The ProRealAlgos proved algo development process

Let's talk trading ideas. Let's talk about your sources of trading ideas and let me share with you our own favourite sources of ideas. I will also share with you 3 interesting ideas from our own backlog, that we haven't had the time to test yet. Enjoy.

• prorealcode.com
• quantifiedstrategies.com
• Books from the video
• Ideas spreadsheet (make a copy)
• Join the community chat

Need to learn the essentials of ProRealTime and the basic coding language? Here's a video covering the must-haves. If you already know the basics of ProRealTime and have some basic understanding of programming, you can skip this video.

I'm going to show you the interface of ProRealTime and the features you will use when coding algos. This is not a course on programming, but I'll quickly show you the basics of what you need to know.

Open the playground and pick "Two moving averages". Change the periods and run it. Then write one of your own ideas as a condition and see it mark the bars it fires on.

The interpreter is a strict subset of ProBuilder, never a superset, so anything that runs here is valid code in ProRealTime. When you write something it does not carry it says so instead of calling it an error.

Open the drill

We get into how to code and evaluate trading ideas in ProRealTime by coding the three ideas we covered two videos back. Carl will show you the 4 metrics that ProRealAlgos use to evaluate trading ideas quickly and effectively.

• Trading idea #1

In this video we start adding some filters, features and finess to the trading idea that we coded in the last video, to make it even better. Slowly turning it into a viable system that will create a passive income stream for you for years. So grab a coffee and join me.

• Full code including new filters and features

The Stochastic MACD indicator is a combination of the stochastic oscillator and the regular Moving Average Convergence / Divergence. Let me show you how to use it with ProRealTime to find the best possible entries and exits.

• SMACD + Divergence + Backtester .itf files

When building a trading strategy the possibilities are limitless. You can add hundreds of different indicators, filters, oscillators, codes and features, but I have found that new developers find it hard to know where to start, just because there are so MANY possibilities.

So I am going to help you get started by guiding you with some simple filters on trading hours, days and seasons.

• Full code of the DAX40 strategy

One bar in the playground is one hour, counting from midnight on Monday 1 January 2024, so CurrentHour and CurrentDayOfWeek carry real values and the whole of chapter 2D is practisable.

Load "Trade only certain hours" and set the market to each Mystery series in turn. One of them pays only inside a narrow window of the day. Move the window until you find it, then reveal and check you were right.

Open the drill

In this learning video Carl goes through several methods to test and evaluate your algos, and their different advantages and disadvantages. There are no perfect ways, but there are definitely methods to avoid.

Walk-forward testing is the antidote you need to avoid overly optimizing your backtest and to make the best use of the available historic data. It is widely recognized as the de facto method for backtesting and optimization.

• Full code (.itf) of the Walk-Forward example

Set the market to Pure noise and load "Seven parameters on noise". The series has no edge of any kind. Nudge the parameters until the equity curve turns up. It will, because with seven knobs and a hundred trades something always fits.

Then change nothing except the seed and run it again. Across five seeds on data with no edge whatsoever, the same code scores profit factors from 0.71 to 1.34. That spread is overfitting, measured on your own screen rather than asserted.

Open the drill

Load "Buy after three red bars" and run it against each Mystery series. One of them has that exact rule planted in it and the rest do not.

Over six seeds the planted one scores a profit factor around 1.84 and pure noise around 0.86. On any single seed those two overlap. That is the real lesson: one backtest cannot separate them, and you need more than one look.

Open the drill

A system you licensed is in drawdown. You get what you would really have: the record you were shown before you bought it, and the record since. Keep it running, or stop it?

One scenario in three keeps its edge the whole way. One loses it the day you go live. One never had an edge and the backtest was a good run on noise. Because the bars are generated the right answer is a fact rather than an argument, which is the only reason a drill like this can exist at all.

Scoring well above chance is hard, and that is the finding, not a flaw. When you cannot reliably tell a dead system from an unlucky one, what saves you is position size and a rule written before you started.

Open the drill

Decide this while nothing is at stake, because you will not decide it well halfway down. A usable rule names a number and a horizon, not a feeling.

Three of the pieces are already in the ProRealAlgos support answers: run the systems together, because one is in drawdown while another is making money; check in weekly or monthly rather than daily; and remember that stopping an algo does not close its open positions.

Every page of the reference ends with a section called common errors and gotchas. There are 1,168 of them across 301 instructions, each one a specific way that correct looking code does the wrong thing.

They also work as a test suite. Running the code from those sections through this playground found three places where it was more forgiving than ProRealTime, including accepting ADX[14](close), which the platform rejects.

Open the reference