Contest duration discussion

It’s great that CG is taking feedback from community, but I was a bit puzzled the question of contest duration isn’t even brought up. Is there any particular reason contests have to be 10-14 days? Every other AI contest out there is usually at least 1 month (RAIC, MiniAICup, Halite, theaigames, …), and the first few major AI contests like Google AI Challenge lasted around 4 months!

I’m sure I’m not the only one with a busy schedule here. With the way it’s organized right now, I had to make a deliberate choice not to participate in BotG beyond the first few hours, because I knew the more complex ideas would require more time than I could set aside in such a narrow time frame. I’m not saying I would’ve won if I had more time, but I certainly would try at least.

Of course, being able to code up complex ideas bug-free in a time constrained environment is a challenge in itself, but I think this idea works better for 2-3 hour long bursts (see codeforces.com), not so much in week+ scenarios, where it’s more of a competition on who can postpone their responsibilities the most. I can’t help but wonder how many more people would take part and what kind of strategies would people think of if they had more time…

With a longer duration you get better competition, bigger participation, more exposure. Why not do it?

4 Likes

We used to ask this question in previous surveys (2 years ago) and the answers were divided (like the answers for ideal time between contests were in the survey I sent last week).
Some people prefer short contests because they can perhaps dedicate all of their time to it and they feel that they would be at a disadvantage in a month long contest. And others feel like you do.

So we stick to 10-day contests because we felt it’s the right balance. Also because the games are released in multi after a contest, so the competition can continue (granted, without any prize).

That being written, we offered all the creators to choose other durations (or other types of contest). Right before the release of BotG (or maybe just after, I don’t remember exactly), the creators actually asked to add at least 1 week to BotG. We refused as we had already announced it as a 10-day contest. Maybe we should have gone for it.

I would also argue that it’s a tad simpler in terms of communication to keep the same duration, because everyone is used to it. :blush:

1 Like

Didn’t CG also try really long durations with PR2?

For me, 10 days is fine.

before T.J. yeah :smiley:
What I heard it is that it was difficult to keep the momentum of players for a month. That explains also why we’re reluctant to go further than 10 days.

I remember CSB contest. It was on 2 weeks. The second week was very hard. As you said, keeping a momemtum during 2 weeks is pretty hard. After 1 complete week you got a hard time to have new ideas and your motiviation is broken.

If you take CSB as an example, during the second week, pb4 and Jeff06 (2 players in the final top 3) were playing with Neural Networks. They knew they won’t win with that but i suppose they had nothing more to do.

I disagree on the notion that momentum is lost with longer durations. I participated in Halite for 3 months, and having more time for it was a godsend. It was possible to test various ideas without having to throw out the majority of them because of lack of time, and made the meta develop in a much more interesting manner, as well as being able to absorb more complexity. If that competition had ended in 10 days, it would have been so much less interesting. I also think it makes it more fair for a lot more people, as it is much easier for someone to dedicate 10x more time than someone else over 10 days, but over a month+ it’s more difficult.

I definitely think it would be interesting for CG to try a longer duration for a contest in a modern setting, at least once, and see how it works out.

5 Likes

Yet that is what CG observed.

That’s not so obvious to me. Over any contest duration one could always assume that the others spent more time.
With longer durations the question of who has more time might turn into the question of who hasn’t given up yet.
In Halite 2 when I checked the leaderboard, the #2 had not submitted in the last 21 days, one could argue that you had a time advantage over him. Whether that is true or not I don’t know, but it also goes in favor of what CG observed: people stop playing. There was also a guy who was #1 for the first month and then just stopped playing.

2 Likes

Regardless of whether an ‘objective’ answer exists to the question or not, I still think @inoryy has a point: if you send a survey regarding the contest format, it would also be good to ask about how everyone feels about contest duration, and see if the sentiment still holds with previous polls. Perceptions and communities change over time. I would have liked to express my opinion about that in the survey.

Momentum is definitely lost with longer durations and it would be a shame if a contest “failed” because of that, since a lot of work goes into creating one. It’s ridiculous to reject the observed trend by cherry picking yourself as a counterexample and projecting your own particular obsessions onto others, but I digress…

I think an interesting idea to test the waters would be to hold a classical 10-day tournament and if the creators receive some valuable feedback during and after the contest, such as in the case of BotG, then the updated game could extend into a second much longer tournament.

1 Like

I like the 10 day contest format. We have longer “contests” in multis, so I believe the effort from the CG team should be on making the multis more engaging.

3 Likes

And it’s even more ridiculous to distort and twist my words to such an extent, by keeping projecting your own unpleasantness with the continuous negative bias you hold against me, for whatever reason I don’t quite care knowing about.

