What's wrong with my solution for Devu and a Journey on Metro?

double solve(int n, int k, List<Edge>[] g) {
	long[] d = new long[n]; // d[v] = distance from 0 to v
	dfs(0, -1, g, d, 0);    // calculate distance using dfs

	double[][] mc = new double[n][n];

	for (int from = 0; from < n; from++) {
		double pr = 1.0 / g[from].size();
		for (Edge e : g[from]) {
			mc[from][e.to] = pr;
		}
	}

	double[] p = pow(mc, k)[0]; // some illegal Markov chain magic
	double ans = 0;
	for (int i = 0; i < n; i++) {
		ans += d[i] * p[i]; // expected value
	}

	return ans;
}

Or this is the right way but i have some mistakes in another part of code? May be problem with precision.

I've got some questions about editorial. Please, can someone prove me: Why minimal number of swaps <= 1 and why it's optimal to do swaps which decrease the string size as less as possible?

There seem to be problems with grading of Devu and a Journey on Metro, several people have complained that outputs that are (within precision margin) the same as expected ones are judged as incorrect.