🗒️ 👩🏻‍🤝‍👨🏼 ⚙️ Job Contest Results or Game of Thrones in the Battle of PS4 Ⓜ️ ⭕️ ⚗️

Those who are not the first day on Habré probably know how we at RUVDS love to arrange all kinds of activities. What we just did not do. They started the server in the clouds, made a collaboration with the creator of Duke Nyukem, even brewed their own admin beer and prepared hubraburgers, just not to mention. All these projects are born out of a small - simple idea. So here, it all started with the fact that I wanted to see how our readers migrated to a remote place, and how they got comfortable there. From the simple idea of “let's make a post where we will post pictures in comments”, the idea grew into a full-fledged competition , with a telegram bot, landing page and even a prize fund specially written for it. In this post, we will summarize the results of the competition, announce the winner and tell you what kind of payoff war we had to face and how we got out of this.

For several weeks they urged you to share photos of your workplaces and moderating the contest, we felt that while sitting at home behind the monitors, we were able to visit several hundred people. In total, more than 700 people from 211 cities from around the world took part in the competition. There were funny moments, for example, when a participant sent a photo from those places where the moderator lived.

The workplace of Mikhail working in Tenerife

Each photo you sent of the workplace was as individual as the people who work for them. Sometimes it was even interesting to fantasize about the personality of a person who manages to front-end a small laptop behind the screen or a person who is quietly working in the midst of a mountain of wires, microcircuits, and all that.

The workplace of Elijah, an electronic engineer from New Jersey.

But there were some unpleasant moments. Cheating votes for yourself, dirty tricks to the participants going to the top honestly. It will not be out of place to tell how they dealt with this.

The mechanics of the competition and the fight against cheating

Wrapping up is a classic issue of contests with likes. Usually they are held in VK, Instagram or Facebook, which do not like to give data to the developer and do not allow the organizers to edit the vote. Therefore, it was decided to use Telegram - immediately after sending the photo to the contest, the user was asked to subscribe to the channel and be aware of the news.

There were two ways to vote in the contest: by putting a Like on the site (with authorization via Telegram) or directly on the channel where photos of new participants were published.

Voting on the channel

We wrote a special bot that aggregated likes from the channel and site. In the self-written voting system, it was possible to track literally every like and not just ban those who win votes, but selectively remove likes from bots.

At the beginning of the contest, we did not particularly check the logs, naively believing that we were in our closed, friendly party for IT specialists. But a few days before the end of the contest, several participants suddenly got +1000 votes. And when we opened the logs to track the markup - the real meat began.

Intrigues, frauds, investigations

With the start of the struggle for the main prize of the PlayStation4, the likes began to wind up mercilessly. For an hour, players from the top ten began to appear 100-300 likes per hour. Then they took only three, who suddenly gained +1000 likes, looked at their logs and banned them, removing them from the competition.

This is what the log looks like with an obvious cheat: on the right column, you can see how many likes were added per minute. Even if a person has a popular channel in which he advertises himself, Elizabeth II can achieve such cohesion and influx only by making an appeal to the nation.

Realizing that there were a lot of markups, we double-checked everyone who won intermediate prizes - almost everyone had a lot of likes from the bots and had to be removed from the competition.

But we made a strategic mistake by talking about it on the channel. After that, some kind of madness began - each of the top ten participants added literally hundreds of obvious bot likes, all of them at once. It became obvious that someone was trying to massively substitute the players who pulled ahead. Having tracked on whom the furious wrapping ended, they put an asterisk in this mind on this participant - most likely, this is his business.

It was already impossible to ban participants, and we began to clean the vote, removing likes from bots. I had to track the logs for 2,000 lines and real votes could fall under the distribution. I had to create an algorithm for determining whether the bot liked or a person. To remove the bots, they made a special command in telegrams - it was enough to enter the command and insert unique IDs of likes that need to be deleted.

We read logs, calculate bots

In the framework of this contest, the log is a simple xls file in which all the data of a particular participant is indicated. Inside each log:

unique id of the set like
username in telegram like
his nickname in telegram
exact date and time like

The easiest screening is through date and time. If there is a clear pattern (for example, 11 likes per minute), then this is most likely a cheat and these voices can be safely removed.

A clear pattern is visible - exactly 10 per minute. This is a 100% wrap.

If a clear pattern is not visible, but all the same, likes were typed too quickly, I had to write to the participant and directly ask where and how he was promoting his post. If a player occupied the first place at that time (and it constantly changed when we cleaned the bots), then they could also ask for screenshots or a link to the call to vote in a chat or channel. Then we correlated the time of the post, the number of views and how many likes from that moment arrived. If the data did not converge, this became the cause of the strike.

One participant with a too suspicious jump in votes sent in confirmation of the screenshots of his Whatsapp, where he asked his friends in a personal vote. In the screenshot, people unsubscribed that they voted and wished them good luck. Messages "Like, good luck!" were dated on the 13th, but there was not a single vote in the file for the 13th - they all started on April 14th.

People massively unsubscribe that they like.

And in the logs there is not a single like during this time - it is quite possible that someone unsubscribed and did not vote, but so massively - it is unlikely. If you rely on Occam’s Razor, it’s just an attempt to fool the rest of the participants.

We told each participant why he was leaving the competition. The participant from the example above, after explaining the reasons, left 30 minutes and brought already new screenshots, where he wished good luck and promised to vote on April 14, but nevertheless dropped out of the game.

There were cases when a series of suspicious likes really turned out to be real. There were two such cases. For example, designer Vova from MentalGrowth is the author of the popular design channel .

He honestly gained votes by posting requests for pushing on his channel, but among real people, the voices of bots were wedged. Vova himself claims that he had nothing to do with it and the subscribers of his channel decided to help him like that by catching up with bots. And we wholeheartedly believe him - but one way or another, his record was cleared and, perhaps, real voices also got under the distribution. Therefore, we say hello to Vova and recommend subscribing to his channel.

Removed bots from Vova. My heart was bleeding, but what to do, the wrapping is a wrapping.

Another difficult case was with a participant Artem, who ultimately bypassed Vova by five votes and won PlayStation4. A clear pattern was not tracked in his file, but at the same time the likes came in large numbers and this was doubtful.

What if there is no pattern, but there is still doubt?

They asked him to show the posts how and where he was promoting his workplace. There were two points of voice jump in the logs, and their time coincided with posts on the channel. But the conversion was suspiciously high: out of 448 views, almost 200 votes. Then we selected 25 random voters, wrote them in a personal message and asked: did they like, to whom and how did they find this person. Bots never reply to messages. Almost all of the 25 responded who they voted for and where the call for help was published.

So, we came to the main point in this post: congratulations to Artem, senior software developer at Intel, with a victory!

We also publish those who scored the maximum number of votes in the overall classification

Vova, designer at MentalGrowth, Moscow

Elizabeth, designer at Jolocom, Berlin

Konstantin, Python Back-end developer at BSW, Tomsk

Alexander, analyst at NORBIT, Kazan

Danil, “Python strangler” and PM on freelance, Crimea

Boomburum, Head of User Relations at Habr, Moscow

We shake hands with the rest of the participants, including those who were mortally offended by the results - we did everything we could. We, as a whole RUVDS team, thank all the contest participants You are all bright and individual, thank you for the invitation to visit and see you soon. Take care of yourself!

Job Contest Results or Game of Thrones in the Battle of PS4

The mechanics of the competition and the fight against cheating

Intrigues, frauds, investigations

We read logs, calculate bots

More articles: