Gate Square “Creator Certification Incentive Program” — Recruiting Outstanding Creators!
Join now, share quality content, and compete for over $10,000 in monthly rewards.
How to Apply:
1️⃣ Open the App → Tap [Square] at the bottom → Click your [avatar] in the top right.
2️⃣ Tap [Get Certified], submit your application, and wait for approval.
Apply Now: https://www.gate.com/questionnaire/7159
Token rewards, exclusive Gate merch, and traffic exposure await you!
Details: https://www.gate.com/announcements/article/47889
#量化 #量化交易 #量化策略
Overfitting is a necessary path for personal quantification.
Every beginner in proprietary quantitative trading will encounter a problem - overfitting.
Seeing some strategies from great masters, the principle is simple, the logic is clear and straightforward.
Clean and effective, it's inevitable to feel a yearning.
At this point, many people may worry: Is overfitting really scary? Should we do everything possible to avoid it?
In fact, overfitting is not so scary.
Quantitative strategy development is essentially a process of "first imitating, then surpassing."
It's like playing a game; at first, you'll start with simple strategies and then continuously complicate it, adding your own understanding.
Once the experience accumulates to a certain stage, complex content can be stripped away, leaving only the truly valuable core logic.
This is very similar to starting a business.
In the early stages, you might replicate someone else's business model, and after it runs smoothly, gradually optimize it, eventually developing your own unique model in the later stages.
Quantitative strategies are the same, starting with "copying" and then evolving.
When just starting out, there is no need to overly worry about issues of overfitting or strategy complexity; what matters more is to "take action first."
Only by continuously iterating and validating in practice can one form their own system.
Strategy development is actually a process of "negation of negation", a process of spiral ascent.
We cannot achieve it all at once and directly mine simple and effective alpha.
In the early stages, one should first do "mindless addition," continuously increasing features and trying out ideas. Once a certain scale is accumulated, then do "subtraction."
Keep the essential parts and remove unnecessary redundancy. This is the process of growth: first addition, then subtraction.
So, if you want to become stronger in personal quantification, don't overly worry about whether it's overfitting at the beginning.
Only by continuously practicing can one form their own understanding and style, and then truly mature.