To investigate variability, predictability, and smallest worthwhile performance enhancement in elite biathlon sprint events. In addition, the effects of race factors on performance were assessed. Data from 2005 to 2015 including >10,000 and >1000 observations for each sex for all athletes and annual top-10 athletes, respectively, were included. Generalized linear mixed models were constructed based on total race time, skiing time, shooting time, and proportions of targets hit. Within-athlete race-to-race variability was expressed as coefficient of variation of performance times and standard deviation (SD) in proportion units (%) of targets hit. The models were adjusted for random and fixed effects of subject identity, season, event identity, and race factors. The within-athlete variability was independent of sex and performance standard of athletes: 2.5-3.2% for total race time, 1.5-1.8% for skiing time, and 11-15% for shooting times. The SD of the proportion of hits was ∼10% in both shootings combined (meaning ±1 hit in 10 shots). The predictability in total race time was very high to extremely high for all athletes (ICC .78-.84) but trivial for top-10 athletes (ICC .05). Race times during World Championships and Olympics were ∼2-3% faster than in World Cups. Moreover, race time increased by ∼2% per 1000m of altitude, by ∼5% per 1% of gradient, by 1-2% per 1m/s of wind speed, and by ∼2-4% on soft vs hard tracks. Researchers and practitioners should focus on strategies that improve biathletes' performance by at least 0.8-0.9%, corresponding to the smallest worthwhile enhancement (0.3 × within-athlete variability).