One can define a single$-$agent version of fictitious play for the learning from expert advice problem as

follows: At each time step $t,$ the agent selects the action $i^t$ based on the following rule:

$Re{g}_{FP}(T){\displaystyle \underset{t=1}{\overset{T}{}}}1[i_t{i}_{t+1}]i^{t}inargmi{n}_{\text{i}\text{i}\text{n}[n]}{\displaystyle \underset{?}{\overset{?}{}}}$