Value per acre seems like a better metric than just value per lot, but it seems like the metric you'd really want to maximize is value per infrastructure cost,

Value per acre seems like a better metric than just value per lot, but it seems like the metric you'd really want to maximize is value per infrastructure cost, which isn't quite the same thing. For instance, narrow, deep lots would have less street frontage than square lots and therefore have lower infrastructure cost and be more efficient. Similarly, having large areas of undeveloped forest or park land don't necessarily increase costs much if they don't require major infrastructure. (One could reasonably debate whether undeveloped space significantly increases the need for roads elsewhere to compensate for reduced density.)