Nobody can hand you proof that an indicator works for you. Your market, timeframe, session, sizing and patience are all inputs, and no vendor has any of them. What you can do is run a test that would genuinely be capable of failing — which is what separates testing from looking for reassurance.
Here's a method that takes about a week of evenings.
Step 0: write the rules down before you look at a chart
If the rules are in your head, you will edit them as you go. Every skipped trade becomes "that one didn't count", and the test becomes a search for confirmation.
Write, in one place:
Market and timeframe. One of each. Not "whatever looks good".
Entry condition. Exactly what must be true, on a closed candle.
Stop placement. A rule, not a feeling — a structure level or a volatility multiple.
Target or exit rule. Fixed multiple, trailing rule, or time-based exit.
Filters. Session hours, minimum volatility, no-trade conditions.
Sample size. Decide now. 100 trades is a reasonable minimum; 30 tells you almost nothing.If you can't write the entry as a sentence someone else could follow, you don't have a testable rule yet.
Step 1: check it doesn't repaint
Everything downstream is worthless if the historical signals aren't the ones that were live. Run the bar replay check in the repainting guide first. Five minutes, and it invalidates a surprising number of tools before you spend a week on them.
Step 2: bar replay, one candle at a time
TradingView's Bar Replay is the honest version of a backtest, because it only shows you what existed at that moment.
Rewind to a start date you chose before looking at the period — otherwise you will pick a period that suited the tool.
Step forward candle by candle. No scrolling ahead.
When the entry condition completes on a closed candle, log the trade whether or not you like the look of it.
Move the stop and target only as your written rule permits.The discipline of not scrolling right is the entire exercise. Hindsight arrives instantly and invisibly.
Step 3: log what actually matters
A results-only journal ("win, win, loss") teaches nothing. Per trade, record:
| Field | Why it matters |
|---|
| Date, market, timeframe | Lets you split results later by condition |
| Direction | Long-only edges are common and easy to miss |
| Entry, stop, target | Reconstructs the trade without memory |
| Risk in R | Normalises everything to one unit |
| Outcome in R | The only comparable measure of result |
| Market condition | Trending, ranging, news-driven |
| Rule followed? | Yes/no — the most important column on the sheet |
Measuring in R (multiples of the amount risked) rather than dollars means the record stays valid when your account size changes.Step 4: read the sample honestly
With 100 or more logged trades, the numbers you want are expectancy and the shape of the losses, not the win rate. A 40% win rate at 2.5R is a stronger tool than an 80% win rate at 0.25R, and the second one feels far better while you trade it.
Run your numbers through the win rate and expectancy calculator and then through the risk of ruin calculator. The second one is the important one: it asks whether your sizing survives the worst losing streak the sample implies, and that question kills more strategies than expectancy does.
Then check:
The longest losing streak. Could you have taken it at real size without deviating? If not, the test result is not achievable for you.
Rule adherence. Split your log by the "rule followed?" column. If the deviations outperform, your written rule is wrong. If they underperform — which is more common — the tool isn't the problem.
Concentration. If two outlier trades produced most of the result, you have a small-sample artefact, not an edge.Step 5: subtract the costs
A strategy is only real after spread, commission and slippage. On short timeframes those costs frequently consume the entire edge, and this is where most promising tests quietly die.
Run your average trade through the fee and break-even calculator using your broker's real numbers, then recompute expectancy net of costs. If the edge survives, it's worth continuing.
Step 6: forward test on a small live sample
Replay removes hindsight but not emotion. Trade the rules at the smallest size your broker permits, or on paper if you'll take paper seriously, for 20 to 30 trades.
You are testing one thing here: whether you can follow the rule when money is moving. Most failures at this stage are behavioural, and knowing that is more useful than another hundred replay trades.
What would make you stop?
Decide in advance. "I abandon this if expectancy after costs is negative over 100 trades" is a test. "I'll see how it goes" is not — you'll keep adjusting until something looks good, which is curve-fitting performed by hand.
Run this on us
We'd rather you did this with SimpleAlgo V5 than take our word for anything. The signals finalise on the candle close, so bar replay reflects live behaviour, and the stop and target levels are drawn on the chart, which means every trade in your log has an objective entry, stop and target instead of a judgement call.
The 7-day money-back guarantee exists mainly so there's time to run Steps 1 through 4 before deciding.
We publish no win rates, no return figures and no backtest statistics — for our own tools or anyone else's — because no vendor can prove those numbers to you and you should not trade on a number you cannot verify. Everything above is educational. Trading carries a substantial risk of loss. SimpleAlgo is not affiliated with, endorsed by, or sponsored by TradingView. TradingView is a trademark of TradingView, Inc.
