Lay Out a One-Screen Performance Dashboard

[Fairman]

Lay Out a One-Screen Performance Dashboard

Goal: a summary sheet that answers, in one glance, whether you are on track. Design it before you build it.

Top row, volume and result: number of trades, total R, expectancy in R, and the t-statistic. These tell you what has happened and how reliable it is.

Second row, quality of the edge: win rate, average win, average loss, profit factor, and profit factor without the best trade. These show shape and fragility.

Third row, risk: maximum drawdown in R and percent, current drawdown, longest losing streak, and the largest single loss in R, with a flag if it exceeds 1.05R.

Fourth row, behavior: process score, the percentage of clean trades, expectancy of clean trades versus flagged ones, and number of trades beyond the daily limit.

Fifth row, costs: average cost in R, total costs as a share of gross profit, and average slippage on stops.

Sixth row, trend: rolling expectancy over fifty trades, plotted, with a review line.

Rules for the layout. Group related numbers. Use conditional formatting for thresholds, green when within plan, amber when close, red when beyond. Show sample sizes beside every statistic, so you do not over-read a thin number.

Add a date range selector, so you can look at this month, this quarter, or all time.

Print or save a copy monthly.

Exercise: sketch the dashboard on paper with your ten most important numbers, and choose a threshold for each.

[FXS London Scalper]

Top row, volume and result: number of trades, total R, expectancy in R, and the t-statistic.

Putting the t-statistic in the top row next to expectancy is smart: it tells at a glance whether the expectancy is distinguishable from zero yet. Adding the trade count needed for t above 2 at the current expectancy makes the "keep collecting data" message explicit.

[Fairman]

The sample-size reminder is useful, but I would avoid presenting a fixed “trades needed for t above 2” number as if expectancy alone determines it. The estimate also depends on the dispersion of returns; under simple independent observations, the rough t-stat scales with expectancy divided by standard error, so a noisy strategy needs more data than a steady one with the same average. In trading, clustered setups and changing conditions can also make trades dependent, which weakens the usual simple calculation.

On the dashboard I would show the trade count, expectancy after costs, and an uncertainty band together, with the date range and unit clearly labeled. A t-statistic can be a useful diagnostic, but crossing a conventional threshold is not proof that an edge is real—especially after trying many variations or selecting the best-looking period. The “profit factor without the best trade” is a helpful fragility check, and I would pair it with a view of the worst losses and drawdown so removing one winner does not become the only stress test. A compact screen should support a decision to keep observing, review execution, or pause; it should not make a thin sample look more conclusive than it is.