Backtestify
Sign in
All articles
Verify Youtube StrategyNovember 11, 202613 min read

How to Verify a YouTube Trading Strategy Before You Risk Money

A step-by-step way to verify a YouTube trading strategy: turn it into rules, backtest with real costs, read the metrics, and check it out of sample.

How to Verify a YouTube Trading Strategy Before You Risk Money

To verify a YouTube trading strategy, turn the video into written rules, run those rules on real historical data with realistic costs, read the results through drawdown and profit factor rather than win rate alone, and then check the strategy on data it has never seen. If it survives all four, it deserves a small live test. If it fails any of them, the video was a good story and nothing more.

This guide walks through each step in order, with the numbers to look for and the mistakes that make weak strategies look strong.

TL;DR

  • Write the strategy as exact rules (market, timeframe, entry, exit, stop, size) before you test anything. If the creator cannot answer those questions, there is nothing to test.
  • Include commissions, spread and slippage in every backtest. Zero-cost tests overstate results.
  • Judge a strategy by maximum drawdown, profit factor and risk-adjusted return together. A high win rate alone says very little.
  • Hold back part of the data, or use walk-forward testing, so you know the rules were not just fitted to the past.
  • A very good result is a reason for more scrutiny, not for celebration.

Table of Contents

Why a YouTube video is not evidence

A video shows you the trades that worked. It rarely shows every signal over several years, the losing streaks, or the costs. Quant researcher Rick Hardy made this point when he began testing popular video strategies: viewers are shown examples where the conditions line up and price moves as predicted, and then a test of around 100 trades is declared a staggering result.

When he ran those ideas through a full backtest, the numbers looked very different. A MACD plus 200-period moving average setup on hourly Bitcoin over one year (October 2021 to October 2022) lost 9.2% with a 1:2 risk-reward ratio, with a maximum drawdown of 23.4%. With a 1:1 ratio it lost 4.4%. A second video strategy, using Chandelier Exit and ZLSMA on 5-minute Bitcoin candles, lost 7.2% over its test year with an 11.3% maximum drawdown.

Comparison of backtested Bitcoin video strategies

That is two strategies, one asset and one year, so it proves nothing about all YouTube strategies. It does show why you test before you trade. Some creator strategies hold up and some do not, and you cannot tell which from the thumbnail.

Step 1: Turn the video into rules a computer could follow

The first job is removing ambiguity. Phrases like "wait for a strong candle" or "enter when momentum looks good" cannot be tested. Write down:

  1. Market and timeframe: for example, ES futures on the 15-minute chart, or EUR/USD on the 1-hour chart.
  2. Entry condition: the exact signal, including which candle you act on and whether you enter at its close or the next open.
  3. Exit condition: the stop loss, the take profit or the trailing rule, and any time-based exit.
  4. Position size: a fixed number of shares or contracts, or a fixed percentage of the account risked per trade.
  5. Trade filter: whether you take every signal, or only the first of the day, or only in certain sessions.

If you cannot fill in one of these from the video, check the description, pinned comments and the creator's other videos. If the creator never states it, you have found the first red flag. A strategy that depends on unstated discretion cannot be verified, only believed.

Some people extract rules from transcripts with scripts. One open-source example, yt-trade-distill on GitHub, pulls transcripts, extracts structured rules from each video and attaches an evidence quote and confidence score to every rule. Whatever you use, read the final rule list yourself and compare it to the video. Automated extraction can misread what was said.

In Backtestify you can skip the translation step to code: you describe the strategy in plain English and we turn it into testable rules. You still need the answers to the five questions above, because the plain-English description is only as precise as you make it.

Step 2: Use the right data and add real trading costs

A backtest is only as good as the data under it. Clean historical data free of errors matters because bad data can make the analysis misleading. Match the data to the market the video trades: a strategy shown on futures should not be tested on a stock proxy, and a 5-minute strategy needs intraday data, not daily bars.

Then deal with costs. Your test needs an explicit execution model covering commissions, bid-ask spread, slippage and fill timing. Otherwise the tester can assume zero commission and perfect timing, which overstates performance compared with real trading. The same source argues that most overly optimistic backtests come from this kind of execution mismodeling rather than from the strategy logic itself.

This matters most for short-term strategies. A scalping setup that makes many trades a day pays the spread and commission on every one. Run the test twice, once with costs set to zero and once with costs you would really pay at your broker. If the edge vanishes when modest costs are added, the strategy was never robust.

Check how long the test period is, too. A strategy tested on one calm year says little about a volatile one. Cover trending periods, choppy periods and at least one sharp sell-off. Our guide on how much historical data a backtest needs covers why a long record can still mislead when you try many variations.

Step 3: Read the metrics together

No single number tells you a strategy is good. Look at these side by side.

MetricWhat it showsRough reading
Maximum drawdownLargest peak-to-trough declineCan you sit through it without quitting?
Profit factorGross profit divided by gross lossAbove 1.5 is generally good, above 2.5 excellent
Sharpe ratioReturn per unit of volatilityBelow 0.5 weak, around 1.0 solid, around 1.5 strong
Win rate with reward-to-riskHow often it wins and how much it wins versus losesMust be read as a pair

The Sharpe ranges above come from a quantitative vetting guide, which adds that a Sharpe above 2.0 deserves forensic examination rather than excitement.

Why win rate alone misleads

A strategy that wins 80% of the time can still lose money if the average loser is five times the average winner. A strategy that wins 35% of the time can be profitable if winners are three times losers. Always read win rate next to average win and average loss.

