#pragma GCC optimize("Ofast,unroll-loops")
#include <bits/stdc++.h>
#define ll long long
#define pii pair<int, int>
#define pll pair<ll, ll>
#define F first
#define S second
using namespace std;
ll N, K, mxC;
double ans = 1e15, dp[505][505], res[505];
pll A[505];
double calc(int i)
{
for (int k = i; k >= 0; k--)
for (int l = K - i; l >= 0; l--)
dp[k][l] = 1e15;
dp[0][0] = 0;
for (int j = 1; j <= N; j++)
for (int k = i; k >= 0; k--)
for (int l = K - i; l >= 0; l--)
{
if (k > 0)
dp[k][l] = min(dp[k][l], dp[k - 1][l] + double(A[j].S) / double(k));
if (l > 0)
dp[k][l] = min(dp[k][l], dp[k][l - 1] + double(A[j].F) / double(i + 1));
}
return dp[i][K - i];
}
signed main()
{
cin >> N >> K;
mxC = N;
for (int i = 1; i <= N; i++)
{
cin >> A[i].F >> A[i].S;
if (A[i].S == -1)
mxC--, A[i].S = 1e15;
}
sort(A + 1, A + 1 + N, [&](pll p1, pll p2)
{ return pll{p1.S, p1.F} < pll{p2.S, p2.F}; });
int L = 0, R = min(K, mxC);
int lmid = floor(L * 0.618 + R * 0.382), rmid = floor(L * 0.382 + R * 0.618);
while (R - L > 9)
{
double lans = calc(lmid), rans = calc(rmid);
if (lans <= rans)
{
R = rmid;
rmid = lmid;
lmid = floor(L * 0.618 + R * 0.382);
}
else
{
L = lmid;
lmid = rmid;
rmid = floor(L * 0.382 + R * 0.618);
}
}
for (int i = L; i <= R; i++)
ans = min(ans, calc(i));
cout << fixed << setprecision(10) << ans << '\n';
}
