Southeastern European Regional Contest of ACM ICPC World Finals took place on 15th of October. However, there were a few major problems during the competition caused by some decisions of the Romanian judges. Here are some key points that led to absolutely unfair and unpredictable results:
1). Although it was declared that contest will be held in multisite mode testing of all the solutions was done on Bucharest PC^2 server. In spite of all requests on people organizing contests in Vinnitsa tests were never sent there. During the contest teams from Ukrainian site were submitting their solutions not directly to the main server, but using an additional online server of PC^2.
2). In the interval from 20th to 30th minute of the competition, server in Bucharest was rebooted and as a result all information about submits before that time was lost. It made a significant impact on penalties of many competitors.
3). During the contest participants had almost no ability to observe the results. All the teams in standings were shown as "TeamXX", where XX is the team number. The monitor was also refreshing very rarely.
4). During the contest some of the problems were rejudged. The problem "J" was rejudged almost at the end of the contest and problem "E" was rejudged only after the contest, while during the contest the clarification was sent that some tests do not fit the input format specification and teams had to deal with it by theirselves (instead of fixing wrong tests and making a rejudge at the moment it was found out). Moreover, some of the submits were judged long after they were sent. As an example, the team of Kharkiv National University had to wait more than 90 minutes before receiving the verdict for their submission of problem "C". It happened only after the direct question to the jury about why the problem is still in queue for judging.
I should also mention that the problem "I" was almost exact copy of a problem from SEERC 2010 contest, and the problem "J" was seen before on one of Moscow school olympiads. Apart from that many problem statements were prepared badly and had ambiguities in statements.
5). During the contest server was not available for significant amounts of time. Moreover, about 4 hours after the restart of the contest (about 4:30 after statements were distributed) the connection between servers in Vinnitsa and Bucharest was lost and as a result the solutions from teams participating in Vinnitsa site were not sent for judging to the server in Bucharest. During the contest those solutions were collected by organizers in Vinnitsa and later sent to Bucharest but the jury there refused to test those solutions even though it was a force-majeure situation. As you can see, the contest for participants in Vinnitsa appeared to be unexpectedly shorter.
As a result of the decisions of Bucharest side of the jury the results of SEERC-2011 can not be treated as fair because of all the problems listed above and the situation described in paragraph 5, in particular. All of that led to the situation in which teams were not in equal conditions. After such violations any regular competition would be at least considered unrated or the results would be split. (In this case the latter is impossible because it was a contest for one region).
I should also note, that it is impossible to recover all the penalty times, beacuse of loss of the data.
1). Although it was declared that contest will be held in multisite mode testing of all the solutions was done on Bucharest PC^2 server. In spite of all requests on people organizing contests in Vinnitsa tests were never sent there. During the contest teams from Ukrainian site were submitting their solutions not directly to the main server, but using an additional online server of PC^2.
2). In the interval from 20th to 30th minute of the competition, server in Bucharest was rebooted and as a result all information about submits before that time was lost. It made a significant impact on penalties of many competitors.
3). During the contest participants had almost no ability to observe the results. All the teams in standings were shown as "TeamXX", where XX is the team number. The monitor was also refreshing very rarely.
4). During the contest some of the problems were rejudged. The problem "J" was rejudged almost at the end of the contest and problem "E" was rejudged only after the contest, while during the contest the clarification was sent that some tests do not fit the input format specification and teams had to deal with it by theirselves (instead of fixing wrong tests and making a rejudge at the moment it was found out). Moreover, some of the submits were judged long after they were sent. As an example, the team of Kharkiv National University had to wait more than 90 minutes before receiving the verdict for their submission of problem "C". It happened only after the direct question to the jury about why the problem is still in queue for judging.
I should also mention that the problem "I" was almost exact copy of a problem from SEERC 2010 contest, and the problem "J" was seen before on one of Moscow school olympiads. Apart from that many problem statements were prepared badly and had ambiguities in statements.
