Early number skills are critical predictors of academic achievement, which is why focusing on their instruction from the very beginning of education is recommended. Young children's number knowledge is also strongly influenced by home numeracy practices. This 12-week quasi-experimental study tested whether early number skills could be enhanced by a play-based intervention implemented at kindergarten (aged 4-6 years) by the teachers and whether providing numerical games to families delivered added value. A total of 569 children from 46 kindergarten classes were assigned either to the first treatment condition in which games were played only at kindergarten, or to the second treatment condition in which games were played at kindergarten and at home, or to the business-as-usual control condition. Measures of numerical ability were collected at pretest and posttest and analyzed through item response theory and multilevel modeling. Results indicated that playing the games at kindergarten allowed children with average and above-average initial performance to make more progress than children in the control group, while providing the games at home allowed low achievers from various backgrounds to progress more than in the other conditions. Implications for early mathematics instruction and for home-based intervention studies are discussed.