Abstract Apparent nitrogen fertilizer replacement values of grass–clover leys ( NFRV GCL ) and farmyard manure ( NFRV FYM ) were studied in a long‐term (24 years) experiment. This paper reports the results for grass–clover leys ( GCL ). Five rates of farmyard manure ( FYM ) and four rates of fertilizer nitrogen (N) were applied to two arable rotations (RotA – arable+silage maize; RotB – arable+ GCL ). Polynomial response surfaces fitted to dry matter ( DM ) yield and N offtake in the three arable crops were used to assess NFRV GCL by interpolation, at high (Method H) and low (Method L) N fertilizer rates, always comparing RotA and RotB at equal FYM rates. In Expt1 (started in 1980), arable crops in RotB required 188, 246, 270, 295 and 312 kg N/ha less fertilizer per cycle than in RotA, to match the DM yields found at highest fertilizer rate in RotA (Method H). These values refer to FYM rates of 0, 20, 40, 60 and 80 t/ha per cycle, respectively. Corresponding fertilizer savings in Expt2 (started in 1981) were 287, 323, 341, 346 and 337 kg N/ha per cycle. These NFRV GCL values represent savings of 50–83% (Expt1) or 77–92% (Expt2) in N fertilizer applied to the arable phase. Slightly lower NFRV GCL was found by Method L. Extra N offtake in RotB over RotA was a poor estimator of NFRV GCL , showing much lower values and different responses to FYM and fertilizer than NFRV GCL . Mixed leys enable large fertilizer savings in arable rotations.