/Quantative-Trading-Models

These are trading results and arbitrage models from Southern China Center for Statistical Science (SC2S2), Sun Yat-sen University

Primary LanguageJupyter NotebookMIT LicenseMIT

Quantative-Trading-Models

Python 2.7 vnpy hmm statsmodels

These are trading results and arbitrage models from Southern China Center for Statistical Science (SC2S2), Sun Yat-sen University

Research Details

We have completed designed the algorithm of intraday volatile mean reversion strategy and run it on RB1805,RB1810 object on vnpy.

See strategyIntraVolMeanRev.py for more details trading algorithm.

Meantime, we have implemented various statistical arbitrage model including

  1. Cross-star Arbitrage
  2. Hidden Markov Model based on Factors Decomposition
  3. Paired Cointegrative Arbitrage

Results

  • intraday volatile mean reversion strategy on RB1810
  1. Backtesting

Trading results from 05/2017 to 07/2017 on RB1805

see backtesting_strategyIntraVolMeanRev.ipynb for more details

  1. Minic Panel

Minic trading results on 05/06/2018 on RB1810

  • Hidden Markov Model arbitrage on CSI300

First, we illustrate the basic conception of HMM and write the augmented expected log-likelihood as

  • is the prior state probabilities
  • is the state transition probabilities
  • is the output probabilities

A example for the probabilistic parameters of a hidden Markov model is presented as below(Omit partial output probabilities for simplicity)

Based on RiceQuant, we obtain daliy open,high,low,close and volume of CSI300 (data) from 01/01/2005 to 31/12/2015 and denoted three feature-factors as,

  1. Computing logged daliy spread
Factor1 = np.log(np.array(data['High'])) - np.log(np.array(data['Low']))
  1. Computing each 5 days logged return spread
Factor2 = np.log(np.array(data['High'][5:])) - np.log(np.array(data['High'][5:]))
  1. Computing each 5 days logged volume spread
Factor3 = np.log(np.array(data['Volume'][5:])) - np.log(np.array(data['Volume'][5:]))

After that, we preset six potential states of CSI300 and begin our assessment in HMM_arbitrage.py

Finally, we demonstrate our trading return based on previous HMM prediction

  • Cross-star Arbitrage on dominant futures Pb
  1. Pb Dominant Futures Duration
start end symbol
2017/10/19 2017/10/23 PB1711
2017/10/24 2017/11/16 PB1712
2017/11/17 2017/12/19 PB1801
2017/12/20 2018/1/17 PB1802
2018/1/18 2018/2/14 PB1803
2018/2/22 2018/3/14 PB1804
2018/3/15 2018/4/17 PB1805
2018/4/18 2018/5/23 PB1806
2018/5/24 2018/6/26 PB1807
2018/6/27 2018/7/23 PB1808
2018/7/24 2018/8/20 PB1809
2018/8/21 2018/9/19 PB1810
2018/9/20 2018/10/19 PB1811
  1. Pb Dominant Contract Backtesting Return(ignore trading fees)

we run our Cross-Star strategy on Pb Dominant contract from 2018/05/05 to 2018/11/10. For more details in Integrate_Pb.ipynb

Set up

  • vnpy(vnpy-1.8)
  • hmmlearn
  • numpy
  • pandas
  • matplotlib

Contact

license