A set of statistical patterns characterizing the conditions for failure of Type 304 stainless steel clad light water reactor fuel is formulated using a 450-assembly cycle data base from the Connecticut Yankee reactor and an information-theoretic (entropy) criterion of pattern formation. The pattern files, consisting of features formulated from output data obtained from the deterministic fuel performance code FCODE-BETA/SS, are partitioned into five failure and six nonfailure patterns. The failure probabilities of the patterns span the 0.17 to 37.47% range, as compared with the data base average of 2.86%. Features that contribute to failure patterns include large swings in the linear power at high burnup, frequent stress cycling at the ridge, and low coolant pH at high linear power. Contributing to nonfailure are low fission gas release, high axial uniformity in linear power, peak burnup, and clad creep strain at the ridge. The feature describing cycling of the linear power agrees qualitatively with previously identified factors contributing to failure of stainless steel clad fuel in the Connecticut Yankee reactor. From an operational standpoint, the fuel failure probability can be reduced by minimizing the number and magnitude of power ramps and maintaining a neutral coolant pH.