What a backtest tells you
A backtest applies a strategy to historical data and records simulated trades or portfolio results. Its usefulness depends on the strategy version, data, instruments, date range, execution assumptions and measurement method.
It helps answer, “How would this strategy have behaved in that period, under these assumptions?” It cannot tell you what next month will bring. If many strategies were tried and only the best result was selected, some apparent success may come from fitting that particular history too closely.
CoinScreener attaches published historical evidence to its strategy products. Read the corresponding version and assumptions rather than relying on an old screenshot or a number quoted without a date.
Follow an equity curve
The lesson begins with 100 hypothetical units and no deposits or withdrawals. Compare the short sample with the complete example, which includes a substantial drawdown before recovery.
An equity curve plots the value of a simulated or live portfolio through time. Check whether its value includes open-position gains and losses or only closed trades; the choice changes what the curve can show.
A new high
Equity has risen from 100 to 112. Total return is 12% so far; that does not describe the path still ahead.
Illustrated walkthrough · all scenarios
Full example
- A new high. Equity has risen from 100 to 112. Total return is 12% so far; that does not describe the path still ahead.
- Drawdown. Equity falls from a peak of 112 to 90: a 19.6% drawdown. Drawdown is measured from the running peak, not the initial account value.
- Recovery takes time. Equity reaches 113 and finally exceeds the earlier peak. A profitable endpoint can conceal a difficult recovery.
- Read the whole record. The example ends at 120: a 20% total return alongside a 19.6% maximum drawdown. These hypothetical observations show the whole path, not just the profitable endpoint.
Short sample
- A small window. Equity starts at 100 hypothetical units, with no deposits or withdrawals.
- A promising start. This short stretch looks promising, but it has not shown how the portfolio behaves through a substantial decline or a long recovery.
- The short sample ends. The gain is 12%. This window contains no major downturn. Switch to the full example to see what the short window missed.
This is a teaching curve, not a Polaris result. Its observations are not trades or calendar years. Do not annualise it or infer trade statistics that the example does not contain.
Return and drawdown answer different questions
Total return compares ending equity with starting equity, assuming a consistent treatment of cash flows and costs. In the full lesson, 100 becomes 120: a 20% total return.
Drawdown measures a decline from an earlier equity peak. The same curve rises to 112 and then falls to 90. That decline is (112 − 90) ÷ 112 ≈ 19.6%, even though the ending result is positive. Maximum drawdown is the largest such decline in the observed series.
Recovery time also matters. A strategy can finish profitably after a long period below its previous peak. Decide whether the path, exposure and uncertainty are acceptable, not just whether the endpoint looks impressive.
Sample size and market coverage
A handful of winning trades provides limited evidence. A larger sample can be more informative, but most trades might come from the same strong rally or quiet trading range. Hundreds of trades in similar conditions still leave other market conditions untested.
Read the date span, markets covered, number of trades and distribution by market. A portfolio result can conceal a weak component or depend heavily on one unusually profitable period. A shorter-history asset cannot provide evidence from a period before it existed.
A missing statistic does not mean zero. The product may withhold a value when evidence is insufficient. Avoid filling a blank with an assumption of good or bad performance.
Win rate, payoff and costs
Win rate is the percentage of measured trades that were winners, according to the stated definition. It does not tell you how large the wins were. For example, eight gains of 1 and two losses of 5 give an 80% win rate but a net loss of 2 before costs.
Read average outcomes, payoff distribution and costs where available. Fees, spreads, slippage and holding costs can change the result. A simulation with optimistic execution assumptions may overstate what was executable.
Check how the model handles a candle that crosses both a stop and a target. OHLC data alone does not reveal the intrabar order. Finer data or an explicit assumption is needed. A polished replay does not eliminate that ambiguity.
Historical versus live results
Historical results simulate a strategy over past data. Live results follow it after it begins operating. Live does not necessarily mean exchange-executed: a live strategy record can track reference trades without placing orders in your account. Check how entries, exits and costs are measured before treating either record as an achievable account return.
When ranking results, look for comparable market coverage, dates, costs and sizing, and identify the strategy version behind each record. When comparing a strategy's backtest with its later live period, acknowledge that they cover different market conditions. Keep open-position gains and losses separate from completed trade results.
A difference between live and historical performance is a reason to investigate the assumptions and market conditions. It is not automatically proof that either record is wrong, nor a reason to promise a return to the historical average.
How to assess a strategy's evidence
- Identify the published strategy and version.
- Read dates, markets and sample size before headline performance.
- Inspect return, drawdown and recovery together.
- Review the trade list and per-market results where available.
- Check costs, size assumptions and treatment of ambiguous bars.
- Compare the live record using compatible measurements.
- Decide what the evidence cannot answer and whether you need more observation.
The strategy's evaluation interval tells you how often it checks for setups. It does not tell you how long trades last or how often new Signals appear. Check the distribution of trades over time: a monthly average can hide long quiet periods and bursts of activity.
Common mistakes and reasons to wait
Avoid choosing a strategy from return alone, cherry-picking a favourable period, treating a backtest as an account statement or assuming a small live sample proves the historical result will continue.
Wait when assumptions are missing, the sample is too narrow for the intended decision or you cannot distinguish simulated and live figures. Do not invent performance for a strategy whose published evidence is temporarily unavailable.
Reading evidence in CoinScreener
Open a strategy's Performance view and inspect the published backtest, the trades it covers and its assumptions. Review any live statistics separately. Use the current labels, dates and version when assessing a result.
The product screenshot below is a dated illustration of how to read the interface. Use the current product evidence, with its dates and version, for an assessment. Risk management explains why even a plausible historical strategy still needs position and portfolio limits.
Start with the original simulation assumptions, then compare the equity curve with drawdown beneath it. These are Polaris Gold historical simulation results, not an account statement, current live results or a forecast.
Common questions
Does a profitable backtest mean the strategy will profit next month? No. It describes selected historical conditions under assumptions, not a forecast.
Is maximum historical drawdown the worst possible loss? No. A future decline can exceed the observed historical maximum.
Can I compare two strategies with different test periods? You can inspect each, but a direct ranking from returns alone can be misleading. Compare equivalent periods and assumptions where possible.
Sources and further reading
CoinScreener's performance screens and terminology were checked on 5 September 2026. The arithmetic examples are original and can be checked against the displayed curve. Refer to each strategy's current Performance view for its methodology and results.