A novel approach to the estimation of equivalence ratio in spark-ignited internal combustion engines is proposed, based on a certain cycle-by-cycle statistical analysis of indicated cylinder pressure data. A classical (“zero-dimensional”) first law cylinder pressure thermodynamic model is re-parameterized into a difference equation in which crank-angle is the independent variable, and parameters in the combustion model are estimated from observation of cylinder pressure alone. An efficient algorithm based on statistical maximum-likelihood techniques is derived to produce estimates of combustion duration, ignition delay, and total heat released on a cycle-by-cycle basis. It is then shown that the statistical signature of the estimated parameters (combustion duration in particular) can be exploited to develop a Bayesian estimator for equivalence ratio. Experimental results from a CFR engine are presented together with potential simplifications to the required data processing requirements.