Accelerometers measure complex movements of children's free play moderate-vigorous physical activity (MVPA), including step and non-step movements. Current accelerometer technology has introduced algorithms to measure steps, along with counts. Precise interpretation of accelerometer-based cadence (steps/min) cut-points is necessary for accurately measuring and tracking children's MVPA. The purpose of this study was to assess the relationships and agreement between accelerometer-based cut-points (cadence and counts/min) to estimate children's MVPA compared to measured values. : Forty children (8⁻12 years; 25 boys) played 6⁻10 games while wearing a portable metabolic analyzer and GT3X⁺ to measure and estimate MVPA, respectively. Correlation, kappa, sensitivity, and specificity assessed the relationships and agreement between measured and estimated MVPA. : Games elicited, on average, 6.3 ± 1.6 METs, 64.5 ± 24.7 steps/min, and 3318 ± 1262 vertical (V) and 5350 ± 1547 vector-magnitude (VM) counts/min. The relationship between measured and estimated MVPA intensity was higher for cadence ( = 0.50) than V and VM counts/min ( = 0.38 for both). Agreement using V and VM counts/min for measuring PA intensity varied by cut-points (range: 6.8% (κ = -0.02) to 97.6% (κ = 0.49)), while agreement was low using cadence cut-points (range: 4.0% (κ = 0.0009) to 11.3% (κ = 0.001)). : While measured and estimated values were well correlated, using cadence tended to misclassify children's free-play MVPA.