> This English version was automatically translated from the Chinese original. Please refer to the original Chinese article if any wording differs.

My thoughts on the real-time experience of quantitative procedures

When backtesting a quantitative program, you behave like a calm, self-disciplined top student with a bright future.

As soon as the real offer arrives, I feel like a person knocking on the door on time every morning:
"Hello, I want to test your faith today."

Many people think that the most difficult part of quantitative trading is writing strategies. Later I discovered that the real difficulty is watching the strategy lose money according to the rules, and you can't just intervene casually.

1. The backtest is a movie clip, and the live test is a live broadcast

The backtest curve is decent.

The market trends of several years are compressed into a picture. The loss is just a small depression, the retracement is just an indicator, and the maximum retracement even looks a bit professional.

But the actual offer is not like this.

The real offer is unfolding day by day. Lose a little today, grind it out tomorrow, and fail to rebound the day after tomorrow. A centimeter of sideways movement in the backtest may mean two months of silence in reality.

Only then do you realize that it’s not that you can’t accept the retracement, you just can’t accept it and clock in every day.

2. Mathematical expectations are beautiful, but human feelings are noisy.

Quantitative strategies are about long-term probability.

It may not need to be right every time, just that it has an advantage in the long run. Sounds reasonable, even a bit advanced.

The problem is, people don't live by long-term probabilities.

What people see is:

It fell as soon as I bought it.
It went up as soon as it was sold.
I made several losses in a row.
It shouldn't have been believed yesterday.
Shouldn't we believe it even more today?

Strategies count expectations, people count heartbeats.

3. The backtest world is too clean, and the live floor is a bit slippery

In the backtest, the transactions were very obedient, the slippage was very gentle, and the handling fees were very polite.

It's different in the real market.

Slippage, handling fees, bid-ask spreads, delays, insufficient transactions, and liquidity problems will all eat away at profits bit by bit.

The backtests of some strategies seem to have good annual performance, and the actual operation seems to be doing joint public welfare for brokers and the market.

Especially for strategies with frequent transactions, friction costs are not a small problem, but come to collect utility bills every day.

4. Overfitting is the best way to disguise yourself as genius

Many strategies are good at backtesting, not necessarily because they have discovered patterns, but also because they have made the historical data happy.

Adjust the parameters, change the indicators, choose the time period, and finally a beautiful curve will come out.

Looks like a genius.

But as soon as the real offer started, the market immediately said:
"You memorized last year's answers."

The most troublesome part of overfitting is here. It will not tell you that it is unreliable at the beginning. It will usually give you a good-looking backtest chart first.

5. The market will change its temper, and the strategy may not keep up.

Strategies are often suitable for a certain market environment.

Trend strategies like trends, shock strategies like shocks, and momentum strategies like the strong to stay strong.

But the market is not responsible for cooperating.

Sometimes it is trending, sometimes it is oscillating, sometimes it is pretending to be trending, but it is actually oscillating; sometimes it is pretending to be oscillating, but then it suddenly punches you.

When the strategy fails, there is little sound. It usually gets worse slowly, leaving you wondering every day: is this a normal retracement, or is it really not working anymore?

This is the most torturous part.

6. The difficulty in making a real offer is not running the program, but the other party’s debt.

The advantage of the quantitative procedure is that there is no need to temporarily slap the head during the session.

The downside is that you'll end up trying to slap your head off the board.

Do you want to stop?
Do you want to change the parameters?
Do you want to add filter conditions?
Do you want to lower your position first?
Do you want to pretend to be rational?

Many strategies die not from the market, but from the inability of humans to tolerate them.

Of course, blind persistence may not be right. What is really difficult is to distinguish clearly: whether you are observing discipline now, or whether you are using discipline to package stubbornness.

7. The more real the money, the more dazzling the curve.

In simulated trading, losing 10% is called risk control.

In the real offer, if you lose 10%, I will open the account repeatedly before going to bed tonight.

When the capital is small, the fluctuations are like numbers. With larger funds, the fluctuations begin to take on a life-like flavor: rent, salary, travel, and year-end bonuses can all be automatically converted.

Therefore, the strategy is not just to be able to run mathematically, but also to be able to endure it psychologically.

A high-yield strategy that you can't stick to may end up being inferior to a mediocre but long-term strategy.

8. Good strategies can also make people feel uncomfortable

Many people mistakenly believe that good strategies should always give positive feedback.

But the reality is that good strategies can also go sideways, retrace, underperform the market, and make you wonder if you have bought a program that automatically creates anxiety.

The most anti-human thing about quantification may not be losing money, but the lack of feedback for a long time.

It won’t comfort you every day. It just plays by the rules and leaves you to deal with your emotions.

Conclusion

I feel more and more now that quantitative procedures do not make trading easier, but change the pain.

It reduces on-the-fly decision-making but increases the difficulty of waiting, doubting, and persevering.

Backtesting tells us: This strategy may have worked in the past.
The real offer asks us: Can you really bear it?

Therefore, if the real-time experience of the quantification program is not good, it does not necessarily mean that quantification is useless.

It just reminds us that money in the market is never made comfortably. Programs can help us enforce rules, but they cannot bear uncertainty for us.

In the end, what is really important may not be to find the most beautiful backtest curve, but to find a system that you understand, can afford, and are willing to implement for the long term.