How to test a trading strategy without fooling yourself.
Guides to forex backtesting, walk-forward validation and the specific ways a backtest flatters you — plus findings from our own archive, including a volume filter we tested, could have sold, and did not ship because it did not work.
Start here
The head questions, answered properly.
Forex Backtesting: How to Test a Strategy Without Fooling Yourself
How forex backtesting works, the five ways a backtest quietly lies to you, and the checks that separate a real edge from a curve fit. For traders who want a verdict, not a pitch.
AI Forex Backtesting: What It Actually Does, and What It Cannot Do
What AI genuinely contributes to forex backtesting — rule generation, parameter search, pattern discovery — and the specific claims to distrust when a tool says it found an edge.
Best Forex Backtesting Software in 2026: An Honest Comparison
A candid comparison of forex backtesting software — MT5 Strategy Tester, Forex Tester, QuantConnect, StrategyQuant and TradingView — including where each one genuinely beats the others.
How to Backtest a Forex Strategy Without Writing Code
A practical route from a strategy you can describe in a sentence to a graded out-of-sample result, with no programming — and an honest account of what the no-code approach costs you.
Platform comparisons
Where the alternatives are genuinely better, and where they are not.
QuantConnect vs QuantParadox: Which Fits Your Forex Research?
An honest side-by-side for forex research — coding requirements, asset coverage, out-of-sample defaults and live execution, including where QuantConnect is plainly the better platform.
LEAN, Backtrader, VectorBT or a Hosted Platform: Which Backtesting Engine?
Open-source backtesting engines like LEAN, Backtrader and VectorBT versus hosted platforms — what running your own engine actually costs in time, and when the control is worth it.
MT5 Strategy Tester: What It Does Well and Where It Misleads
An honest assessment of MetaTrader 5's Strategy Tester — modelling quality, broker data variance, the out-of-sample gap, and when a dedicated backtester is worth it.
TradingView Backtesting: What Pine Script Results Do and Don't Tell You
What TradingView's Pine Script strategy tester measures well, where its intrabar assumptions flatter results, and how to sanity-check a strategy report before trusting it.
Method
The checks that separate a real edge from a curve fit.
Walk-Forward Optimization in Forex: The Only Test That Survives Contact
What walk-forward optimization is, how to choose window sizes, and why an in-sample-only backtest tells you almost nothing about how a strategy will behave next month.
Overfitting in Forex: How Beautiful Backtests Get Manufactured
How curve-fitting produces spectacular backtests that fail live, the specific signs that reveal it, and why the number of variants you tested belongs in the verdict.
Is My Trading Edge Real, or Am I Just Lucky?
How to tell whether a run of profitable trades reflects a genuine edge or ordinary variance, how many trades it takes to know, and why most track records are far too short to say.
Spread and Slippage: The Costs That Decide Whether an Edge Exists
Why realistic cost modelling flips more backtest verdicts than any other factor, how to set spread and slippage assumptions, and what a zero-cost backtest is really telling you.
The Backtest Metrics That Lie to You
Why win rate, profit factor and Sharpe ratio routinely mislead when read without sample size and context — and the small set of numbers actually worth reading first.
How Much Historical Data Do You Need to Backtest Forex?
How much price history a forex backtest actually requires, why bar count matters more than calendar years, and the data-depth traps that silently shorten your test window.
Position Sizing Changes Your Backtest More Than Your Entry Does
Why position sizing dominates backtest outcomes, the difference between fixed-lot and risk-based sizing, and how sizing choices quietly rewrite the result of an unchanged strategy.
Backtesting for a Prop Firm Challenge: Testing the Rules, Not Just the Edge
Why passing a prop firm evaluation is a different problem from having a profitable strategy, and how to test against daily loss limits, trailing drawdown and consistency rules.
Look-Ahead Bias: How a Backtest Sees the Future Without Anyone Noticing
Look-ahead bias is a backtest using information that was not available when the trade was decided. The five ways it gets in, how to detect it, and why it flatters every result.
In-Sample vs Out-of-Sample: How to Split a Backtest So the Result Means Something
In-sample data tunes the strategy; out-of-sample data grades it. How to choose the split, what invalidates it, and why a single out-of-sample look is all you get.
How Many Trades Does a Backtest Need Before the Result Means Anything?
Thirty trades is a coin-flip sample. How trade count sets the width of the confidence band around a backtest, what the number depends on, and how to reach it honestly.
