Reading the cumulative P&L chart
The /app chart plots the running sum of every closed trade's net P&L in USD. Real numbers from your subaccount — not a simulated $1000 backtest.
What each point is
The y-axis is cumulative net P&L in USD. The x-axis is the close timestamp of each trade. Every trade contributes one point — the point's y value is the running total up to and including that trade's net P&L (net = gross P&L minus dYdX fees minus funding).
A point at ($1234, 2026-04-12) means: "by April 12, 2026, the cumulative net P&L was $1234". The line connecting points is interpolation only — there is no 'P&L between trades'.
The strategy filter
The three pills (all / v6.4 / v7) above the chart let you scope the curve. 'All' shows the unified curve across both strategies. Clicking 'v6.4' recomputes the running sum from scratch over v6.4 trades only — so the curve in that view is what v6.4 produced standalone, not a slice through the combined.
Same logic for 'v7'. This is the cleanest way to compare strategies head-to-head over the same time window.
What good vs bad looks like
A healthy run is a slow upward staircase — many small green steps interspersed with occasional small red steps. The slope is what matters; absolute values depend on how much capital you started with.
A sudden cliff downward usually means a stop loss got hit on a big position (the SL is bounded per-coin but the dollar amount depends on your equity at the time). A long flat period means few or no signals were triggering — usually because the regime gate is filtering them out.
Comparing to backtest
The backtest is a $100-starting-capital walk-forward simulation; the live chart is on your real capital. They're not directly comparable in absolute terms. What IS comparable: drawdown shape, win rate, average trade size relative to equity.
If your live curve looks flatter or red while the backtest looks bullish, two common causes: (a) regime gate is just filtering everything out for that period — wait for it to re-engage, or (b) you got unlucky with the SL distribution. Neither means the strategy is broken; backtests show occasional 6-9 month flat periods historically.