I'm starting to teach myself quantitative finance and I've got several questions ( marked in bold ) regarding the replicating portfolio of a security in the binomial model. I'm following, among others, the classical book " Stochastic Calculus for Finance I: The Binomial Asset Pricing Model ". First, I'll start with some notation so that there is no confusion. As always, let dd , uu , rr be the down-factor, up-factor, and risk-free interest rate, respectively, verifying that d<1+r<ud < 1 + r < u . Then, if VnV_{n} , 0nN0 \leq n \leq N , is the value at time nn of a security that has a unique payoff VNV_N at maturity, we know that the discounted process Vn(1+r)n\dfrac{V_n}{(1+r)^n} is a martingale under the risk-neutral probability measure P~\widetilde{\mathbb{P}} , where the probability of heads is p:=1+rdudp := \dfrac{1+r-d}{u-d} , and thus we can easily compute the value of each VnV_n via Vn=E~n(VN(1+r)Nn).V_{n} = \widetilde{E}_{n}\Big(\frac{V_{N}}{(1+r)^{N-n}}\Big). If d=dnd = d_{n} , u=unu = u_{n} , and r=rnr = r_{n} are not constant numbers but an adapted stochastic process, the result is the same but now the discounted process is given by Vn(1+r0)(1+rn1)\dfrac{V_n}{(1+r_{0})·\dots·(1+r_{n-1})} , and the risk-neutral pricing formula still holds true provided that P~(wn+1=Hw1,,wn):=pn:=1+rndnundn,\widetilde{\mathbb{P}}(w_{n+1} = H|w_{1},\dots,w_{n}) := p_{n} := \dfrac{1+r_n-d_n}{u_n-d_n}, P~(wn+1=Tw1,,wn)=1pn.\widetilde{\mathbb{P}}(w_{n+1} = T|w_{1},\dots,w_{n}) = 1- p_{n}. To show the previous result, one usually constructs the following portfolio: suppose that VV is, for instance, an European call. Start with X0X_{0} wealth, buy Δ0\Delta_{0} shares of the underlying, and invest (or borrow) the remaining money at the risk-free rate rr . At time 11 , sell the portfolio and reinvest the money doing the same strategy. At time n+1n+1 , the value of the replicating portfolio is given by Xn+1=ΔnSn+1+(1+r)(XnΔnSn).X_{n+1} = \Delta_{n}S_{n+1} + (1+r)(X_{n}-\Delta_{n}S_{n}). Now that the notation is clear, my first question is the following: 1) Do we need to replicate the derivative security using the underlying security SS ? I'm aware of the hedging benefits of combining a derivative and its underlying, but since here the goal is to construct a replicating portfolio, could it be constructed trading with another security? What are the advantages of using the underlying over the rest of securities? The only benefit I see is that you only need to model the prices of one stock. Now suppose that we want to price zero-coupon bonds using the binomial model. Assume that the interest rates form an stochastic adapted process, in such a way that 1 dollar invested at time nn yields (1+rn)(1+r_{n}) at time n+1n+1 . Let Bn,mB_{n,m} be the value at time nn of a zero-coupon bond that pays 11 dollar at time mm . Since the risk-neutral pricing formula also applies here, we can easily conclude that Bn,m=E~n(1(1+rn)(1+rm1)).B_{n,m} = \widetilde{E}_{n}\Big(\frac{1}{(1+r_{n})·\dots·(1+r_{m-1})}\Big). However, here goes another question: 2) How would one construct a replicating portfolio in this case? In what securities does it make sense to trade? Finally, I see that in the book that I mention at the beginning, a portfolio process is constructed by trading in the zero-coupon bonds and the money market via the following equation: where Δn,m\Delta_{n,m} is the number of zero-coupon bonds of maturity mm held by the investor between times nn and n+1n+1 . I understand that this portfolio process, properly discounted, is a martingale, and hence there can't be arbitrage when trading in the zero-coupon bonds and the money market. My final question is: 3) How is this formula related to the proof that Bn,m=E~n(1(1+rn)(1+rm1))B_{n,m} = \widetilde{E}_{n}\big(\frac{1}{(1+r_{n})·\dots·(1+r_{m-1})}\big) ? If some of my questions are not clear enough, please let me know. Thanks a lot!