Quantitative strategy development, high-quality community, sharing of trading ideas and other related content
Hello everyone, today we share the 5th strategy of 2023 - non-timing CTA single factor strategy.
This strategy is the fifth in the 2023 general series. Today I am not going to talk about algorithms, indicators, or appearance issues with you. The main discussion is the application of the stock multi-factor framework system to futures. This is very rare in public policy materials in the market.
In the past CTA real offer or research process, we always determine what kind of products to use, and then use timing strategies to open and close positions. However, whether you are a B-side or a big C-side, most of these timing strategies are used Much the same with minor differences, a small part is heterogeneous, the only performance attribution is that the variety or cycle is different.
1. Strategic logic
In the stock multi-factor framework, a multi-factor model is a quantitative model used to analyze and explain asset price movements. It is based on the assumption that price changes in assets can be explained by multiple factors. These factors can be macroeconomic indicators, industry indicators, corporate financial indicators, etc., and their relationship with asset prices can be modeled and analyzed through statistical methods.
The core idea of the multi-factor model is to express the relationship between the return on assets and multiple factors as a linear equation. This equation can be written as follows:
R = β1F1 + β2F2 + ... + βnFn + α + ε
Among them, R represents the rate of return of assets, F1, F2, ..., Fn represent the values of multiple factors, β1, β2, ...,
βn represents the coefficient of each factor, α represents the constant term (also known as excess return), and ε represents the error term.
Of course, I won’t go into details about the specific advantages and disadvantages here. After all, there are a lot of academic explanations on the Internet. Let’s focus on the application of futures. In the commodity futures market, multi-factor models can be used to analyze and explain changes in commodity prices and help investors formulate trading strategies.
The process of applying a multifactor model to commodity futures is as follows:
Factor selection:
First, you need to choose a factor that is applicable to the commodity futures market. These factors can include macro
Economic factors, supply and demand factors, seasonal factors, volume price factors, etc.
Data collection and processing:
Collect and organize historical data related to selected factors.
Model building:
Build a multifactor model to explain commodity price changes
Model Evaluation and Validation:
Evaluate and validate the constructed multifactor model
Trading strategy formulation:
Based on the analysis results of the multi-factor model, formulate corresponding trading strategies
I won’t go into details about the specific process here. I will explain it in detail in our offline courses in the fourth quarter. For example, in the last step of “trading strategy formulation”, according to different factor weights and price forecasts, contracts can also be traded up and down. Constraints, factor exposure constraints, and quadratic programming for combination weight assignments in multi-factor models, etc.
In this strategy, we only use a single factor - the momentum factor, to conduct a cross-sectional long-short combination backtest on the case workspace. As shown below:
We are long TOPN and short 1-TOPN according to the sorting of factor values in the cross section (because it is a single factor, so there is no so-called scoring link). We didn't choose all the varieties brainlessly, but chose several representative ones from different sectors, as well as individual non-popular varieties.
There are 13 varieties in the working area of this case. I adopt the simple logic of more first half and empty second half, as shown in the figure below:
In fact, the logic is not complicated. In this phase of the strategy, we have no logic of timing entry, no logic of adaptive exit, and no endless parameter optimization (0 parameter). Some are just factor validity, stability, significance, etc.
Secondly, this CTA strategy strictly depends on the type of factor. If the factor is fundamental, reversal, etc., it is not actually a trend-following CTA strategy.
2. Performance
combination
Ur
Cf
SC
The main performance stretch is that the weight market value of crude oil is equivalent to the sum of other varieties, and crude oil has been in a retreat since May 2022. This also means that the momentum effect of crude oil has diminished in the last year. In the multi-factor research process, the volume-price reversal factor and the risk control factor can be added to neutralize the momentum factor, so as to achieve cross-sectional performance stability.
3. Summary
-
A fan friend asked me a question. He asked me that the factor can directly backtest the results, so why should I study the factor alone? This question is actually very good. Most people don’t really understand the difference between research factors and backtesting. Let me briefly say:
Although the results can be directly backtested, it is still necessary to study the factors separately.
The main reasons are:
1. Analyze the internal logic of the factor and understand the source of its Alpha. It is difficult to understand its principles directly by looking at the backtest results.
2. Analyze the economic significance of the factor and judge whether it is sustainable. Backtesting cannot judge future performance.
3. By studying the instability and periodicity of factors, optimize the timing of their use. Backtests reflect historical performance only.
4. When combining multiple factors, it is necessary to study the correlation between the factors. The backtest does not take factor combination effects into account.
5. Study the robustness of factors in different market environments, and adjust the use strategy. Backtesting is based only on certain historical market conditions.
6. Compare similar factors to find the best choice, or find a new factor as an alternative. Backtests do not support factor comparisons.
7. The research industry or specific factors, the backtest data is not rich enough. Therefore, studying factors individually can give us a deeper understanding of their working mechanisms and optimize their use. Direct backtesting has limitations, but factor research can make up for these shortcomings and achieve better investment results. Both should be used in combination.
-
In terms of fund management, I noticed a problem. In the past, the calculation was based on the reinstatement price, but the funds have not been reinstated. Therefore, in fact, the price should be calculated by deweighting the price, as shown in the figure below:
-
This strategy is a daily cross-sectional single-factor strategy, with few factors and no capital management, so it seems unstable, but it provides us with rare ideas and inspirations for CTA quantification at the factor combination level. In the offline courses in the fourth quarter, we will also provide you with as many factors as possible and 50+ full-variety CTA combination strategies.
Due to differences between platforms, the backtest performance is subject to the QMT version! ! !
This strategy is only for learning and communication, and investors are personally responsible for the profit and loss of real trading transactions! ! !