In genreral, reliability refers to a condition where testing result is stable during several testings excluding the changes in athlete´s performance. It means, if we test the athlete once and then after very short period of time (we expect no changes in the physical performance of the athlete) then the test result is reliable. For example, testing the canoist on water 1000 meter distance can have low reliability, since waves and current can influence the result and the magnitude of that influence is impossible to detect during each test. Standardization of the various cofounders is vital for having high reliability of the test.
Articles > Reliability of the performance test