Bro look closely at the year, it's 14 years ago.

I always use cin and cout. It really happened for me that one time I submitted with C++17 and it got TLE mid contest while the exact same code got AC using C++23. After that, I started using C++23 but do you think I should change to C++17?

Even more clickbait T-T

The revision 1:

PS: If the image couldn't load, click here.

So he isn't biased at all? Don't apply double standards. If you truly have no bias, then you should also be suspicious of jamesben in plagues's post. That account was also registered only 2 weeks ago, which just happens to coincide with the start date of the Huawei Challenge. Why don't you mention that?

Link to my comment.

As I said, the three accounts TemperanceAlouette6042, MatildaDamaris6674, and PiaElka37401 share several unusual similarities. All three usernames appear to follow the pattern of algorithmically generated names, and all three accounts were created around the same time. This makes me wonder whether they could be alt accounts of someone who dislikes plagues.

Interestingly, their registration time also coincides with the period when plagues started pointing out some unusual coincidences regarding the top 18 participants of the Huawei Challenge.

Regarding TemperanceAlouette6042's claim that I'm applying a double standard: jamesben's username is actually formatted differently from those of TemperanceAlouette6042, MatildaDamaris6674, and PiaElka37401. The three accounts I mentioned all follow a remarkably similar naming pattern, whereas jamesben's username does not.

Note: These are merely my observations and suspicions; I'm not confirming or accusing anyone of anything.

got dozens of threats to rape my mother from brave internet warriors?

orz Um_nik for bravely protecting your mom from those warriors.

I don't deny that a lot of plagues's evidence consists of fairly strong signals like "the scores are exactly the same." Information like that is, of course, quite convincing. And as far as I know, some of the people he accused did indeed violate the contest rules.

However, a lot of plagues's other evidence is based only on weak correlations, such as "the scores are very close," "the running times are the same," or "the scores follow similar trends."

His blog feels like it mixes some far-fetched claims in with a number of correct facts, making it look as if everything he says must be true.

In any case, publicly naming and shaming people in this kind of post based on weakly correlated evidence is absolutely irresponsible. The reason is that, apart from the two people involved and the organizers, nobody else can actually verify what happened. Most people are simply following the crowd.

With ordinary CF submission records, on the other hand, at least everyone can inspect the evidence and reach their own conclusions. So I don't think it is always unreasonable to suspect someone of cheating when the relevant source code is publicly available to everyone.

I think what you're saying makes a lot of sense. However, it seems pretty clear that most of the evidence he found consists of scores that are merely close to each other rather than exactly identical.

I noticed that, for many people, the majority of their submissions are concentrated within a range of no more than 1,000 points, while the publicly displayed scores have a precision of 0.001. In other words, there are roughly 1,000,000 possible displayed score values within such a range, and many people have made far more than 200 submissions.

So we can consider the following simple probability model. Suppose we uniformly and randomly choose 200 integers from 1 to 1,000,000 to form a set A, and independently do the same to form a set B. What is the probability that there exists a pair of numbers a in A and b in B such that |a — b| <= 50?

There are 200 × 200 = 40,000 possible cross-group pairs. For any particular pair, the probability that their difference is at most 50 is approximately 101 / 1,000,000. This gives an expected number of about 4.04 such close pairs, and the probability that at least one such pair exists is approximately 98.24%.

Of course, I know that this model differs considerably from the actual situation. In particular, scores in a heuristic problem are obviously not uniformly distributed. In reality, submissions from different contestants may be concentrated in similar score ranges because they use similar general approaches, start from the same baseline, or simply converge toward similar levels of performance. If anything, this clustering could make close scores even more common than the simplified uniform model suggests.

Therefore, I am not claiming that this calculation proves that any particular pair of similar scores is merely a coincidence. My point is only that, when two contestants have hundreds of submissions, the mere existence of a few pairs of scores that differ by something like 0.01, 0.02, or 0.05 is not statistically surprising and, by itself, seems like fairly weak evidence of code sharing.

I do agree that exactly identical high-precision scores are a different matter and should be considered separately. Under the same simplified model, exact matches would be much less common than merely close scores. So I think it is important not to treat “very close scores” and “exactly the same score” as if they carried the same evidential weight.