What are the considerations when building a "model trading strategy"?

In the field of investment, especially in the field of quantitative investment, the strategy of model trading can be roughly divided into trend trading type (including positive and negative, trends of different periods), regression trading type (including shock model, probability regression), hedging trading type (including Index and individual stock hedging, futures and spot hedging, different types of hedging).

All strategies are essentially predicting the future, because the basis for the next step of operation can only be a profitable judgment based on existing historical information. If the future cannot be predicted (that is, the future has nothing to do with history), then all Therefore, it is a misunderstanding to say that you only respond and do not make predictions.

However, model trading is a probabilistic event, and it can never be a decisive prediction. Therefore, even if there is a possibility of direction, the only way to obtain real unilateral income is through stop loss. No matter what trading strategy is adopted, stop loss is a must. The only way to make the strategy profitable.

Since stop loss is necessary, if you choose a single strategy, there is no possibility of divergent losses. Therefore, it is definitely impossible and unreasonable to try to attribute the bankruptcy of the US long-term capital fund to the misuse of normality. It is impossible for a few Nobel Prize winners to be so stupid that they can't stop losses but can use normality to make profits!

No matter what strategy model is adopted, there is a firm belief in one of the most basic laws of nature. The trend trading type believes that the world is mutated, non-linear, and has a specific structure, whether it is human nature or matter, the universe, time and space. in this way;

The regression trading type believes that the world will eventually tend to conform to the mean value of the normal distribution. No matter how specific it is temporarily and locally, the final result is always random and average;

The hedging transaction type is essentially the continuation of normality, and any positively related factors will eventually return if there is enough deviation.

The essential meaning of all algorithmic trading, from the perspective of information theory, is to obtain useful specific information for the future from seemingly random historical data as much as possible. For trend models, it is to obtain breakthrough trend information. For regression models, it is to obtain events that deviate from the mean. Come and wait for the return.

The most amazing thing is that the trend model and the regression model are strategically contradictory to each other, which is also a way for the world to balance. Structure and chaos, mutation and stability always run through every space-time and event, so it is impossible for one trading strategy to have an exclusive advantage.

Many people have a fatal misunderstanding, that is, they think themselves too smart. They develop a variety of contradictory models, and then automatically choose the most suitable strategy model in different time and space according to different conditions for arbitrage trading. All cases can be covered.

As everyone knows, even for one trading strategy, we can only rely on probability and stop loss to make profits. For multiple strategies, we must face the probability of strategy selection. If at a specific moment, two contradictory strategies happen to be chosen (there are many such cases), then the original stop loss will be destroyed, and the entire system may become divergent, leading to irreversible consequences !

Therefore, in the field of quantitative investment, any bankruptcy can only happen for the following reasons:

1. The leverage is too large, and the position will be liquidated when the funds are large enough to withdraw;

2. The liquidity is too small, even under the correct strategy, it is impossible to stop loss and trade;

3. Mixed strategies lead to divergence.