Parameter Stability: Why the Best Setting in a Backtest Is Usually the Wrong One
A strategy whose result collapses when a parameter moves one step was fitted, not found. How to read a parameter surface, what a plateau looks like and how to pick a setting that lasts.
What we measured
Findings from our own archive — including the ones that went against us.
We Tested Whether Volume Filters Improve Forex Setups. They Did Not.
A decade-scale test of whether requiring above-average volume improves forex setup performance. The result went against a feature we could have sold, so we published it.
Does Session Timing Change Forex Setup Performance? We Measured It
What a decade of minute data says about the same setup across Asian, London and New York hours — and why session is one of the few filters that survives testing.
Market structure
Sweeps, levels, flips and breaks — defined tightly enough to test.
Liquidity Sweeps: How to Tell a Real One From a Fake
What a liquidity sweep is, why most of the ones marked on charts are ordinary wicks, and the test that separates a sweep that reversed from one that simply kept going.
How to Backtest a Liquidity Sweep Strategy Properly
A step-by-step method for backtesting liquidity sweep setups, including the definition problem, the same-bar ambiguity that inflates results, and how to count your variants honestly.
Stop Hunts in Forex: What Actually Happens at Your Stop
Whether brokers hunt stops, why price so often reverses right after taking yours, and what to do about it — separating the market-structure explanation from the conspiracy one.
Do Support and Resistance Levels Actually Work?
How to test support and resistance without fooling yourself: defining levels mechanically, avoiding hindsight selection, and what actually changes when a level has been touched before.
Support Turned Resistance: Does the Flip Actually Hold?
Flip zones are one of the few structural ideas that survived our own testing. Here is what a flip is, how to define one mechanically, and what we measured about how they behave.
Order Blocks and Smart Money Concepts: What Survives a Test
Order blocks, smart money concepts and institutional narratives, assessed by what can actually be defined mechanically and measured rather than by whether the story sounds plausible.
Fair Value Gaps: How to Define One and Whether They Fill
What a fair value gap is, the three-candle definition that makes it mechanical, and how to measure fill rates honestly instead of counting only the gaps that happened to fill.
Break of Structure: Defining It So You Can Actually Test It
Break of structure and change of character explained mechanically — how to encode swing points without look-ahead, and why most structure-break backtests quietly use the future.
Breaker Blocks: When a Failed Order Block Becomes the Level, and How to Test It
A breaker block is an order block that failed, was traded through, and is then expected to hold from the other side. The definition, the parameters that decide it and how to grade it.
Inducement: The Liquidity Taken Before the Real Move, and Whether It Can Be Tested
Inducement is the minor high or low swept before price reaches the level a trader wants. What the concept claims, how to define it mechanically, and where it leaks hindsight.
Premium and Discount: Splitting a Range at Equilibrium, and What the Split Can Prove
Premium and discount divide a range at its midpoint and say buy low, sell high in structural terms. Which range, which midpoint, and whether the location of an entry changes its result.
Change of Character (CHoCH): Defining the First Sign of a Reversal So It Can Be Graded
A change of character is the first structural break against the trend in force. What separates it from a break of structure, the swing rule it rests on, and how to measure what follows.
Candles and chart patterns
What survives when the pattern stops being drawn by eye.
Doji Candles: What They Signal and What They Don't
What a doji actually is, why the textbook indecision reading is weaker than it sounds, and how to test whether doji candles change what happens next on your instrument.
Engulfing Candles: Testing the Reversal Claim Honestly
The bullish and bearish engulfing pattern, the definitional choices that change your sample size tenfold, and how to measure whether it beats the base rate on your instrument.
Pin Bars: The Rejection Candle, Measured Rather Than Assumed
How to define a pin bar mechanically, why the wick-to-body ratio you choose decides your results, and how to test rejection candles against the base rate on your own data.
Do Chart Patterns Work? Testing Triangles, Flags and Head and Shoulders
Head and shoulders, triangles, flags and wedges — why they are so hard to encode, what happens when you make the definition mechanical, and how loose detectors fool everyone.
Double Tops and Double Bottoms: Where the Definition Breaks
Double tops look unmistakable in hindsight and are surprisingly hard to detect in real time. What a strict definition requires, and what we found when we audited a loose one.
How to Backtest Candlestick Patterns Without Fooling Yourself
A method for testing any candlestick pattern: the base-rate comparison, the definitional tolerances that decide your sample, cost sensitivity, and counting the variants you tried.
AI tools
What the technology is genuinely good at, and where the marketing outruns it.
