Performance Metrics Deep Dive

When you click into the Metrics card on the results dashboard, a full modal opens with four panels arranged in a 2x2 grid, plus a collapsible Raw Metrics row at the bottom. This is the most numerically dense view in the platform, and every element is there for a reason.

The four panels are: Equity Path (top left), Outcome Summary (top right), Risk-Adjusted Metrics (bottom left), and Metric Interpretation (bottom right). Together they answer the question: what did this strategy do, and should I believe it?

At the top, a subtitle tells you the sample size: "Historical simulation outcomes from N trades." The red disclaimer — "Past performance does not predict future results" — is not decoration. It is the single most important sentence on the screen, and the rest of the modal is designed around that reality. Every metric shown is historical. The platform's job is to help you assess whether the historical pattern is likely to persist, not to promise that it will.

The Equity Path chart plots your portfolio value over time, starting from the initial capital and marking a point each time a trade closes. It is not a candle chart of the asset — it is the simulated value of your account over the backtest period.

What the shape tells you

A smooth, upward-sloping line suggests consistent compounding. A line that rises in spikes and sits flat for long stretches suggests the strategy depends on catching a few large moves — the returns are concentrated, not distributed. A line that climbs steeply and then falls back indicates a strategy that gave back a significant portion of its gains, which matters for drawdown analysis.

The platform deliberately does not draw a straight best-fit line or overlay a benchmark. The equity path is for reading the character of the returns: are they smooth or jagged? Concentrated in one period or spread across the whole dataset? Steadily compounding or front-loaded?

Comparison mode

If you have run a baseline comparison (e.g., Walk-Forward Analysis versus a standard backtest), the equity path will overlay the baseline as a dashed line behind the current result. This makes it immediately visible whether the current configuration improved on or degraded from the baseline — and where in the timeline the divergence happened.

Caption: "Shape illustrates return concentration and recovery behaviour"

This caption is a nudge to read the chart qualitatively, not just check whether it goes up. A chart that ends higher but with three 40% drawdowns along the way tells a very different story from one that ends lower but never dropped more than 10%.

The Outcome Summary panel contains four numbers. Each one seems straightforward, but the interpretation is more nuanced than it appears.

Total Return

Calculated as (final portfolio - initial portfolio) / initial portfolio x 100. This is the headline number most people look at first — and it is the most misleading. A 200% return over seven years on Bitcoin during a period when Bitcoin itself rose 500% is not evidence of a strategy working. It may be evidence of a strategy underperforming buy-and-hold.

Total return also hides volatility. Two strategies can produce 80% returns, but if one had a maximum drawdown of 50% and the other 15%, they are fundamentally different propositions. This is why the platform shows Total Return in the Outcome Summary but puts it as a footnote in the Metrics grid card — to prevent anchoring on a single number.

Net P/L

The absolute currency amount gained or lost. This is Total Return expressed in portfolio units (e.g., USDT). On its own, it has the same limitations — it tells you the destination without describing the journey.

Trades

The total number of completed trades. This is the sample size, and it is more important than most traders realise. Statistical confidence in any metric requires a minimum number of observations. Below 30 trades, the metrics on the rest of this modal are unreliable. At 50-100, patterns start to become meaningful. Above 100 on daily or higher timeframes, you have a solid analytical foundation.

Trade count also constrains what kind of analysis is possible. The Behaviour card's profiling, the Trade Structure analysis, and the Monte Carlo simulation all need enough trades to produce stable results. If the number here is low, interpret everything else with proportional scepticism.

Win Rate

The percentage of trades that closed in profit, displayed alongside the raw win/loss count (e.g., "44% (27W / 35L)"). Win rate is almost meaningless in isolation. A 30% win rate with a 4x payoff ratio is highly profitable. A 70% win rate with a 0.4x payoff ratio loses money. The platform always shows win rate next to the Outcome Summary precisely because it needs context from the Risk-Adjusted panel below to mean anything.

The Risk-Adjusted panel is where the real analysis begins. These four metrics describe the structure of the edge — not just whether the strategy made money, but how it made money and with what consistency.

