How to assess the effectiveness of a trading system: key metrics and practical tips

Hello everyone, I was wondering if someone has some useful general guidelines to evaluate the effectiveness of a trading system. For example: A good trading system should generate consistent benefits both in real time and backtesting. To what extent is it reasonable to go back in time to consider that the backtest is valid? 1 year, 2 years, more? A good system should be profitable over several years, but it should also generate benefits on a considerable volume of operations. I mean that, if you only do two operations a year, is it less reliable than a system that performs 1001 annual operations. What volume of operations would be good? If good systems generate consistent benefits, what would be an acceptable drawdown? No more than 6% or 11% of the total balance of the account? I read that, when more people use the same system, especially if it is automated, its effectiveness decreases. Large operators can anticipate operations or go against the system to make it less effective. Some in forums do not agree with this, which reassures me a little. But let�s suppose that this is true: how many people would be too large operators operating the same system

I appreciate any answer. It is that the more I read, the more different opinions I find. Some say that the 2-year backtesting is enough, others that at least 5 years. What is the reality?

At 2 years you have a decent sample if your system operates on a daily basis. But if you do few operations a month, best minimum 5 years. The more historic, the more solid the system is.

The truth is that not even 10 years of backtesting are you safe. The market changes. The only thing that's worth the forward testing. That's where you really see if it works.

Do 3-year backtest to get a good sample and then forward demo test for 6 months. If you survive both, it is viable.

On the drawdown, do you think it's reasonable for a system to have a 10% drawdown?

10% is within tolerable, but it depends on your aversion to risk. I am already uncomfortable with more than 7%.

It all depends on your time horizon. If you operate swing, an 11% drawdown can be normal. If you do intraday, it�s too much.

Drawdown is the only truth that matters. Don't notice how much you can earn; ask yourself how much you're willing to lose without going crazy.

Interesting... the number of operations also worries me. Is a very active system better or a more selective one?

Better asset, systems with few operations are dangerous because they don't generate enough statistics, with few operations everything seems nice until it explodes.

Not necessarily. I operate a system that makes about 8 entries a month, but has 4 years running solid. Quality before quantity.

A lot of activity usually translates into more commissions and psychological wear. If the system is good, you don't need to operate like a hamster on the wheel.

So, is there an ideal number of operations per year to validate a system?

More than 300 operations per year are beginning to be statistically significant. Less than 100 are unreliable.

Don't look for magic numbers. Look for consistency in different market conditions. That's worth more than any specific number.

If your system only has 40 trades a year and has small drawdowns and sustained profits, why would you want to complicate it?

And on the subject that many using the same system can ruin it, is this true or paranoia?

When too many follow the same automatic strategy, the institutional ones smell it and position themselves against it. The market always adapts.

There's some myth about that, too, and in order for it to affect, there would have to be millions behind the system, the retailer doesn't move that much.

The problem is not how many people use it, but how obvious the system is. If everyone enters the same level, it is easy to hunt.

I meant whether the institutionals look at the number of people or the total volume of money operated.

They always look at the volume, not how many traders are behind it. They don�t care if it�s a single trader with $501,000 or a thousand with $1001. Money moves the market, not names.

Even if they are many with small accounts, the danger is that they all enter the same technical levels. That generates attractive liquidity zones to be hunted.

It's simple: the more predictable your entry point is, the easier it will be for big players to clean you up.

So, are very popular systems doomed to fail?

Exactly, which works well today, when it becomes massified, it stops working, which is why banks constantly change their models.

Not necessarily. There are robust strategies that survive decades, but they are often more complex and adaptive, not simple media crossing systems.

That's why internet selling systems usually last a short time. When all fools buy them, they become useless.

So a manual system has more life expectancy than an automatic one?

Totally. Human discretion can adapt to market changes much better than a blind robot following fixed rules.

Robots work while the environment remains stable. As soon as there is a disruptive event, only a trained trader knows how to improvise.

But eye: most humans improvise badly. That's why many robots beat mediocre traders.

Is it better then to have a semi-automatic system with human intervention?

That's ideal. You automate mechanical execution to eliminate emotions, but you maintain strategic control.

I use robots to shoot orders, but I'm the one who defines when to activate or shut them down according to the macro context.

The semi-automatic is the perfect midpoint: the precision of the machine with the flexibility of the human.

Another question: does it make sense to pay for commercial systems that sell on the internet?

In 99% of cases, it's throwing money away. Good systems aren't sold, they're protected.

If the system was so good, why would they sell it for $301? They live by selling illusions, not operating.