When Does a Grid Basket Run On? Size, Speed and Price State in a Gold (XAUUSD) Grid-Trading Model: Backtest Evidence and a Pre-Registered Live Test

Grid (averaging-down) trading models add larger orders as price moves against an open position and close the whole group, or basket, together. We study one automated gold (XAUUSD) grid model using every closed order from 2 January to 9 October 2026, rebuilt into baskets and matched to one-minute prices. A case is a basket at the moment its largest single order first reaches a lot level x; the outcome is whether the basket later opens an order of at least 2x ("runs on"), and the floating loss still ahead of it. The model changed its order-size steps on 12 August 2026, so the two settings are analysed separately. In the current setting (2,091 cases from 788 baskets), 29.4% of cases at 0.15–0.25 lots ran on, against 48.6% at 0.7 lots and above: a difference of 19.2 percentage points (95% basket-bootstrap interval 6 to 31; 14.5 points in a worst-case treatment of six baskets that could not be rebuilt). The share of baskets whose later floating loss exceeded 0.05% of the account balance rises from 3.7% at 0.15 lots to 50.9% at 0.7 lots. Time taken to reach a level is associated with running on in the earlier setting but not in the current one, and a 24-measure model ranks out-of-sample cases only slightly better than chance. All backtest results are exploratory. Hypotheses for a live test of the running model were published on 21 October 2026, before the first live case; this paper reports the design and the backtest only.