Sharpe Ratio (Annualised)

The Sharpe Ratio measures return per unit of volatility. In the platform, it is computed from per-trade returns (not the equity curve) and annualised based on average trade duration:

1. Calculate the percentage return of each trade relative to portfolio value at entry.
2. Compute the mean and standard deviation of those per-trade returns.
3. Divide mean by standard deviation to get the Sharpe per trade.
4. Annualise: multiply by the square root of (trades per year), derived from 365.25 / average trade length in days.

The platform assumes a risk-free rate of zero, which is standard for crypto backtesting where the alternative to holding a strategy is holding stablecoins.

Thresholds: The Sharpe value is colour-coded. Green (above 1.0) indicates acceptable risk-adjusted returns. Grey (0.5 to 1.0) is marginal. Amber (below 0.5) means the strategy is not generating enough return to justify its volatility — even if the total return headline looks good.

The Sharpe is capped to a range of -10 to +10 to prevent extreme outliers from distorting display.

Profit Factor

Profit Factor is gross profits / gross losses. It captures the full distribution of outcomes in a single ratio.

• Below 1.0 — the strategy lost money overall.
• 1.0 to 1.2 — marginal; transaction costs or slippage could easily flip this to a loss.
• 1.2 to 1.5 — modest edge. Worth investigating further.
• 1.5 to 2.0 — solid edge. This is where most viable strategies sit.
• Above 2.0 — strong edge, but verify with out-of-sample data.

If there are no losing trades, Profit Factor is stored as 999 (effectively infinity). This is a flag for suspicion, not celebration — a strategy with zero losses on a historical dataset almost certainly found a data artefact.

Colour coding: Green above 2.0, grey between 1.5 and 2.0, amber below 1.5.

Expectancy

Expectancy is the average amount you can expect to gain (or lose) per trade, calculated as:

(win rate x average win) - (loss rate x average loss)

This collapses the whole win-rate-versus-payoff interaction into a single currency value. Positive expectancy is the absolute minimum requirement for any strategy. Negative expectancy means you will lose money over enough trades, regardless of how the equity path looks right now.

Expectancy is expressed in portfolio currency (e.g., USDT124.21), which makes it concrete: on average, each trade is expected to produce that amount. Multiply by trade count to roughly cross-check against Net P/L.

Payoff Ratio

The Payoff Ratio is average win / |average loss|. It measures asymmetry: how much larger are the winners compared to the losers?

• Below 1.0 — losers are bigger than winners. The strategy needs a high win rate to compensate.
• 1.0 to 1.5 — roughly symmetrical. Win rate needs to be above 50%.
• 1.5 to 2.5 — moderate asymmetry. Classic range for trend-following strategies.
• Above 2.5 — high asymmetry. The strategy wins big but infrequently.

Payoff and win rate are mathematically linked. The break-even win rate for any payoff ratio is: 1 / (1 + payoff ratio). With a payoff of 2.0, you need to win just 33% of trades to break even. With a payoff of 0.8, you need 56%.

Contextual Insight

Below the four metrics, a one-line caption provides a qualitative summary. Examples include "Returns show reasonable risk-adjusted consistency" or "High absolute return with low risk-adjusted ratio." This is generated from the metric values, not manually written, and it is meant to nudge your attention toward the most salient feature of the risk-adjusted picture.

The Metric Interpretation panel is the platform's attempt to tell you what the numbers mean as a whole. It is split into two sections: Characteristics and Implications.

Characteristics

This describes the outcome structure — the pattern formed by the interaction of the metrics. The platform evaluates the metrics in a priority order and assigns the first matching pattern:

• "High absolute return with low risk-adjusted ratio" — appears when total return exceeds 100% but the Sharpe is below 0.5. The strategy made money, but the ride was volatile relative to the gain. The return might be concentrated in a few lucky trades.
• "Low win rate compensated by larger winners" — appears when win rate is below 45% but Profit Factor is above 1.5. This is the classic trend-following fingerprint: losing often, but winning trades are large enough to more than compensate.
• "Payoff-driven outcome structure" — appears when the payoff ratio exceeds 2.0. The strategy depends on asymmetry rather than accuracy.
• "Strong risk-adjusted performance" — appears when the Sharpe ratio is above 1.0. The return-to-volatility relationship is healthy.
• "Moderate outcome structure" — the default when no specific pattern dominates.

