You pick an asset, a timeframe, and (optionally) a session window. The tab pulls real historical candles for that selection and computes standard quant statistics on them — the same kind of numbers a prop-trading or research desk looks at before sizing a position, not predictions or advice.

Everything below is grouped in the order it naturally builds up: first "how much did price move and which way", then "was that move good risk-adjusted", then "how sure can we be this isn't just noise".

Sample Size is how many candles went into every calculation on this page. Everything else — every ratio, every percentage — is only as trustworthy as this number. Under ~30 candles, treat all other numbers as noise; the tab will say "not enough data" rather than show them.

Bullish % / Bearish % are simply the share of candles that closed up vs. down. On their own they say nothing about *how much* was gained or lost — a market that's 70% bullish but loses more on the 30% red candles than it makes on the green ones can still be a net loser. That's exactly why the return-size metrics below exist alongside these.

Average Return is the mean gain/loss per candle. Median Return is the middle value — use it alongside the average to spot skew: if the median is much smaller than the average, a few huge outlier candles are pulling the average up (or down).

Average High-Low and ATR (Average True Range) both measure typical candle range, i.e. how far price swings within a period regardless of direction. ATR is the more standard version (it also accounts for gaps between candles).

Volatility and Standard Deviation measure how spread-out the returns are around the average. Volatility here is that same spread expressed as a %; Standard Deviation is the raw price-unit version. This single number is the "risk" half of every ratio in the next section — Sharpe, Sortino, and Calmar all divide some version of return by some version of this spread.

All three answer the same question — "how much return did you get per unit of risk taken" — but define risk differently, so they can disagree:

Sharpe Ratio = return ÷ total volatility (punishes upside and downside swings equally).

Sortino Ratio = return ÷ downside volatility only. It's always ≥ Sharpe for the same data, because it ignores the "risk" of moving up. A big gap between Sharpe and Sortino tells you most of the volatility was upside surprises, not downside pain.

Calmar Ratio = annualized return ÷ Maximum Drawdown (see below). It answers "how much did I earn relative to the single worst losing streak", which matters more to someone who can't stomach a big drawdown than average volatility does.

Read them together: high Sharpe + low Sortino gap + high Calmar = smooth, consistent performer. High Sharpe but low Calmar = decent on average, but with one nasty drawdown dragging down the risk-adjusted picture.

These four are computed from the same two underlying numbers: the win rate (≈ Bullish %) and the size of average wins vs. average losses.

Average Win / Average Loss are the building blocks — mean return of the up periods and mean return of the down periods, separately.

Profit Factor = total gains ÷ total losses. Above 1 means winners outweighed losers in total; below 1 means the opposite even if win rate looked fine.

Expected Value = (win rate × avg win) + (loss rate × avg loss). This is literally "what you'd expect to make per period, on average" — the single most direct answer to "was this historically profitable".

Kelly % is derived from win rate and the win/loss size ratio — it's the textbook-optimal fraction of capital to risk per period if this historical pattern holds going forward. In practice people use a fraction of Kelly (e.g. 1/4) because the historical win rate is an estimate, not a certainty — which is exactly what the confidence metrics further down quantify.

Maximum Drawdown is the worst peak-to-trough decline across the whole sample — the deepest hole you'd have been in if you'd held through the sample period. It's the denominator of Calmar Ratio above.

MAE (Maximum Adverse Excursion) is, on average, how far price moved against a candle's own eventual direction before it closed — i.e. how much "heat" you'd have taken holding through it. MFE (Maximum Favorable Excursion) is the opposite: how far it moved in your favor before closing. Comparing MAE to MFE gives a feel for how much a candle "tests" you before resolving.

These five are all one chain of logic, not five separate ideas:

Z-score asks: "how many standard errors is the average return away from zero?" — the bigger the number (positive or negative), the less likely the average return we measured is just random noise around a true average of zero.

P-value converts that Z-score into a probability: specifically, the odds you'd see a result this extreme (or more) purely by chance if the true average return were actually zero. Lower = more confident there's a real effect, not luck.

Confidence Score is just (1 − P-value) × 100 — the same information as P-value, flipped into an intuitive 0-100 scale where higher is better.

Confidence Interval uses the same standard error as the Z-score to give a range: "95% of the time, the true average return would fall somewhere in this band." A wide band means the point estimate (Average Return) is uncertain even if the sign is consistent.

Data Quality is the blunt, sample-size-only version of all of the above — a quick sanity check before you even look at the p-value.

Rule of thumb: a great Sharpe Ratio or Kelly % computed on a Low/Insufficient Data Quality sample should be trusted much less than a mediocre one computed on High Data Quality — always check this row first.

Session Behavior re-runs the bullish%/avg-return/avg-range numbers separately for Asia, London, New York, Friday Close, and Monday Open (sessions defined in UTC, so they don't shift with daylight saving) — useful for spotting whether a market's character actually changes by time of day.

Gap Analysis looks only at the jump between one candle's close and the next candle's open, and how often price later trades back through that pre-gap level within 24h/48h/72h ("filling" the gap) — a classic mean-reversion tell.

Return Distribution is a histogram of every individual return in the sample — it shows the shape behind the Average/Median/Std Dev numbers above (a fat middle with thin tails looks very different from one with frequent large outliers, even if the average is identical).

Seasonality and the Hourly Return Heatmap break the same average-return calculation down by weekday and by UTC hour respectively, to reveal patterns that a single blended average would hide.

1. Check Data Quality and Sample Size first — decide how much to trust everything else.
2. Look at Average/Median Return and Bullish/Bearish % for the basic direction and size of the edge, if any.
3. Check P-value / Z-score / Confidence Interval — is that edge statistically distinguishable from zero, or could it just be noise?
4. Use Sharpe / Sortino / Calmar to judge whether the edge was smooth or came with rough patches.
5. Use Profit Factor / Expected Value / Kelly % to translate the edge into "would this have actually been worth trading, and how much of it."
6. Use Session Behavior / Gap Analysis / Seasonality to see if the edge concentrates in specific times or conditions rather than being uniform.