Granular materials have frequently been used as representations of natural fault gouges. Although they can reproduce proper avalanche behaviors, the universality of the scaling exponent of avalanche size remains debatable. As a core issue in both amorphous plasticity and geophysics, avalanche universality may help reconcile the avalanche behaviors of earthquake and granular materials into the same universality class. We examine numerically the signatures of stress avalanches emerging from quasi-static shear of granular materials with different size polydispersity. A persistent serrated plastic flow phenomenon is observed in our models with varying polydispersity. The stress drop is well defined by a truncated power law distribution P(s)~s −τ exp(−s/s max). The exponent τ and cutoff stress drop s max show a clear dependence on polydispersity, which reflects a tuned criticality. We further calculate the effective temperature from the statistics of energy fluctuations. The effective temperature volatility can be used to explain the tuned critical behaviors of granular gouge. Plain Language Summary Granular materials are ubiquitous in nature and important to a wide range of industrial processes. When driven at slow rates, granular materials deform via intermittent dynamics which alternates slow elastic loading with rapid slips. Such serrated behaviors have attracted attentions from researchers in material science, process engineering and powder technology. They have aroused great interest for geoscientists since key clues may be offered for the origin of earthquake physics. The avalanche-like slip events of granular materials reveal a strong statistical similarity with earthquakes, having made granular materials widely adopted as an ideal laboratory model material for the study of earthquake physics. Different granular systems subjected to different loading conditions present similar scaling laws, such as the power law distributions of event size. However, the scaling exponents describing avalanche distributions vary greatly in range, posing challenges to the existence of universal scaling across diverse granular systems. In this study, we examined granular systems with varying size polydispersity from monosized spheres to highly polydisperse packings and found the signature of tunable avalanche statistics. We further use size polydispersity as a control parameter to generate continuously adjustable avalanche statistics.