Oh okay, Thank you for taking the time to answer. I understand the problem that my code faced now a bit more. I guess long double was not enough to remedy this, probably is the same as double in my operating system (Edit: I checked, long double is 18 and double is 15 for me).

I gained new insights because of my mistake. Thank you for some of them! Now I am just confused about how even <= works for doubles and how many digits does it compare, because sometimes 0.0 does equal 0.0 but sometimes not, I understand that it did not matter whether I used double or long double, as the important digits were accurate, <= would read the inaccurate digits too, I think but I am not sure any more.

Interesting. I used epsilon then it gave me correct answers. What about long double precision, I used long doubles. I defined double to be long double.

#include<bits/stdc++.h>
using namespace std;
#define int long long int
#define double long double
void IO(string s = "") {
    ios_base::sync_with_stdio(0);
    cin.tie(0);
    if (!s.empty()) {
        freopen((s + ".in").c_str(), "r", stdin);
        freopen((s + ".out").c_str(), "w", stdout);
    }
}
void solve() {
    int n;
    cin >> n;
    double speed;
    vector<pair<double, double>> segments(n);
    for (auto &[x, y] : segments) {
        cin >> x >> y;
    }
    speed = 1.0L/segments[0].second;
    for (int i = 1; i < n; i++) {
        speed = max(speed, (i+1.0L)/segments[i].second);
    }
    bool f = 1;
    for (int i = 0; i < n; i++) {
        if (!(segments[i].first <= (i+1.0L)/speed && (i+1.0L)/speed <= segments[i].second)) {
            f = 0;
            break;
        }
    }
    if (f) {
        cout << fixed << setprecision(7) << speed << endl;
    } else {
        cout << -1 << endl;
    }
}
int32_t main() {
    IO("A");
    int t;
    cin >> t;
    for (int i = 1; i <= t; i++) {
        cout << "Case #" << i << ": ";
        solve();
    }
}

It was my check that had the problem, "(i+1.0L)/speed <= segments[i].second)" was the condition where 36.000000 <= 36.000000 was false. I understand that I could have just kept the information of both minimum speeds and maximum speeds and used them but because the time complexity allowed it, I thought that my check would have been better and well it came first to mind than the minimum and maximum speed check. Oh, how wrong I was. It is a bit scuffed that I took ints in as doubles but that should not be that problematic?

I was using only long doubles. Most times that is enough. I was aware of the epsilon fix but was too lazy and did not realise I had a precision problem because it was in the final testcase and delayed feedback of that. I guess now I have an actual example to fallback on why to use epsilon.

I had no time to do the contest fully and I had a floating point inaccuracy in the first problem final submit testcase (36.000000 <= 36.000000 was false, I thought setprecision remedied this but I was wrong). From rank 2200 to 5500. Oof. I forgot that the feedback to the submission is given after the contest. If only, the format was like Codeforces. Eh I guess better luck next year.

I jinxed it. This was my first and last attempt. I became quite ill on the first contest day. One of the most unpleasant experiences in my life has been this moment. I am kind of depressed now but eh the second day went better. I am still feeling ill. I am ashamed, I could have done so much better. :(

I just hope that everything goes well and that I get a medal.

It is quite funny to see that the former CEO of Alameda Research is on that table. I guess committing massive fraud is an achievement as well. Though probably the table was made before.

My prediction was correct. I understand why I am getting downvotes as this was a pessimistic prediction but sadly, it is accurate. This was my own opinion and prediction. It is sad to see that the results under my country's own name are like this.

Estonia will probably not get anything, they are rather unmotivated and show no development since March.

Will there be an editorial? If there already is one, where?

It seems that the unordered_set is running in its worst time complexity O(n) instead of O(1) so the algorithm is possibly O(n^2), I reckon a lot of hash collisions. If you just change the unordered set to a normal set instead which would run in O(n log n), it is a lot faster. https://codeforces.me/contest/1760/submission/206029716 . I hope someone smarter can tell you better why it is specifically slower here.

You can solve it using prefix sums and going through all ways.

Sorry, I answered the Lex person who asked about the 1st problem in today's competition. I got the staring contest 50 points. Was pretty perplexed about it, how to do it in n + 25 queries. Sorry, I should have not answered him.

When will the standings be shown?

I got 78/100 by just DFSing. I chose all nodes that had only 1 edge as starting nodes. For the second task where there is an intersection, I did BFS to find the most distant node from that intersection and then used DFS from that node but that only works when there is only one intersection.

Hello.

Can the team leader be from a team of three students? For example, if I decide to compete, can I be the team leader, or does it have to be someone who does not participate, which would complicate this further?

deleted

Did they change tests for C. I had done a typo in my code, a <= was == but it went through during the contest and it was accepted, it was accepted after the competition as well, but just today it turned into being not accepted. From a positive rating change to a negative rating change, nice.