5). During the contest server was not available for significant amounts of time. Moreover, about 4 hours after the restart of the contest (about 4:30 after statements were distributed) the connection between servers in Vinnitsa and Bucharest was lost and as a result the solutions from teams participating in Vinnitsa site were not sent for judging to the server in Bucharest. During the contest those solutions were collected by organizers in Vinnitsa and later sent to Bucharest but the jury there refused to test those solutions even though it was a force-majeure situation. As you can see, the contest for participants in Vinnitsa appeared to be unexpectedly shorter.
As a result of the decisions of Bucharest side of the jury the results of SEERC-2011 can not be treated as fair because of all the problems listed above and the situation described in paragraph 5, in particular. All of that led to the situation in which teams were not in equal conditions. After such violations any regular competition would be at least considered unrated or the results would be split. (In this case the latter is impossible because it was a contest for one region).
I should also note, that it is impossible to recover all the penalty times, beacuse of loss of the data.
Date: 2011/10/27
Subject: SEERC Appeals
These included :
(1) rolling back to an earlier version of PC^2 prior to the contest.
Именно переход на непроверенную в режиме multi-site более раннюю версию системы и привёл к многочисленным сбоям.
(2) allowing contestants at the VNTU site to continue working for an additional 30 minutes after submissions at the UPB site were halted
Вопрос к участникам соревнований в Виннице.
Верно ли это по факту? То есть действительно ли после того, как время контеста истекло, участники оставались на своих местах в течение получаса? Иначе говоря, через какое время после начала контеста были собраны решения - через 5 часов или через 5-30?
Если первое, то ни о каком продлении контеста речи не шло, все решения были сданы вовремя, а время организаторы добавляли сами себе для решения возникших технических проблем. В этом случае руководство ACM ICPC было введено в заблуждение румынской стороной.
(3) ending of the contest due to poor performance of the VNTU server.
При том, что сервер в ВНТУ изначально был "промежуточным звеном" и не мог тестировать задачи (кстати, вопрос: а какова в таком случае была его функция?), в данном случае существовало намного более адекватное решение: протестировать присланные организаторами в Виннице посылки, после чего не учитывать штрафное время вообще, то есть распределять места по количеству сданных задач, по аналогии с ситуацией в верхней части таблицы NEERC Northern Subregional 2008. Учитывая "пересдачу" задач на старте, учёт штрафного времени вообще кажется несколько странным.
Итого как минимум два из трёх принятых решений (первое и третье) существенно исказили результаты SEERC. Что же касается второго решения - вопрос в том, действительно ли оно было принято в такой форме (то есть действительно ли участники писали 5 часов 30 минут). Если да, то даже чисто теоретически вряд ли можно рассчитывать на что-то большее, чем предложенный вариант.
С учётом всей имеющейся информации (как открытой, так и закрытой), предложенное организаторами решение будет относительно адекватным только в случае, если к нему будет приложено решение о расширении квоты SEERC до 6 команд.
(кстати, это английская ветка).
Есть мнение, что не был разъяснен следующий факт (поправьте меня, если что): в последние полчаса соревнования решения были ОТПРАВЛЕНЫ на проверку на стороне Винницы, однако не были проверены. Это единственное, на что вообще здесь можно подавать апелляцию. (Поскольку обязанностью Жюри по тому самому пункту является проверка любого 'run'.)
Кроме того, видимо, не было разъяснено, что сервер в Виннице только перенаправлял решения на проверку, а фактически проверяющим не являлся.
Дополнение: приведённая ниже цитата показывает, что время на 30 минут увеличено _не_было_: сам prof. Tapus признаёт, что увеличить время контеста на 5-10 минут не удалось, и ещё 20 минут шла борьба с сервером - то есть получается, что для участников ничего увеличено не было.
Заметим, что это следует из ответа румынской стороны:
With accept from contest director (prof. Nicolae Țăpuș) the decision was to extend the contest for 5 to 10 minutes in order for the VNTU competitors to submit their problems. However, due to now heavy load on the servers, it was not possible to add that time. After 20-30 minutes of struggling to get everything working on the servers, the decision was to end the contest and leave things as they are.