In this work a study of the influence of the fuel/air equivalence ratio and engine rotational speed on the cycle-to-cycle variations in combustion in a natural gas spark ignition engine is presented. The study considers both classic estimators of cyclic dispersion and a new one, based on the burned mass and burning rate. The engine experimental conditions were as follows: Intake pressure 0.5bar, while fuel/air equivalence ratio was changed from 1.0 to 0.63, and engine rotational speed was varied from 1000rpm to 2500rpm. For each equivalence ratio and engine speed, a diagnosis model is used to process the experimentally obtained combustion pressure data in order to provide combustion relevant results such as the mass burning rate at a cycle level. A procedure based on the use of genetic algorithms is used to obtain a very accurate and objective (without human intervention) adjustment of the optimum parameters needed for combustion diagnosis: angular positioning and pressure offset of the pressure register, dynamic compression ratio, and heat transfer coefficients. The model allows making the diagnosis of series of 830 consecutive engine cycles in an automatic way, increasing the objectivity of the combustion diagnosis. The paper focuses on using the values of the mass fraction burned computed from the pressure register and especially on the analysis of the combustion cycle to cycle variation in the natural gas fuelled engine. A new indicator for the study of cycle-to-cycle variations is proposed, i.e. the standard deviation of the mass fraction burning rate. The values of this new indicator are compared with other classic indicators, showing the same general trends. However, a deeper insight is provided on the combustion cyclic variation when the values of the new indicator are plotted as a function of the mass fraction burned, since this allows analyzing the cyclic variation along the combustion development in each cycle from a mass fraction burned of zero to one, with a relevant value at mass fraction burned of 0.5. More important is that the consideration of the dependence of the combustion variables (density, flame front surface, combustion speed) on the mass fraction burned allows ensemble averaging of all registered cycles for each value of mass fraction burned. This permits using the ensemble averaged mass fraction burning rate as an estimator of combustion speed.The analysis of the general trends of cyclic dispersion when engine speed and equivalence ratio are modified (1000, 1750 and 2500rpm; 0.7, 0.8, 0.9 and 1.0) indicate that cycle-to-cycle variations show, as expected, a strong dependence on the engine rotational speed, increasing the variation with engine rpm. However, when the standard deviation of mass fraction burning rate is plotted as a function of mass fraction burned, there is a linear dependence on engine rpm, but only a very weak dependence on equivalence ratio. This means that the proposed estimator of cyclic dispersion is sensitive to only flow turbulent intensity and not to equivalence ratio.