Measuring the performance of the compression-ignition internal combustion engine of agricultural machines (in particular, the torque delivered instantaneously) is an essential requirement for monitoring: (a) the exploitation of the engine mechanical-energy potential (in terms of generated torque) and (b) its correct operation (in terms of global efficiency, fuel consumption and possible ageing). Due to the many important technical and economic implications, the instant acquisition of the engine torque is therefore a critical point in any operational monitoring system, as well as in predictive maintenance models. Torque measurement is by no means a simple task, especially in old agricultural machines lacking of default data acquisition devices/on-board electronics, and many critical issues arise from the fact that it involves rotating components (shafts), which are often difficult to be accessed. For this reason, an indirect torque measurement methodology, based on a predictive model relying on the exhaust gas temperature, is preferable. An accurate measurement of temperature data is of primary importance to precisely calculate the torque, which means performing an accurate thermocouple choice, placement, and data elaboration. This is made even more challenging by the fact that the temperature of the exhaust gas is often in a transient state due to variable engine regime necessary for machine operation. The study presented here illustrates some considerations about the trend and the equation of the experimental measurements of the exhaust gas temperature, considering three different positions for the thermocouples on the exhaust line, and proposes an optimal technical solution in terms of sensitivity and promptness of response.