[maple3142]

In V8 engine, Math.random is implemented using xorshift128, which is a completely linear PRNG with 128 bits of state. Since you can recover 1 bit from each output, 128 outputs should be enoguh for you to setup a GF(2) linear system and solve for the state. Once you have the state, all you need is to simulate how V8 output future random values from state.

There are a series of challenges called "fastrology" from PlaidCTF 2023, which is about predicting Math.random() given some partial output (e.g. Math.floor(Math.random() * k)) with several variants. You can try to find writeups for that challenges, those solvers should be easy to adapt to predicting Math.random()>0.5 .