Because JITs are able to generate code that optimistically assumes variables have certain types (with appropriate deoptimization checks), which is much less exp

Because JITs are able to generate code that optimistically assumes variables have certain types (with appropriate deoptimization checks), which is much less expensive than performing full dynamic lookups on every single operation.