The　security　and　fairness　of　blockchain　are　always　threatened　by　selfish　mining　attacks.　To　study　such　selfish　mining　attacks,　some　necessary　and　useful　methods　need　to　be　developed　sufficiently.　In　this　paper,　we　provide　an　interesting　method　for　analyzing　dynamic　decision　of　blockchain　selfish　mining　by　applying　the　sensitivity-based　optimization.　Our　goal　is　to　find　the　optimal　dynamic　blockchain-pegged　mining　policy　of　the　dishonest　mining　pool.　To　this　end,　we　consider　a　blockchain　system　with　two　mining　pools:　the　honest　and　the　dishonest　mining　pools,　where　the　honest　mining　pool　follows　a　two-block　leading　competitive　criterion,　while　the　dishonest　mining　pool　follows　a　modification　of　two-block　leading　competitive　criterion.　To　find　the　optimal　blockchain-pegged　mining　policy,　we　develop　the　sensitivity-based　optimization　to　study　dynamic　decision　of　blockchain　system　through　setting　up　a　policy-based　Poisson　equation,　and　provide　an　expression　for　the　unique　solution　of　performance　potentials.　Based　on　this,　we　can　characterize　monotonicity　and　optimality　of　the　long-run　average　profit　with　respect　to　the　blockchain-pegged　mining　reward.　Also,　we　prove　the　structure　of　the　optimal　blockchain-pegged　mining　policy.　The　methodology　and　results　derived　in　this　paper　significantly　reduce　the　large　search　space　of　finding　the　optimal　policy,　thus　they　can　shed　light　on　the　optimal　dynamic　decision　research　on　the　selfish　mining　attacks　of　blockchain　systems.