AI Forex Trading Tools: What They Do and What They Can't
A plain assessment of AI forex tools — signal generators, chat assistants, pattern recognisers and auto-optimisers — and the questions to ask before trusting any of them.
AI Chart Pattern Recognition: Where It Helps and Where It Fails
Machine chart reading is strong at describing what happened and weak at saying what happens next. Why that distinction matters, and how to evaluate a pattern recogniser.
Can ChatGPT Build a Trading Strategy? What It's Good At
Language models are excellent at turning a vague trading idea into precise rules and poor at telling you whether the idea is any good. How to use one without being misled.
Stocks and equities
The data problems that make an equity backtest flatter itself.
Stock Backtesting: How to Test an Equity Strategy Properly
What makes backtesting stocks harder than forex: survivorship bias, corporate actions, gaps, borrow costs and liquidity limits — and how to handle each one honestly.
Survivorship Bias: Why Your Stock Backtest Is Too Good
Testing a stock strategy on today's list of companies excludes every one that failed. What that does to your results, how to detect it, and what a point-in-time universe fixes.
Splits, Dividends and Why Your Stock Backtest Is Wrong
Corporate actions rewrite historical prices. How split and dividend adjustment work, the two defensible dividend conventions, and how to verify your data before trusting a result.
Metrics and risk
The numbers that decide position size, and the ones that only look like they do.
What Is a Good Sharpe Ratio for a Trading Strategy?
What Sharpe actually measures, the numbers worth taking seriously at retail scale, and why a backtested Sharpe above 3 is usually a sign of a search rather than an edge.
Maximum Drawdown: How Much Is Too Much?
What max drawdown measures, why the backtested figure is almost always smaller than the one you live through, and how to size a strategy around the drawdown you can tolerate.
Expectancy and R-Multiples: The Only Edge Number That Travels
Why expectancy in R is the one performance figure that compares across instruments, timeframes and account sizes — and how to compute it without flattering yourself.
Win Rate vs Profit Factor: Which One Should You Trust?
Win rate is the most marketed metric in trading and the least informative. What profit factor adds, where it also misleads, and the pair of numbers that actually describes a strategy.
Is a 1:3 Risk-Reward Ratio Actually Better?
The most repeated rule in retail trading, tested properly: what happens to win rate when you widen the target, and why the ratio is an output rather than a setting.
Risk of Ruin: The Number That Decides Position Size
How to calculate the probability that a positive-expectancy strategy still blows up an account, and why the answer depends far more on position size than on the edge.
Deflated Sharpe Ratio: Correcting a Backtest for Every Strategy You Didn't Keep
The best of many backtests has a Sharpe ratio inflated by selection. What the deflated Sharpe ratio corrects, what it needs as inputs, and why the number of trials matters.
Proving it works
Getting from a promising backtest to something you would trade.
Why Your Backtest Doesn't Match Live Trading
The six reasons a strategy that tested well trades badly, ranked by how often each one is the real cause — and how to tell which is yours before changing anything.
Monte Carlo for Trading Strategies: What It Can and Can't Tell You
Resampling a trade sequence turns one historical result into a distribution. What that genuinely shows, the independence assumption that breaks it, and how to run it honestly.
How Long Should You Forward Test Before Going Live?
Forward testing is measured in trades, not weeks. How to work out the number your strategy actually needs, and what paper trading can and cannot confirm.
How to Evaluate Someone Else's Backtest Before Buying
The seven questions that separate a real result from a sales asset, in the order that eliminates the most claims fastest — and what an honest vendor answers without hesitating.
Multi-Timeframe Analysis: The Backtest Leak Nobody Checks
Using a higher timeframe to filter a lower one is standard practice and the most common source of look-ahead bias in retail backtesting. How the leak works and how to close it.
When a Strategy Stops Working: Telling Edge Decay From a Normal Losing Streak
Every strategy has losing runs; some have stopped working. How to set the line in advance, which signals separate decay from variance, and what to do when it is crossed.
Traps that flatter a backtest
Four setups that produce beautiful results and untradeable strategies.
Why Martingale and Grid Strategies Backtest Beautifully
Grid and martingale systems produce the smoothest equity curves in retail trading, for a structural reason that has nothing to do with edge. How to see it in a backtest.
Renko and Heikin Ashi: Why Their Backtests Are Not Real
Smoothed and price-derived charts produce the cleanest-looking backtests in retail trading, and the results are usually untradeable. What the transformation actually does to a test.
