Misfire seriously impacts on performance, emissions, engine life, etc. When the engine is running at different condi- tions, causing an unstable behavior of combustion, it will bring challenges to the accuracy and robustness of misfire detection. In this pa- per, an analysis of misfire detection algorithms based on instantaneous power and instantaneous torque is presented. The instantaneous crankshaft(engine speed)signal has been recorded. The optimal angular position and the selected final algorithm to detect misfire, with- in the engine cycle, for the combustion time evaluation, and its variation with the engine operating condition have been put forward. Some related factors affecting misfire algorithm detection are finally compensated. During the engine is running, roughness threshold self-adap- ting is realized to improve the detection accuracy. Finally, the misfire detection algorithm is proved in the car, and the test result shows that misfire event and its corresponding misfire cylinder can be precisely detected.