In order to explore the selectivity of different mesh size codends for Larimichthys polyactis, the traditional selective method (cover net method) was adopted, and a selection experiment was carried out with 5 different mesh sizes (35 mm, 45 mm, 55 mm, 60 mm, 65 mm) codends on canvas stow net in the middle of Zhoushan fishing ground. The Logistic selection model was used to fit the selectivity curve of L. polyactis and the model parameters were estimated by the maximum likelihood method. The results showed that:① during the experiment, the main catch target was L. polyactis, the number and weight accounted for 66.94% and 84.48% respectively. The body length range was 71 to 190 mm, and the average body length was 124 mm; ② with the increase of mesh size, the dominant body length of L. polyactis in the codend was larger than that of the cover nets; ③ when the mesh size for L. polyactis was 35 mm, 45 mm, 55 mm, 60 mm and 65 mm, its L₅₀ was 80.425 mm, 103.496 mm, 119.711 mm, 122.676 mm and 137.026 mm, its SR was 50.202 mm, 52.540 mm, 61.120 mm, 41.730 mm and 63.560 mm, respectively. L₅₀ showed a gradual increase trend; ④ the relationship between the codend mesh size (m) of L. polyactis and L₅₀ needs to be enlarged, and L₅₀:L₅₀=L₅₀=1.776 m+20.300 (R²=0.980); ⑤ the number escape rate of 5 mesh size were higher than the weight escape rate, when the mesh size of codend ≤ 45 mm, escape rate in number and weight of L. polyactis was less than 5%, when the mesh size of codend ≥ 55 mm, escape rate in number and weight of L. polyactis more than 10%, even up to 50%. Comprehensive analysis shows that the mesh size of 55 mm can effectively release juvenile fish of L. polyactis, but the mesh size still needs to be enlarged.