Long-term changes of El Niño/Southern Oscillation (ENSO) are studied using an ensemble ocean reanalysis that includes eight ensemble members from 1871 to 2008 and coupled model intercomparison project phase 5 (CMIP5). The presence of greater “weather noise” in the ensemble member forcing relative to the ensemble average has a conflicting impact on ENSO. In the western Pacific, higher wind speed cools sea surface temperature, which leads to a weaker zonal temperature gradient and hence weaker ENSO. In contrast, stronger episodes of westerly wind, which get averaged out in the ensemble mean, can lead to a stronger than average ENSO in some ensemble members. The ensemble reanalysis has weak El Niño throughout the record, whereas strong El Niño occurs at the beginning and end of the record. The strength of La Niña is weaker than for El Niño and has less variability, a prominent asymmetry in the reanalysis. ENSO variability is also analyzed in the CMIP5 historical experiments. Results show that most of the models have a realistic representation of the strength of ENSO; however, the location generally extends too far to the west. None of the CMIP5 models investigated show a significant change in the strength or location of ENSO from the mid-nineteenth through the twentieth centuries. One distinguishing difference between the CMIP5 models and the ensemble reanalysis is that ENSO in the reanalysis has prominent asymmetry between El Niño and La Niña, whereas ENSO in the CMIP5 models tends to have fairly symmetric El Niño and La Niña.