The same backtest can show different win rate, profit factor, and Sharpe figures without changing a single trade. Change the unit from trades to days, and you changed the question the metric is answering.
For a prop trader, that is not bookkeeping. Prop firms fail accounts day by day and path by path, so a per-day lens often says more about survival than a per-trade headline does.
What changes when you switch the denominator
Per-trade metrics treat each closed trade as one observation. Per-day metrics collapse everything that happened inside a trading day into one daily result. The trade list can stay identical while the measurement unit changes.
Sharpe is the clearest conceptual example. A per-trade Sharpe uses trade-by-trade variability in the denominator. A per-day Sharpe uses daily variability. Same strategy, different volatility bucket, different number.
The same logic applies to win rate and profit factor. Count winners and losers by trade, and you get one answer. Count winning and losing days, and you get another.
Neither basis is fake. They are just not interchangeable.
| Basis | What one observation means | Best used for |
|---|---|---|
| Per trade | One closed trade | Execution quality, trade frequency, cost sensitivity |
| Per day | One trading day | Path smoothness, daily pain, prop-rule survivability |
A small example with the same 12 trades
Take a toy strategy that closes the same 12 trades over four days.
| Day | Trades closed | Net day |
|---|---|---|
| 1 | +1R, +1R, +1R, -1R | +2R |
| 2 | +1R, +1R | +2R |
| 3 | +1R, -1R, -1R | -1R |
| 4 | +1R, -1R, -1R | -1R |
Read trade by trade, the strategy logged 7 winners and 5 losers. That gives it a 58.3% win rate and a 1.40 profit factor.
Read day by day, the same history logged 2 winning days and 2 losing days. That gives it a 50.0% win rate and a 2.00 profit factor because the gross positive days summed to +4R while the gross negative days summed to -2R.
Nothing dishonest happened. The trade list stayed exactly the same. What changed was the bucket.
You do not even need an exact Sharpe calculation to see the implication. A Sharpe built from twelve individual trade outcomes is measuring a different volatility pattern from a Sharpe built from four daily results. The numerator and denominator are now looking at a different frequency, so the number moves.
That is why a metric without its basis attached is incomplete. It is a label with the unit torn off.
Why prop traders should care more about day-based pain
A prop firm does not fail you because your per-trade profit factor looked low on paper. It fails you because a bad cluster of daily or intraday pain hit the rule first.
That is why the day-based lens matters so much in funding. Daily loss limits, overall loss floors, and floating drawdown rules all care about the path the account takes through time, not just the gross sum of winners and losers. If you want the rule mechanics behind that, daily loss limit vs max loss for prop traders, balance vs equity drawdown for prop traders, and why Monte Carlo drawdown matters for prop traders are the relevant companions.
Per-trade metrics are still useful. They help when you are checking execution quality, spread sensitivity, or whether a strategy trades too often for the edge it claims. But when the real question is "can this survive the account rules?" a day-based and path-based read earns more weight.
That is also why a high win rate can still be flimsy. A strategy can win often, look tidy per trade, and still deliver ugly daily clustering. The basis does not create or destroy the edge. It changes which part of the edge you are examining.
How the published realbacktesting numbers should be read
realbacktesting is a trading-software studio for cTrader built around reproducible backtests, so the metric basis is stated instead of hidden in the footnotes.
Across the three published systems, the current public figures already show the denominator effect:
| System | Profit factor per day | Profit factor per trade | Win rate per day | Win rate per trade |
|---|---|---|---|---|
| Guardian | 1.77 | 1.62 | 51.2% | 44.9% |
| Balanced | 1.89 | 1.70 | 50.2% | 45.6% |
| Edge | 1.90 | 1.67 | 50.9% | 44.6% |
Those are the same published strategies viewed through different measurement units. The site reports risk metrics per trading day. The native cTrader report counts per trade. Compare one column with the other as if the basis were identical, and you are doing arithmetic with the labels removed.
The wider proof chain matters more than any one metric anyway. The published systems are measured on cTrader broker M1 data from 2021-06-01 to 2026-06-20, on an 80,000 EUR model base, with real per-symbol spread, commission, swap, and 1 bps slippage. Signal parity is 100% across 13 strategies and 175,401 bars, and drawdown ceilings are enforced at the 95th percentile of 20,000 Monte Carlo simulations, then checked on a 30% out-of-sample hold-out. The full method is laid out on methodology.html, and the prop-account path model lives on funding.html.
The honest limit matters too. This is still a backtest record on 2021-2026 data, not a live track record. A clean denominator does not turn history into a guarantee.
How to stop fooling yourself with metric headlines
If you want the shortest useful checklist, it is this:
- Ask the basis first. If the metric is shown without saying per trade or per day, the number is unfinished.
- Compare like with like. Do not stack a per-trade figure against a per-day figure and pretend the comparison is clean.
- Put path risk ahead of vanity stats when the goal is funding survival. That is where out-of-sample testing in trading and how many trades you need to trust a backtest still matter.
- Keep summary metrics in their place. Profit factor, win rate, and Sharpe are useful lenses, not substitutes for realism, costs, sample quality, or path stress.
That sounds stubborn because it is. A backtest is easy to flatter when the metric label is doing half the marketing for you.
Frequently asked
Which basis is better?
Neither universally. Per-trade metrics are better for trade-level behavior such as execution quality and cost sensitivity. Per-day metrics are better for path questions such as daily pain and prop-rule survivability.
Can daily metrics hide trade-level problems?
Yes. A daily view can smooth over a strategy that trades too often, pays too much spread, or wins through fragile execution. That is why the two lenses are complements, not replacements.
Does per-day Sharpe replace out-of-sample testing?
No. Basis choice tells you what frequency the metric is measuring. It does not tell you whether the edge is real. You still need sample quality, costs, and out-of-sample evidence.
Why does this matter so much for prop firms?
Because prop firms judge the account by the damage it can take through time. The account does not fail on a pretty average across 3,000 trades. It fails on the path that reached the limit first.
The stubborn takeaway
The same backtest can tell two different statistical stories without lying once. If you do not ask what one observation means, the metric is answering a question you never meant to ask.
For a prop trader, that question usually needs to be about survival by day, not elegance by trade.