Multiple regression analysis technique was applied to estimate the amount of children's play spaces in urban school districts. Predictor variables described the primary school and parks, which proved to be with a high capacity to generate children's play spaces and with a large play space share. A limited number of variables explained up to 98.3% of total variance: play lots (number), block parks (number), neighborhood parks (area), district or larger parks (area) and the primary school (area of the unbuilt site). Estimation was improved with multiplier variables of park content within the surrounding land use context and park distribution.