A rising curve is the first thing most gold pages show. What decides it sits above: the Strategy Tester header and the inputs list. Ask about the dates first Before the headline ratio, ask which dates the parameters were chosen on, and whether the report sits inside or outside them. Any search over a few thousand combinations finds one that looks clean on the stretch it searched. Whoever did the work answers in one line; whoever did not quotes the headline ratio again. Read the modelling line Every tick based on real ticks replays what the broker recorded. Every tick interpolates inside the minute; 1 minute OHLC keeps four prices per minute and no order between them; Open prices only skips the inside of the bar entirely. A robot that fills a pending order at a level, moves to breakeven and trails a stop lives inside the bar; on open prices all of that is resolved by convention, not tested. Check the label against the count: half a year of gold on real ticks is tens of millions. The Sarveon Strategy Tester report: 2026.02.12 to 2026.08.31, 100 percent real ticks, 44 085 819 ticks over 3 231 bars, a window outside the one the settings were chosen on. Do the arithmetic yourself Divide the average loss by the sum of the average loss and the average win: that is the share of trades that must close positive just to stand still, and everything above it is margin. In that Strategy Tester run, 14.47 / (9.46 + 14.47) = 60.5 percent required against the 75.32 percent the run actually shows, over 863 trades. Run it on any report that impresses you. A run showing 95 out of every 100 trades closing positive, with an average loss twenty times the average positive trade, needs 95.2 to stand still: already under water. Then ask what shaped it: the worst trade against the average loss, whether volume grows after a loss, whether the average trade clears what your own broker charges on gold. In that Strategy Tester run the worst trade is more than seven times the average loss, -107.42 against -14.47, or 1.07 percent of the 10000 USD deposit; volume stays at the lot you set; the average trade is 3.52 USD at a fixed 0.01 lot. Two drawdowns and a ratio Balance drawdown counts closed trades only, equity drawdown counts floating loss too: in that Strategy Tester run, 423.38 USD (3.27 percent) against 631.06 USD (5.17 percent) on a 10000 USD deposit. Against the deposit those are 4.23 and 6.31 percent; the printed figures are smaller because each is measured from the account peak its drawdown began at. Percentages move with lot size; recovery factor does not: in that Strategy Tester run, net result over maximum drawdown, 3041.93 / 631.06 = 4.82. The month you want to find Open the monthly breakdown and ask which month is worst and why it is still in the window. In that Strategy Tester run one of the seven months closed negative, and it stays: a window with the bad month removed is a selection, not a period. Open the inputs list The description says what a robot does; the inputs say what it can do. Scan for a lot multiplier, a grid step, a distance between averaging orders: one above 1.0 explains a smooth curve by itself. Then ask what no report shows: whether stops sit on the server with the order or only inside the robot. In Sarveon the stop and the target are attached to the order itself. Produce a report I never touched The version of this checklist that settles it is the report you run yourself. Free Demo on the Sarveon page runs in your own Strategy Tester, on your broker's XAUUSD ticks and spread. Set 10000 USD and a fixed 0.01 lot, pick your own range, and work through the checks above on the report you get. Then run the same checks on every gold expert on your list.