Can Scalping Strategies Be Backtested Honestly?
At a 5-pip target a 1-pip spread is 20% of the edge. What scalping backtests need that longer-horizon ones do not, and the assumptions that decide the answer.
Which Timeframe Should You Trade? Test It, Don't Pick It
Timeframe changes costs, sample size, noise and how often you have to be right. How to test the same idea across timeframes without fooling yourself about which one won.
By market
What changes when the instrument is not a currency pair.
Crypto Backtesting: What Changes When the Market Never Closes
Funding rates, exchange-specific prices, 24/7 sessions and tokens that stop existing. The differences that make a forex backtesting setup wrong for crypto.
Backtesting Gold (XAUUSD): The Instrument That Breaks Settings
Gold's volatility, spread behaviour and session structure break parameters tuned on currency pairs. What to change before testing a strategy on XAUUSD.
Backtesting Indices: NAS100, US30 and the Overnight Gap
Index CFDs gap, close, roll and charge financing. The four mechanics that make an index backtest diverge from a forex one, and how to model each honestly.
From another platform's code
Porting a strategy out of a charting tool without inheriting its assumptions.
Testing an Expert Advisor Properly in MetaTrader
Modelling modes, tick data quality, spread settings and the optimiser. The MT5 Strategy Tester settings that decide whether an EA result means anything.
From Pine Script to a Strategy You Can Actually Test
TradingView is excellent for developing an idea and limited for validating one. What survives the move to a real backtest, and the four Pine behaviours that flatter results.
Practice and simulation
What replay and demo accounts teach, and the question they structurally cannot answer.
Trading Simulators: What They Prove and What They Don't
What a trading simulator genuinely teaches, the four ways manual replay flatters a result, and where clicking through history stops being evidence about a strategy.
Demo Accounts vs Trading Simulators: Which One Do You Need?
Demo accounts run forward in real time, simulators replay the past on demand. What each one models honestly, what each one quietly gets wrong, and when to use which.
Bar Replay vs Backtesting: Two Tools, Two Different Questions
Bar replay answers whether you can execute a setup; backtesting answers whether the setup has an edge. Why swapping one for the other quietly wastes months.
How to Practise Trading Without Risking Money (and Without Wasting Time)
A practice sequence that produces something checkable: written rules, a replay log, a mechanical grade, and a small live position. What to skip and what to measure.
Crypto Trading Simulators: What Changes When the Market Never Closes
Round-the-clock sessions, exchange-specific data, funding costs and thin history make crypto practice different. What a crypto simulator models and what it silently omits.
Stock Market Simulators: Where the Practice Transfers and Where It Doesn't
Equity simulators face problems currency ones do not — splits, dividends, delistings and gaps. What a stock simulator has to handle before its history means anything.
Journals and research records
Logging trades and tests so the record can answer something later.
Trading Journals: What to Record So the Log Can Answer Something
Most trading journals record outcomes and cannot answer any useful question. What to capture instead so the log can separate a bad strategy from bad execution.
A Trading Journal Template That Can Actually Be Queried
A field-by-field trading journal template, why each column exists, and the three fields that decide whether the log can ever separate strategy from execution.
Keeping a Backtesting Journal: The Log That Stops You Fooling Yourself
Recording every backtest you run, including the failed ones, is what turns a series of tests into evidence. What to log and why the discarded runs matter most.
Named models
Popular frameworks, defined tightly enough that they could fail a test.
Power of Three (AMD): Turning Accumulation, Manipulation and Distribution Into a Testable Rule
The accumulation, manipulation and distribution model described precisely enough to test, the four decisions it leaves unstated, and how to grade it without hindsight.
The Judas Swing: Defining the False Move Before You Know It Was False
The Judas Swing describes an early session move that reverses. What separates a testable definition from a retrospective label, and how to grade it without hindsight.
Turtle Soup: The False Breakout Rule and How to Grade It Honestly
Turtle Soup fades a breakout of a recent extreme that fails. The definition, the lookback and confirmation parameters, and why the failure cases decide the result.
Quarterly Theory: Splitting Time Into Quarters and Testing Whether It Matters
Quarterly Theory divides every time period into four phases with an expected role for each. How to state that as a testable claim and what a fair test needs.
The Market Maker Model (MMXM): What It Claims and What Can Be Measured
The market maker buy and sell model describes a repeating multi-stage cycle. Which parts are geometry that can be tested and which are narrative that cannot.
