ABSTRACT Change of direction speed (CODS) tests can provide insights about the acceleration and deceleration abilities of soccer players. We compared peak accelerations and decelerations during matches, training sessions and the pro-agility test. Sixteen (n = 16) players from the men’s Portuguese first division were monitored with global navigation satellite systems during half of one season. Peak efforts were retrieved during matches, training sessions and the pro-agility test. Mean paired differences with effect sizes [95% confidence intervals] compared peak efforts registered during matches, training sessions and the pro-agility test. The pro-agility test peak accelerations were similar to matches (p = 0.81; d = 0.09 [−0.68, 0.88]; unclear effect size), but largely different from training sessions (p < .001; d = 1.30 [0.78, 2.05]); increasingly, peak accelerations differ moderately (p = 0.02; d = 0.95 [0.24, 1.83]) between matches and training sessions. The pro-agility test peak decelerations differ largely from matches (p < .001; d = 3.83 [2.85, 5.43]) and hugely from training sessions (p < .001; d = 5.74 [4.37, 8.05]); additionally, peak decelerations were similar between matches and training sessions (p = 0.26; d = 0.43 [−0.31, 1.25]; unclear effect size). The findings of this study indicate that the pro-agility test can be a good predictor of match peak accelerations. Finally, monitoring training sessions may be a more appropriate strategy to predict peak decelerations during matches.