Backtest έναντι πραγματικών αποτελεσμάτων: πώς ξεχωρίζετε ένα αληθινό ιστορικό από μια προσαρμοσμένη καμπύλη

If you have spent any time looking at trading products in Europe, you have seen the same image: a smooth equity curve climbing left to right, a large percentage in bold, and the words "backtested results" in grey type underneath. Understanding the gap between backtest vs live results is the single most useful piece of due diligence a retail reader can do, and it takes about fifteen minutes once you know what to look at.

What is the difference between a backtest and live results?

A backtest is a simulation of how a set of rules would have performed on historical data. Live results are a record of orders that were actually sent to a market and filled at prices someone actually received. The first is a hypothesis. The second is evidence.

That distinction sounds obvious, and it is routinely ignored. A backtest is produced by the same person selling the product, on data they chose, with rules they were free to adjust until the picture looked good. Nothing about that process is dishonest by itself. It only becomes dishonest when the output is presented as a performance record rather than as a research artefact.

The honest framing is uncomfortable: a backtest tells you what a strategy could not have failed at. It says nothing about the future, and very little about whether the strategy would have survived execution.

Why do backtests look so much better than live trading?

Because a backtest silently removes almost every cost and constraint that a live account cannot escape. The gap is not usually caused by fraud. It is caused by a long list of small omissions that all point the same direction.

Put concrete numbers on that. Take a strategy that backtests at 42% annually over four years, trading 500 round-turns a year on gold at 0.10 lots.

• Commission at $7 per round turn per lot: 500 × 0.10 × $7 = $350
• Extra spread of $0.15 per trade beyond the modelled figure: 500 × $0.15 × $10 = $750
• Slippage of $0.20 on the roughly 200 stop-outs: 200 × $0.20 × $10 = $400

That is $1,500 a year of friction the backtest never charged. On a $10,000 account, a 42% backtest becomes roughly 27% live before a single thing goes wrong with the strategy itself. And that is the optimistic version, because it assumes the rules were not fitted to the data in the first place.

How can you spot a curve-fitted backtest?

Look for the fingerprints of a strategy that was tuned until it matched history rather than tested against it. Curve fitting leaves marks, and they are easy to see once you know them.

What does a real live track record have to show?

At minimum: third-party verification, a full trade history rather than a summary, a stated account currency and leverage, and a period long enough to include at least one bad stretch. If any of those four are missing, you do not yet have a track record, you have a claim.

Verification matters because self-reported figures are unfalsifiable. A platform like Myfxbook connects to the trading account directly and publishes the equity curve including every losing trade, which is why the presence of verification and the presence of losses are both good signs. The mechanics of reading one properly are covered in how to read a Myfxbook track record.

You also want to see the risk numbers, not only the return. A 60% annual return with a 55% maximum drawdown is a worse product than a 20% return with an 8% drawdown, and most marketing is built to stop you noticing. Drawdown and Sharpe ratio for European readers walks through why, and what drawdown really measures covers the number itself.

Here is the comparison in a form you can apply directly:

Does European regulation protect you here?

Partially, and less than most readers assume. EU and UK rules do impose real constraints: retail leverage on gold is capped at 20:1 under ESMA-derived rules, negative balance protection applies to retail accounts, and firms must publish the percentage of retail accounts that lose money, which typically sits between 70% and 80%.

What regulation does not do is validate anyone's performance claims. A regulated broker is regulated for how it handles your money and executes your orders. It is not certifying that a signal provider, an EA vendor or a strategy marketer is telling the truth about returns. Those are frequently different entities in different jurisdictions, and the regulated logo at the bottom of the page often belongs to the broker rather than to the party making the claim.

The practical step is to check who is actually regulated and for what activity. The broader framework for that check is in due diligence before trusting a trading provider.

What should you actually do before trusting a performance number?

Work through it in this order, and stop at the first failure.

1. Establish whether it is live or simulated. If the page does not say plainly, assume simulated.
2. Find the verification link. No independent link means the number is a marketing claim.
3. Look for the worst period, not the best. Ask for the maximum drawdown, the date it happened, and how long recovery took.
4. Check the sample. Fewer than a couple of hundred trades or less than two years is not enough to distinguish skill from luck.
5. Add the costs back in. Assume the published figure omits at least commission and slippage.
6. Ask what happens when it stops working. Every strategy eventually does. A provider without an answer has not thought about it.

And the genuinely unflattering conclusion, which almost nobody selling a trading product will write down: for most people, most of the time, a low-cost diversified index fund beats the realistic after-cost outcome of an actively marketed trading strategy. If someone is asking you to accept higher risk, higher cost and more complexity, the burden of proof sits entirely with them, and a backtest does not discharge it.

What is the TIC angle?

TIC publishes performance through independently verified Myfxbook records rather than backtests, and the record includes the losing months, because a curve without them is not a record. That is the standard we think readers should hold everyone to, including us, and you can inspect it at /results.

The wider point is that learning to evaluate a track record is more valuable than any single track record. Once you can tell a fitted curve from a verified one, the marketing stops working on you, which is worth considerably more than whatever the marketing was selling.