Three Types of Quant Trading

Introduction

CTA is probably the most hand-on strategy for trading among varieties of candidates. 

What is Alpha?

Alpha means the ability of absolute return, that is the ability of exceeding the baseline, and the baseline is the major market performance.

Which ever future or equity we buy, we can simply have a measure between them a

Three Types of Alpha-based Strategy

  • A Type Alpha Strategy: Find good indicator

Most popular research method finance research paper. Restrictly seperate the selection of equity candidates and the timing for positioning.

 

1. MACD: exponetial smoothing (time dependent) +

Trend speed = (long-term trend - short-term trend)/time

This is not a equity selection, because this is an indicator of the price, not the indicator of the strength of trend for individual equity.

Major message: we need a rankable measurement between a bunch of equities.

2. 市值因子

3. 市净率因子 PB

Backtest：

When we fixed a specific factor, backtest is the method for validation:

1. candidate pool

2. positioning period （什么时候买）: 5/10/15/ days, 

3. selection out of a rank list (买什么) : quantile into 4 sections, see which part in the ranking list is good for the selection

4. weight of each equity (Portfolio) （买多少）

5. calculate the value (tracking the price at each time epoch)

The procedure:

http://bbs.quanttech.cn/uploads/article/20170627/4f033d10eb6317d49f5228604f226e1f.jpegAn example : Twitter factor.

Note that: The performance of a single factor is a marginal factor.

 

Multiple factors:

(1) Investment fundamental analysis

Sharpe ratio (SR) by day, information ratio (IR)

(2) Ability of prediction for a single factor (Robustness)

Correlation between factor of current period and the return in the next period.

可以用于Plot一个随时间变化等趋势，了解因子得分的半衰期

Machine Learning:

Advanced way to create the ranked list.

Use the current factor to predict next return.

This constructs a regression model.

Research Question: 

Is this machine-learning-weighted factor better than the primal equally-weighted factors?

Under what condition it is better?

B Type Alpha Strategy: Find good portofolio

  • Good for institutional investors. No indicator for equity selection. We first explain the return, the target is no the predict the return, the target is to explain the current market status. This explanation can be used for finding the optimal portfolio.
  • The whole technique is based on, so-called smart-beta. 

Trust one thing： return is explanable: 同期解释同期

Return is a linear function of risk explosure. = 风险溢价＋ 风险补偿

（1）Macro (Fama-French)

市值、估值都是风险：

－ 小市值股票意味着更大的破产风险，欺诈风险

－ 低估值也是风险因为分析师已经给出了对资产的负面评价（理性资产定价解释）

－ 盈利能力： 

－ 投资风格：基金是风险

－ 动量： 

factor是构造出来的，从而得到

这个不是横截面回归，是时间序列回归，针对每只股票估计beta的风险暴露。如果知道了每只股票的风险暴露，就可以计算投资组合的风险暴露，这一步是一个优化问题。

在未来的市场，我要优化的是一个单位风险收益率，让这个优化问题里风险＝0。

要点［对冲的思想］：寻找w_i，把所有的beta_i 优化成0，也就是说对于我没有看法的风险，对冲掉。

 

这个模型承认了自己的无知，但是拥抱它，把它优化成0。

所以我们不选股，就在最大的池子里，把组合优化成0。

高分红，低估值是近期A股市场上表现最好的因子。

 

（2）Fundamental （APT，APM， Active Portfolio Management）

要在横截面上先估计风险暴露，然后再做时间序列回归。

BARRA：经典模型，表达对风险的看法。

QTARME1: 量邦科技。

这个有点深度学习的味道。

 

（3）Simple B-type Alpha

均值－方差优化：给定收益，最小化投资组合风险，保证组合的市场风险暴露是1。

 

风险就是资产，做资产配置不如做风险配置。

对风险的配置就是对资产的配置。 by Bridgewater Investment

 

X Type Alpha Strategy:

  • ad-hoc method for action, the most practical method for taking actions for individual investors. 
  • S1: 建仓日检查上证50成分股中均线金叉突破的股票，也是短线均线突破长期均线，将所有突破的股票等比例买入
  • S2: 每天检查持仓股票，卖出死叉的股票，同时用闲置资金买入当天金叉的股票

不区分选股和择时。

 

In this article, we mainly focus on CTA strategy.

Note that, the majority of the current public quant trading platform fits for X-type Alpah strategy.