Updated October 2026. This article originally ended with plans to trade the strategy live and share signals. I no longer run live trading or signal services. The robustness checklist below is the part that matters, and it works for any strategy.
In this series we turned a TradingView indicator into a strategy (Part 1), tested it across markets and optimised its filters (Part 2), and set out the principles behind it (The Trader’s Mindset).
A great backtest still proves very little.
This part is about the tests that separate a robust strategy from a lucky one.
The problem: overfitting
Whenever you optimise a strategy, with filters or with an algorithm, there’s a risk of overfitting: building something that fits the past perfectly and fails as soon as the market changes.
A strategy is only worth trading if it keeps working on data and conditions it wasn’t built on.
Six robustness tests
1. Simple, understandable logic
A strategy should make sense in one or two sentences.
When it goes through a drawdown, and it will, you’ll only stick with it if you understand why it should work. Complexity often hides fragility.
2. Out-of-sample testing
Test the strategy on a period that played no part in building or optimising it.
If it only works on the data it was tuned on, it has memorised the past rather than found an edge.
3. What-if analysis
Stress the record with realistic problems:
- What if you remove the best 5% of trades?
- What if spreads widen or slippage increases?
- How does it behave in low-volume periods?
These questions expose dependencies that a headline number hides.
4. Monte Carlo simulation
Reshuffle and resample the trades thousands of times to see the range of outcomes the same edge could have produced:
- risk of ruin,
- the spread of possible returns,
- worst-case drawdowns, and
- confidence intervals.
If your backtest is one lucky path, Monte Carlo will show it.
5. Parameter sensitivity
Change the inputs slightly. Does performance stay roughly the same, or does it collapse away from one exact setting?
A robust strategy has a broad plateau of good settings, not a single sharp peak.
6. Different market conditions
Check performance across regimes:
- trending and ranging,
- high and low volatility,
- rising and falling markets.
A good strategy survives every phase, even if it only shines in some.
When is a strategy robust?
Only when it:
- has sound, simple logic,
- performs out-of-sample,
- survives the what-if stress tests,
- stays stable across Monte Carlo runs,
- isn’t sensitive to small parameter changes, and
- holds up across market conditions.
Anything less, and the backtest is a description of the past, not evidence about the future.
One strategy is not enough: build a portfolio
Even a robust strategy has losing streaks.
Combining several strategies that don’t move together can smooth the equity curve and cut drawdowns. Diversify across:
- instruments,
- timeframes,
- signal types, and
- strategy logic.
Then check that combining them actually reduced the drawdown, rather than assuming it.
How to run these tests yourself
You don’t need a quant team for any of this.
- JSK TradeProof runs most of these tests on your own trade export, inside Excel or Google Sheets: Monte Carlo, what-if scenarios, out-of-sample and walk-forward splits, market phases and portfolio correlation.
- My book Seven Tests Before You Go Live explains each test step by step, using nothing but a spreadsheet.
For a worked example, see how the optimised HalfTrend strategy holds up: Half Trend Strategy on NIFTY.
Final thoughts
Finding an edge is only the beginning. The real work is testing it, validating it, and managing it over time.
Discipline and process matter more than prediction.