Drawdown decides whether you can actually trade it

Drawdown is where recovery math hurts. A 40% maximum drawdown needs a 67% gain just to get back to even. Ask what drawdown you would tolerate in a real account, then compare it to the backtest. The same vetting guide notes that live drawdown can exceed the backtest maximum by a factor of two or more, and suggests stress-testing the historical maximum drawdown with at least a two-times multiplier before you size a position.

Be suspicious of perfect results

A profit factor of 1.4 can be more credible than one of 3.8 if the higher number came from tightly tuned entry filters and unrealistically clean fills. Once a profit factor passes about 3.0, curve-fitting and look-ahead bias become plausible explanations, according to that same guide. Our four quick tests for look-ahead bias show how to check whether a result is leaking future information.

Step 4: Hunt for the biases that flatter a strategy

A good-looking equity curve can come from the test, not the strategy. The recurring failure modes are look-ahead bias, survivorship bias, data snooping, unrealistic execution, cost underestimation and regime fragility. One machine-learning-for-trading textbook goes as far as reframing backtesting as falsification: the default stance toward any result is disbelief until it survives repeated attempts to break it.

In practice for a YouTube strategy:

  • Look-ahead bias: the rules quietly use information that was not available at the time of the trade, such as a candle's close when the entry happens mid-candle.
  • Survivorship bias: if the video trades a list of stocks, test the full historical universe including delisted names. Using current constituents only can make results look too optimistic.
  • Data snooping and overfitting: if you tweak the creator's settings until the curve looks great, you have found the best fit to the past. Research on the probability of backtest overfitting shows that the winning configuration among many tried is likely to overstate future performance. Read more in our piece on data snooping bias.
  • Regime fragility: the strategy worked in a trend that no longer exists. Look at results by year, not only the total.

The cleanest defense is discipline: decide your rules and your test period first, then run the test once. Every extra variation you try raises the odds that a good result is luck.

Step 5: Test on data the rules have never seen

Once a strategy passes on the full history, split the data. Tune on an earlier period, then test the fixed rules on a later one. If performance collapses out of sample, the strategy is overfit. Do not adjust and rerun on the same validation set, because that leaks information back into the rules.

Two practical checks:

  1. Hold-out split: reserve a later block of data, for example the most recent two or three years, and touch it only once.
  2. Parameter nudges: change each setting a little, say a moving average from 50 to 45 and 55. If results fall apart, the edge was a coincidence of exact numbers. A real edge usually degrades gently.

For a more rigorous version, use walk-forward analysis, where you optimize on one window, test on the next and roll forward. It is widely described as the gold standard for robustness testing. Our guide to walk-forward analysis explains how to set the windows.

After that comes forward testing: paper trade or trade very small size in real time and compare fills and results with the backtest. Our forward testing checklist lists five checks that keep those results auditable. Backtest results are not guarantees, since live trading adds slippage, liquidity limits and execution delays. Start with a size where the worst case, at twice the backtest drawdown, is an amount you can afford to lose.

Run the verification in Backtestify

This whole process is what we built Backtestify for. You enter a strategy in plain English, we test it on real historical market data, and you see every trade plus win rate, profit factor and maximum drawdown. You can then improve the rules and compare the original and improved versions side by side on recent data, forecast likely price movement with probability and confidence statistics, and export to TradingView Pine Script.

We also publish reports for strategies that win and for ones that lose, so you can see how a popular setup actually behaved. The Tori Trades trendline strategy report is one example of a creator strategy tested this way.

Backtestify

There is a free plan to start. Pro costs $29 per month or $190 per year and adds unlimited backtesting, improvement and forecasting. You can run your first backtest now, or see the full plan details first.

How much evidence is enough?

No test proves a strategy will work. What you can do is set a bar in advance and see whether the strategy clears it. A reasonable bar before risking real money:

  • The rules were complete and unambiguous, with no discretion needed.
  • The backtest includes realistic costs, and the edge survives them.
  • The sample is large enough, with many trades across different market conditions, not a few dozen.
  • Profit factor and Sharpe are decent but not suspiciously perfect, and drawdown fits your tolerance even at double.
  • Out-of-sample or walk-forward results resemble the in-sample results.
  • Small changes to parameters do not break it.

A strategy that clears all six merits a small, time-limited live test. One that clears most but not all may merit a rule improvement and a fresh, untouched test period. One that fails on costs or out of sample should be dropped, however convincing the video was. For the wider method, our guide on evidence-based trading and the step-by-step backtesting guide go deeper.

FAQ

Can I verify a YouTube strategy without coding?

Yes. You need the rules written precisely, but not code. A platform like Backtestify accepts plain-English strategy descriptions and runs them on historical data. The hard part is the rule list, not the programming.

What if the creator will not share exact rules?

Treat that as a warning. If entries, exits and stops cannot be written down, results cannot be reproduced, and a strategy that cannot be reproduced cannot be verified. You can test your best written interpretation, but a good result then applies to your version, not the creator's.

How many trades do I need for a reliable backtest?

There is no universal number, but a handful of trades proves little, and a test of around 100 is the kind of sample that produced the glowing claims described above. Aim for many trades across several market conditions, and check results year by year.

Does a good backtest mean I will make money live?

No. Live trading adds slippage, liquidity limits and delays, and live drawdowns can run well beyond the backtest. Use the backtest to filter out weak strategies, then start small and compare live results with the test.

Sources

Recommended