This is a graph theory problem. The graph is a DAG (Directed Acyclic Graph) and we can use dynamic programming to find probabilities.

$dp[i]$ is the probability of a log floating to lake $i$. In the edge from $i$ to $j$, we will call $j$ the child of $i$. For each child of $i$, add ${\displaystyle \frac{dp[i]}{\text{total rivers leading out}}}$ to its probability ($dp[j]$), as we have an equal probability of selecting a river to flow to.

At the end, output the value for all lakes which have no children. Be careful to output at least 9 digits. Fast input and output methods should be used if possible.