GHPR is the geometric average return per position. Treat each position return as a multiplier (a 2 percent gain is 1.02, a 1 percent loss is 0.99), multiply them all together, take the nth root where n is the position count, subtract 1.
The arithmetic average lies when results vary. Gain 50 percent then lose 50 percent and the arithmetic average says zero. The reality is a 25 percent loss. GHPR catches this. It is the per-position return that, if held constant over the same number of positions, would have produced the same final equity. That makes it the honest figure when you are projecting compounded growth over a long sequence.
A GHPR of 0.5 percent per position does not sound like much. Compounded over hundreds of positions, it becomes a real number. But the metric on its own says nothing about the variance of individual results. A 0.5 percent GHPR can come from a strategy that wins 60 percent of the time with small swings, or from one that wins 20 percent of the time with violent ones. Same average. Two very different rides.
Javlot reports GHPR alongside position count and the rest of the variance picture: drawdown, Sharpe, Sortino. The headline averages, plus the shape of the volatility around them.