#include <bits/stdc++.h>
using namespace std;
double solve(int n, int k, vector<pair<int, int>> &a, int totalB) {
vector<vector<double>> dp(n + 1, vector<double>(n + 1, 1e9));
dp[0][0] = 0;
for (int i = 1; i <= n; i++) {
for (int As = 0; As <= min(i - 1, k - totalB); As++) {
int Bs = min(totalB, i - 1 - As);
if (As < k - totalB) {
double d = a[i].first;
dp[i][As + 1] = min(dp[i][As + 1], dp[i - 1][As] + d / (totalB + 1));
}
if (Bs < totalB) {
double d = a[i].second;
dp[i][As] = min(dp[i][As], dp[i - 1][As] + d / (Bs + 1));
}
else {
dp[i][As] = min(dp[i][As], dp[i - 1][As]);
}
}
}
return dp[n][k - totalB];
}
int main() {
ios::sync_with_stdio(false);
int n, k;
cin >> n >> k;
vector<pair<int, int>> a(n + 1, {0, 0});
for (int i = 1; i <= n; i++) {
cin >> a[i].second >> a[i].first;
if (a[i].first == -1) a[i].first = 1e9;
}
sort(a.begin(), a.end());
for (int i = 1; i <= n; i++) {
swap(a[i].first, a[i].second);
}
double ans = 1e9;
for (int i = 0; i <= k; i++) {
ans = min(ans, solve(n, k, a, i));
}
cout << setprecision(10) << ans << endl;
return 0;
}
