In this paper, the results and main conclusions of a R&D project carried out to achieve the highest level of performance in a 125cc. 2-S engine for a World Championship racing motorcycle are presented and discussed. Both computational (Modelling Techniques) and experimental (steady flow rig and motorcycle roller test bench) means have been combined, trying to accomplish the highest degree of optimization. An important improvement in the engine's performance was obtained as a result.The paper also describes in some detail the organization of the study, the possible strategy of optimization, the characteristics of the new computer code used to perform the calculations of these very difficult engines, and the experimental measurements.With the objective being to achieve the highest level of performance in a 125 cc. 2-Stroke engine for a World Championship racing motorcycle, a wide R&D project has been carried out by the use of experimental and computational techniques. Typically, the development of this kind of engine is based on direct testing on an engine test bench or a motorcycle roller test bench, and then finally on the racing track.The use of computer modelling techniques in the past few years has revolutionized all the engineering fields, and has brought a significant improvement to the designed products. A special field in which computer codes are increasingly used is in the R&D of engines. Computer modelling can save lots of money and time on a development campaign, but its advantage is not limited to that. Both computational and experimental techniques are directly interrelated and have been shown to be complementary with each other. Their simultaneous application becomes a powerful tool in accomplishing the highest degree of optimization. Computation allows fast and accurate determination of the main design parameters; and then, testing on a steady flow rig, an engine bench, etc… permits the final optimization of the design and gives information about factors whose influence is difficult or impossible to consider adequately in the computation.However, engines for racing vehicles belong to a very special world, and the application of computational modelling to these engines faces several serious difficulties. First, racing teams have little time to concentrate on the development of their engines, mainly because they are always in a hurry, urged by the racing calendar, and also because the usual instabilities they suffer in their staff composition and economic support. Second, the engines are very special, their operation conditions very critical, and therefore their modelling is not a mere extrapolation from conventional engines. Third, they are always keeping information as classified and top secret, so long term cooperation with other teams and even inside the team becomes very difficult.The authors had the opportunity to apply the proposed combination of computational and experimental techniques to one of those engines: a 1993 HONDA RS125R. That R&D project was carried out by the University for Jorge Martínez “ASPAR”, one of the Spanish racers who participated in the 1993 World Championship. The access to the information was very difficult and the availability of material and the construction of prototypes were very reduced, but the results in the end were very good.This paper describes the organization of the study, a possible strategy of optimization, the characteristics of the computer code and the experimental measurements which were used, and the results and conclusions we finally got.