These rules are evaluated in sequence, so a strategy can match the first applicable pattern. For the example in the mock — 44% win rate, 1.52 Profit Factor, 1.97 payoff, 1.49 Sharpe — the low-win-rate-compensated-by-larger-winners rule fires first, then the strong-risk-adjusted rule adds to it.

Implications

This answers the question: what should you do next?

• "Returns may be concentrated in few trades" — paired with high return/low Sharpe. Go check the Trade Structure card to see if profit depends on outliers.
• "Metrics alone overstate reliability" — paired with low-win-rate compensation. The numbers look good, but the strategy is inherently streaky. A run of losses is expected and could be psychologically difficult.
• "Behaviour and drawdown analysis required for assessment" — paired with high payoff. Raw metrics are insufficient; you need the qualitative cards.
• "Verify with out-of-sample or walk-forward testing" — paired with strong Sharpe. Good numbers — now prove they are not overfitted.
• "Review behaviour metrics for strategy assessment" — default. No strong signal either way; use the rest of the dashboard.

The Interpretation panel is deliberately conservative. It never says "this strategy is good." It says "here is the pattern, and here is what to investigate next." This is by design — the metrics modal is one piece of a diagnostic system, not a verdict.

At the bottom of the modal, a collapsible Raw Metrics row displays six values in a flat horizontal layout: Return %, Net P/L, Win Rate, Sharpe, Profit Factor, and Expectancy. These are the same numbers from the panels above, presented without context or colour coding.

The purpose is export-friendly reference. When you are comparing multiple backtests or taking notes, the raw row gives you a scannable, copy-friendly summary. It also serves as a check: if any number here does not match what you see in the panels above, something is wrong with the data (which should not happen, but the redundancy is intentional).

One subtlety: the Win Rate in Raw Metrics may show a more precise decimal (e.g., 43.5%) compared to the rounded percentage in the Outcome Summary (44%). Both are correct — the Outcome Summary rounds for readability, while Raw Metrics preserves the full figure.

The 2x2 layout is not random. The top row tells you what happened (equity shape + outcome numbers). The bottom row tells you whether to trust it (risk-adjusted quality + automated interpretation). This is a deliberate progression from facts to judgment.

A practical reading order

1. Glance at Trade count in the Outcome Summary. If it is below 30, everything else on this modal is unreliable. Note it and move on with scepticism.
2. Read the Interpretation panel first, not last. It will tell you the dominant pattern and what to focus on. This saves you from anchoring on Total Return.
3. Check the Sharpe colour. Green means the returns are earned efficiently. Amber means the total return is misleading — the strategy took excessive risk to get there.
4. Look at the equity path shape. Is it smooth or jagged? Are the returns spread across the period or concentrated in one phase? Does it give back gains?
5. Cross-check Profit Factor against Expectancy. A Profit Factor of 1.5 with an Expectancy of $124 per trade tells a consistent story. A Profit Factor of 3.0 with an Expectancy of $5 per trade suggests a few massive winners distorting the ratio.

Common traps

Anchoring on Total Return: A 200% return over seven years sounds impressive until you realise Bitcoin rose 400% in the same period. Total return without context is decoration.

Ignoring the Sharpe colour: A strategy with a 150% return and a 0.3 Sharpe is a rollercoaster. You might not have the stomach to trade it live, even if the backtest looks profitable.

Treating Profit Factor as the whole story: A Profit Factor of 2.0 from 15 trades is statistically meaningless. Always check trade count before reading Profit Factor.

Skipping the Implications: The platform is explicitly telling you what to investigate next. If it says "verify with WFA," that is not a suggestion — it is the logical next step.

The Performance Metrics modal gives you the quantitative foundation. But it is only the first card in the dashboard for a reason. Numbers alone cannot tell you whether a strategy is robust, consistent, or suitable for live deployment. That is what the rest of the cards — Behaviour, Trade Structure, Drawdown, Regimes — are built to answer.