How to Become Profitable (algotrading for beginners)

[AriaS]

1. Backtest/optimize everything you possibly can, across every market you possibly can, until you find something that seems to work out-of-sample (new/unseen time period that you never used for tweaking/optimizing). Use your own or modular algo**.** Don't use the closed commercial algos - they are usually overfitted by their sellers. Also b careful with strategies and markets that suffer from heavy slippage and other execution problems.
2. Validate through many cycles of walk-forward analysis (WFA) on historical data. If it passes this most important reality check, you probably have an edge. After optimizing/tweaking on a certain period ("Optimization-Period"), you will need to decide what setup to choose and test on the "Future-in-the-Past" - a period that follows the "Optimization-Period". You will need a selection criteria. For example, a setup that works well on the period that precedes the Optimization-Period, plus some problematic periods (stress tests), plus additional tests like Monte Carlo, etc. The goal is to see what selection criteria consistently provides a setup that works best on the "Future-in-the-Past". When you eventually trade live, that period will be your real future.
3. Move your WFA process to the present. "Future-in-the-Past" will be the real future now. Trade it on a small live account and keep comparing the live results with their corresponding backtest results every day or two. Live performance and backtest performance must reasonably match.

'We don't start with models. We start with data. We don't have any preconceived notions. We look for things that can be replicated thousands of times.'
- Jim Simons