Hi @reCurse, I did some simple codes for Halite too. And my experience with this time frame was not good.
Let me explain why not. After 2 or 3 days building simple AIs, I realize that the context was for hard engaged developers, like you… When I saw your name at the first position, I thought that I had no chance to win… And I thought that after 3 months, only very good and very optimized algorithms could have a chance… (I’m a Java developer, so, I’m not a optimization fan).

I think that Halite was a challenge to find the real best developer. And I’m sure that I’m not a good one, I don’t have any chances at a 3 month contest…
Just think about the BotG, wiht more one day, probably the first one would be changed, more 3 months, probably, a real good code would be won.

My point is: It is like a small soccer team playing against a big one. The small can win one game, but if you play 10 games, probably the big one will win much more… In a 10 days contest, you have the adrenaline, and you can see what noone saw, and exploit it… And maybe win, or get a good position… Because of this adrenaline, this fast meta change, this time constrain (There is no way to build a perfect code in 10 days) I like 10 days contests…

I’m here for fun, so, I think that a 10 days contests is much more fun… I think that 3 month contests is for companies that are hunting really good developers… I don’t know if it is the CG objective…

But if you really want to prove that you are the best developer (And you are a very good one), just invest 3 months doing the perfect code to the Multiplayer version.

3 Likes

Hi @BrunoFelthes. I really should have precised that I was not advocating for 3 months contests at CG at all, so sorry for the misunderstanding. I was taking Halite as an easy example because of how different the context was compared to CG, my recent involvement and very positive experience with it, as well as participating actively in its community and exchanging feedback with other players.

I just think it’s unfortunate that anything longer than 10 days seem instinctively rejected right off the bat as something that doesn’t work, when there are plenty of examples where it does work very well like @inoryy mentioned, and after talking with many others, I know I am definitely not alone with this sentiment. I also seem to recall positive feedback from some on the duration of TAC which lasted 2 weeks.

I am also not advocating against 10 days contests either, I understand (and truly share!) the thrill of the time pressure and wild evolution of a short contest like that. I definitely want to keep seeing those at CG.

However I also think it would be a good idea to try out a contest for once with a longer duration, but definitely not 3 months. Anywhere between 2 and 4 weeks would be nice to try. I can’t believe I have to write this but following some comments: it is of course my personal opinion based on subjective preferences that I would like to voice on the matter.

1 Like

If loss of momentum is such a problem then how does almost every other AI contest deals with it?
Sure, there are outliers like @BrunoFelthes mentioned, but as far as I can tell Halite as a whole was quite active until the last minute. I know that RAIC is also quite active for the duration. And, while not directly comparable, kaggle ML contests last 2 months at the minimum and are always active throughout the duration. But then again I also disagree that lack of submission equals lack of interest. Maybe the #2 guy on Halite devoted as much time as @reCurse, only he did it during the first month of the contest, when it was more convenient for him?

I’m not asking for 4 month contests either, maybe try 2-3 weeks for the next contest and see how it goes?

4 Likes

As far as i know, players on Halite (or RAIC) and codingame are not the same. There’s some people who play on both formats, but how many are they ?

Everytime someone mention a 1 month+ contest on the chat we can see the same reactions. “It’s too long”, “No time for that”, “There’s no way i can code during X months”. We can’t debate on preferences and tastes. I can understand that some people like X months contests.

But this is not the way of codingame. And the codingame population have definitively a taste for 10 days contest (it’s pretty obvious since they are here :smiley: ). My point of view is that if people want to make a X months contest, they already can and i won’t judge them. But you can’t ask to make a X months contest on Codingame and not expecting limitless rage from the community.

Take a step back a little. It’s like if tomorrow i go on Halite and i ask for a 10 days contest. How do you think the Halite community will react ?

1 Like

This sounds interesting and more manageable imo.

So participants that can only do the initial 10 days still code under the same conditions. Once the contest is over you can go for stage two if you want to …

What? Why are you acting like we’re demanding all contests on CG take X months, when the whole point that started this thread is it would have been good to survey opinions about that as well, and the only suggestions here so far is (way) below 1 month for once. It’s unfortunate to have a kneejerk dismissal for voicing a different suggestion/preference, especially when I have also seen people on CG that would prefer longer duration.

1 Like

As a CodinGamer, I do like the fact that a contest duration should be updated accordingly to the game itself and not try to fit a 3-week contest in 10 days just because it has to be 10 days.

Even if the community want a 3 month contests, I do believe that the game should suit this duration.


As a CG staff member, we’re thinking about ‘checkpoint’ to enable people with different timeframe to compete somehow and still win some prizes (Checkpoint 4h, 24h, 48h, 1 week, 3 week…) even if they don’t win the whole contest. A longer duration will enable us to close the contest, run a final batch, make a screenshot of the leaderboard et re-open the contest for a longer duration.

I’m not sure I’m understandable. :stuck_out_tongue: (But it’s still in debate)

11 Likes

Fair enough

+1
Great Idea!