Indicators
What each one measures, and the control test that says whether it earns its place.
VWAP: What It Measures, and Why Forex VWAP Is Not the Same Thing
VWAP is a volume-weighted average price, which makes it only as meaningful as the volume behind it. Why that matters far more in forex than on an exchange.
Volume Profile: Reading Distribution by Price, and What It Needs to Be Valid
Volume profile shows activity distributed by price rather than time. What the point of control and value area actually mean, and why the forex version rests on tick data.
Moving Average Crossovers: The Most Backtested Rule in Trading
The crossover is simple enough to test exactly, which makes it the best available lesson in parameter search, regime dependence and why simple results decay.
Fibonacci Retracements: Making the Levels Testable Instead of Decorative
Fibonacci levels depend entirely on which swing you anchor them to. How to make that choice mechanical, and what a fair test of the levels requires.
RSI Divergence: Making the Most-Cited Reversal Signal Mechanical Enough to Grade
RSI divergence is price at a new extreme while the oscillator is not. The swing rule, lookback and confirmation that make it a countable event, and the baseline it must beat.
MACD Strategies: Signal-Line Crosses, Histogram Turns and the Zero Line, Graded Separately
MACD gives three distinct signals that are usually tested as one. What each one measures, why they should be graded separately, and the settings question that decides the result.
Bollinger Bands: Mean Reversion or Breakout? The Two Rules a Backtest Must Keep Apart
Bollinger Bands support two opposite strategies, fading the band and following the break. What the bands measure, why the two rules cannot both be right at once, and how to grade each.
Setups and entries
Hand-drawn and clock-anchored setups, made mechanical enough to grade.
Opening Range Breakout: The Parameters That Decide Whether It Works
The opening range breakout has four parameters hiding inside a simple description. How each one changes the result, and what a fair test of the setup requires.
Trendline Trading: Making a Hand-Drawn Line Into Something You Can Test
Trendlines are drawn by eye, which makes them the hardest common technique to test. The algorithmic definitions that make it possible and what they cost.
Swing Highs and Lows: The Definition Everything Else Depends On
Swing points sit underneath market structure, trendlines, Fibonacci levels and stop placement. Why the definition is a parameter and when a swing is actually confirmed.
Swing Trading: What Changes When Positions Are Held for Days
Swing trading holds positions across days, which introduces gaps, overnight financing and a trade count low enough to make validation genuinely hard.
Kill Zones: The Session Windows That Are Supposed to Carry the Move, and How to Test Them
Kill zones are the hours around the London and New York opens where setups are said to work best. The windows, why a clock rule is easy to test, and what a fair test controls for.
London Breakout: Trading the Asian Range at the Open, With the Parameters That Decide It
The London breakout buys or sells a break of the Asian session range when Europe opens. The range definition, the break rule, the stop and the time limit, and the false-break problem.
Exits and trade management
Stops, targets and the management rules that feel safe — graded on the same entries.
ATR Stop Loss: Sizing the Stop to Volatility, and the Multiple That Decides the Result
An ATR stop scales the stop to recent volatility so a rule means the same thing in quiet and active markets. What the multiple changes, how it meets the target, and how to test it.
Trailing Stop vs Fixed Target: Which Exit Wins Depends on a Distribution, Not a Preference
A trailing stop rides long moves and gives part back; a fixed target takes less and keeps it. How each exit changes the same entries, and the intrabar flaw that flatters trails.
Moving the Stop to Break-Even: The Rule That Feels Safe and Usually Costs Expectancy
Moving a stop to break-even turns some losers into scratches and some winners into scratches too. Which happens more is a measured question, decided by what price does after entry.
Data and fidelity
What the bars hide, what the clock gets wrong, and what a test can honestly know.
Tick Data vs Minute Data: What Resolution a Backtest Needs, and What It Can Never Know
Bars hide the order in which the high and low were hit, and that order decides trades whose stop and target share a bar. When minutes are enough and when only ticks are.
Timezones and Daylight Saving: The Backtest Error That Shifts Every Session by an Hour
Broker time, exchange time and UTC disagree for weeks each year. How that corrupts session and daily-bar rules, how to find it in your data, and how to define time so it stays put.
Head-to-head comparisons
Seven full comparisons, each checked against the other product's own published documentation, and each with a section naming the cases where they are the better tool.
Reading about backtesting is not backtesting.
Describe a strategy in a sentence, or paste your own code, and get the out-of-sample verdict on ten years of minute data. One full backtest free, no card.