e then use SVMs to learn from. As explained in Section 1, our proposed approach uses two. This approach could make use of the sources,. This is why we c hoose SVMs as the pre-. Traditional technical indicators, such as exponential mov-. However, those indicators are de-.
Modern exchanges usually use order. An intelligent market mak ing strategy in algorithmic trading 5. W e consider such inf ormation more. The order book is constructed by two priority queues. Suppose we sample the historical. W e denote the series as S , and each snapshot as s i ,.
Prices decrease from bid 1. The arrow links the orders in each queue, by the priority of their limit price. Raw data series such as S is not useful, since changes at. Order book pr essure OBP  is constructed to summa-. Figure 4 gives an. At time point t , if the total size of. The size of the box indicates the queue size at that level. T o formalize OBP ,w e h a v e. Another type of signal comes from market news pieces. News agencies , such as Bloomberg, Dow Jones and. Those pieces of news capture the ev ent in the.More...