If one cTrader broker gives you a clean backtest and another gives you a worse one, that is not automatically proof one of them is lying. Most of the time, the "same strategy" was not run in the same experiment at all.
The result changed because the inputs changed. cTrader itself treats commission, spread, and the historical data source as part of the backtest setup, not background wallpaper (cTrader backtest guide).
The same code does not mean the same test
cTrader backtesting runs a bot on historical market data under conditions you define. In the platform's own settings, that includes the commission, the data source, and whether spreads are fixed or random; it also allows custom CSV data instead of the broker's server history (cTrader backtest guide).
That matters because brokers do not all expose the same market conditions. cTrader's own trading glossary states that trading sessions can vary between brokers because of location, liquidity-provider hours, and local holidays (cTrader trading glossary). Spotware's custom-symbol guide makes the same point from the implementation side: swap rates, commissions, and market hours are all part of the symbol definition itself (cTrader custom symbols).
So when traders say "I ran the same cBot on two brokers," that sentence is usually incomplete. The logic may be the same. The test conditions often are not.
| Input that changed | What actually moved | Why the result can diverge |
|---|---|---|
| Historical data | The bars or ticks the bot consumed | A signal can appear on one data set and not on another |
| Spread | The entry and exit friction | Tight edges can disappear once the spread widens |
| Commission and swap | The cost stack around each trade | Swing systems can drift meaningfully when overnight carry changes |
| Trading sessions | When the market is considered open | Higher-timeframe bars can close at different moments |
| Starting capital and sizing | Position size path | Percentage-risk systems compound differently from the first trade onward |
Tiny broker differences can create different trade lists
Some differences only shave the headline return. Others change the actual trades.
That second case surprises people because it feels unfair. The code did not change, so how can the trade list change? Because many strategies do not trade raw ticks. They trade candles, filters, thresholds, and risk sizing rules built on top of those candles. If one broker's session cuts the day differently, or one symbol has slightly different historical highs and lows, the filter can flip from "yes" to "no" on the margin.
This is especially true on higher timeframes. One H4 candle is not always the same H4 candle across every implementation. realbacktesting documents this openly in its own methodology stack: the cTrader-native report is the figure a buyer reproduces in cTrader, while the independent research engine can land a few points away because higher-timeframe bars are built on different boundaries from the same M1 base data. That is not a contradiction. It is exactly the sort of implementation detail an honest backtest should disclose instead of hiding behind a screenshot.
Costs work the same way. A wider spread does not merely reduce profit in a neat straight line. It can stop entries from filling, pull exits forward, and turn a marginal winner into a loser. That is the mechanical reason why your backtest lies was worth writing in the first place.
Why prop traders should care more than everyone else
An investor looking at a multi-year portfolio can sometimes tolerate messy implementation noise. A prop trader usually cannot.
Prop-firm rules judge the path, not just the destination. A slightly harsher broker feed or cost stack can cluster losses differently, deepen an intraday dip, or delay a recovery long enough to matter under daily-loss or max-loss rules. That is why daily loss limit vs max loss for prop traders is a survival topic, not admin trivia.
This is also why the harshest plausible backtest is usually more useful than the prettiest one. If a strategy only looks acceptable on the friendliest broker conditions, that is not robustness. That is dependence on a flattering setup.
The funding side of the problem is the same. A prop account gets paid only if the return path survives long enough to reach the payout path. That is why funding.html focuses on the route to funded status and the return distribution, not just one headline curve.
How to compare cTrader backtests honestly
If you want to compare two backtests without kidding yourself, be stricter than the marketing pages are.
- Match the symbol, broker, and date range first. If those differ, you are already comparing different tests.
- Match the cTrader backtest settings. Data mode, commission, and spread assumptions belong in the comparison table, not in the footnotes.
- Match the starting capital and sizing logic. On a percentage-risk system, a small difference early can compound into a visibly different curve later.
- Compare the trade list and the daily path, not just the final ROI. A prop trader fails on path stress long before a long-run average can save them.
- Re-run the test on the harsher plausible conditions. If the edge only survives on the softest version of reality, it was probably not much of an edge.
That is also the right standard for published research. If the author cannot tell you which data they used, what the cost model was, and whether you can reproduce it in your own platform, you do not yet have a number worth trusting.
What realbacktesting proves, and what it does not
realbacktesting publishes verifiable, prop-firm-ready cTrader trading systems, validated on five years of real broker data with full out-of-sample and Monte Carlo testing. The useful part is not the slogan. It is the audit trail.
The current published method is explicit: 2021-06-01 to 2026-06-20 broker M1 data, an 80,000 EUR model base, real per-symbol spread, commission, swap, and 1 bps slippage, 100% signal parity across 13 strategies and 175,401 bars, and drawdown ceilings enforced at the 95th percentile of 20,000 Monte Carlo simulations, then confirmed on a 30% out-of-sample hold-out. Those figures live on methodology.html, where the whole point is that a skeptical trader can inspect the method instead of trusting a polished story.
The honest limit matters too. These are still backtests and models on 2021-2026 data, not a live track record. A broker difference does not automatically invalidate a backtest, but it does mean you should ask harder questions about what exactly was tested.
Frequently asked
Does a different broker result mean the strategy is fake?
No. It usually means the data, costs, or session structure changed. The right question is not "are the two curves identical?" but "are the assumptions documented, reproducible, and still acceptable under tougher conditions?"
Should I trust the best-looking broker result?
Not by default. The best-looking broker result is often just the kindest version of the test. For prop trading, the harsher plausible result is usually the more decision-useful one.
Can the same strategy get different trades even if the costs stay close?
Yes. If candles are built on different session boundaries, or if small differences in historical highs and lows push a filter across a threshold, the trade list itself can change.
What is the minimum standard before I trust a published backtest?
You should be able to identify the broker or data source, the date range, the spread and commission assumptions, the sizing basis, and whether the result is reproducible in your own platform. If any of that is missing, the headline number is weaker than it looks.
The stubborn takeaway
When the broker changes, the backtest is not just being "rechecked." It is being rerun under a different set of market facts.
That is why honest traders care less about the prettiest curve and more about whether the method survives when reality gets less cooperative.