Putting It All Together: Software, Data and Sample Size for Breakouts

Now you know everything about my breakout formula.

POI. SPACE. FILTERS. TIME. EXITS. Every component, explained down to the detail I promised.

And if you are even slightly ambitious, if you actually want this to work, you have already written out your own POI and FILTER ideas. On paper. Not in your head.

So what happens next?

What are the next steps for putting the formula into practice?

Two things. That is all.

1. Create your own list of POI and FILTER ideas. You already have all the answers for SPACE, TIME and EXITS from me. Those are solved. POIs and filters are where your own thinking goes.
2. Experiment, experiment, experiment.

That second one is the magic key. Trying many different combinations until you find what sticks, while keeping your number of optimisation inputs between 3 and 6 to keep overfitting and curve fitting away.

Three to six. Not eight. Not “just one more, it really helps.”

I know that for some people this still is not enough, so let me cover the three questions I get asked most: software, data history, and sample size.

Which software should you use for breakout strategies?

Beginner traders are far too preoccupied with this question.

Yes, you need software to build, test, optimise and trade your strategies. But software is a tool. There is no perfect solution, and every trader has a different preference.

The good news: the Mr. Breakouts model is universal. It can be applied in any software. I know traders who have coded the whole thing in Python.

I cannot tell you much about anything other than TradeStation, because it is the only software I have ever used. It meets every criterion I need. It is free once you open an account with them. It gives me all the data history I need. And EasyLanguage is simple, easy, and very well documented on their pages.

I built all my own breakout strategies in TradeStation, along with the 2,041 strategies in our hedge fund. We use it for trade execution too.

I am simply not interested in other platforms. I know traders who have coded my formula in NinjaTrader, MetaTrader and others. I do not know how exactly, and I do not need to. For me TradeStation is the industry standard and the best choice.

Do not spend three months choosing a platform. Spend three months testing ideas.

How much data history do you need?

This question matters far more than the software one.

You need at least 10 years of good quality data history. That is what I build my strategies on, and it is the history we use in the hedge fund.

Ten years, because you need to cover a range of volatility regimes, trends, and economic circumstances. Deep historical analysis is what makes a strategy robust. If you want to maximise the odds that a strategy works in the future, you must confirm it works well across a variety of market conditions, not just the last two years you happen to remember.

The 10-year rule is another golden rule.

On top of quantity, you need quality, and specifically good X-MINUTES data. The formula works best when tested on 30-minute, 60-minute, 90-minute, 120-minute, or similar periods.

Daily data can get tricky. The X-minute timeframes are the most reliable.

So check your data provider meets all of this before you start. With TradeStation I never think about it: high-quality history is included down to 1-minute resolution, and continuous contracts for all futures markets come with one click. That saves an enormous amount of time.

What is a valid sample size?

In algorithmic trading, sample size matters. Every strategy you build has to clear a minimum bar to be statistically meaningful.

The criteria:

• At least a few dozen trades per year.
• With 10 years of backtest history, that means at least a few hundred trades.

In our hedge fund, our sample size over the 10-year history runs anywhere from a few hundred up to 2,000 trades per strategy.

If your beautiful equity curve is built on 34 trades, you do not have a strategy. You have a coincidence.

What about robustness testing?

This is the one thing I have not covered, and it is a standalone topic of its own. I promised you the formula in the book, and I delivered it down to the slightest detail. Robustness deserves its own treatment.

But here is the guidance that matters, in three points:

1. Use the toughest robustness testing procedure you can come up with. You may need to do some studying online. The general advice: combine several methods together and set the bar as high as you can stand.
2. If you already have a robustness procedure, whatever it is, ask how to make it tougher. Not easier. Tougher.
3. If you do not know where to start, a simple multi-market validation is a good starting point.

That is the whole formula, start to finish.

The components are not complicated. Yesterday’s close as a POI, ATR-based space, a simple filter, a time window, a USD stop and target. None of it is exotic. What separates traders who make this work from traders who do not is not cleverness.

It is the willingness to test hundreds of combinations, keep the input count low, demand 10 years of data, insist on a real sample size, and throw away everything that does not survive.

Experiment. Measure. Discard. Repeat.

That is the job.