Let <code>dp[x][t]</code> be the maximum points we can get from the subtree rooted at node <code>x</code> and the second operation has been used <code>t</code> times in its ancestors.

Note that the value of each <code>node <= 10⁴</code>, so when <code>t >= 14</code> <code>dp[x][t]</code> is always <code>0</code>.

General equation will be: <code>dp[x][t] = max((coins[x] >> t) - k + sigma(dp[y][t]), (coins[x] >> (t + 1)) + sigma(dp[y][t + 1]))</code> where nodes denoted by <code>y</code> in the sigma, are the direct children of node <code>x</code>.