Testing ideas honestly

Risk, psychology & your plan · All levels · 7 min read

You will learn to

The five checks before trusting a strategy
Why most ideas fail
Using the Lab

Five checks

No look-ahead: only information available at the time (see repainting).
Fees and slippage included.
Out of sample: build on one period, test on another you never looked at.
Against random: compare with random entries using the same stops and targets. Many "edges" are just the market drifting up, or the exit doing all the work.
Enough trades, with a confidence range — not just an average.

✓ What our backtest found

Of everything tested in NexusCore (candles, S/R, supply & demand, trendlines, confluence, chart patterns across 15M–4H), only 4H chart patterns beat random entries out of sample. That is normal: most ideas do not survive honest testing.

Lab → Backtest runs your own rules this way (fees, next-candle entries, random comparison, confidence range).

Check yourself

1. Why compare a strategy with random entries?

To make it look better
To check the result is not just drift or the exit shape ✓
Required by exchanges

If random entries with the same exits do as well, the entry adds nothing.

2. A strategy made +0.5R per trade over 12 trades. You should…

Go live with full size
Treat it as unproven — far too few trades ✓
Delete it

Twelve trades cannot separate skill from luck.

Terms in this lesson

Out-of-sample: Data not used while building a strategy, kept for testing it.
Overfitting: Tuning a strategy to past noise so it fails on new data.

Practise it on real charts. In the app, this lesson comes with real examples (reveal what happened next), a graded drill and the track test. Open the Academy in NexusCore — free →