The three secrets of keeping a quantized model alive

The three secrets of keeping a quantized model alive

First of all, there is a general consensus that there is no such thing as a long-term model. There is a profound law here that financial markets have a self-regulatory function. If a model survives, then it will be like a snowball, taking more and more money out of the market, and reaching a critical point, the market will automatically adjust and change, causing the model to fail or become less efficient. So the next question is: How can a model slow down aging? After talking to a lot of bigwigs, I've summed up the secret principle of the golden rule, which I dedicate to you here, selflessly.

• The first principle is that the model should be built on stone, not sand.

What is the cornerstone of the model? The original idea behind the design of the model. Each model captures a certain law of the market. So, you need to ask yourself, is this law of the market long-term or limited short-term? For example, you design a high- and low-pull model based on the characteristics of narrow swings during periods of very low trading. Once the characteristics of the market change, the model obviously does not adapt.

• The second principle is to carefully maintain and upgrade it as if it were a woman.

Models are like women, they need regular maintenance and upgrades. Even if you find a market law that can be sustained over the long term. For example, too much will always fall and too little will always fall. This law is sustained over the long term. For example, a period may be 5% of the time, and a period may be 3% of the time. This involves optimization of parameters. In the meantime, adding data on new markets and then optimizing the overall picture.

• The third principle: symphony is better than solo.

In any case, a single model is unsustainable. Almost all the quantitative people I talk to agree with this. So, the combination of different models is very important. When a model A has a retracement, it is likely that model B is making money, at which point the net value curve of the entire model combination will be relatively smooth. It can be a combination of models with different strategies, a combination of models with different cycles, or a combination of models of different varieties. In general, the symphony is better than the solo. If you can fully understand these four secrets, you will succeed. Although I can't guarantee that your model will live forever, it is still hopeful to become an evergreen tree in the model.