A full submit is between 100 and 200 matches depending on your league. I think that cgSpunk benchmarks put a bigger strain on CG servers than a submit.
But then the AI would appear in that personās withdrawn AI list (his past agentIds) and I could play against it in the IDE?
There would still be the problem of AIs that look at the time to stop working one hour after the submit. But CG could kill this by always starting the Virtual Machine at a certain time.
Well the strain on the servers because of cgSpunk is clearly caused by the priority system that Maxime was talking about. If something can be done on that side, it wonāt matter that you run 500 games in cgSpunk anymore since you will be on the lowest level of priority (well ok, it will still matter but then, the system can just delay the matches scheduled on a certain priority and all will be good on that side). Anyway, I guess thatās another problem.
Not the play number but the number of opponent changes, much softer restriction. Just to make clear that the solitary IDE testing is different from an arena release. This would give a nudge to publish a version that seem to work (but itād be perfectly OK not to publish if you want to bet that your bot will do well anyway). IMHO thatās a setting that balances different interests.
PS. There is no such thing as āhonestyā, just abiding by the rules. And we donāt have to define what the āIA hidingā is. Just make some rules that encourage a kind of behavior that make CG fun to play.
Hello all,
We do not forbid AI Hiding but we donāt promote it. As written above, there is no easy way to prevent it anyway. Actually, as pointed out by reCurse and the great article he shared, itās not even sure that AI hiding has to be prevented. Weāll remain attentive to trends about this practice in future contests. I hope that this clarifies CG stance.
Personally, I wouldnāt do it. Play to win, yes, but play to win with glory 
Keep Calm
and
Submit Your IA
10 Likes
So as probably everyone in top10 of CotC contest already noticed, thereās at least one top3 contender who is currently blatantly hiding his true AI, namely Recar and Iām going to take that as an opportunity to bump this thread and give my 2 cents on the topic.
In any sport youāll find examples of plays that while within the rules (borderline), are still looked down upon by the community. Whether itās faking injuries in soccer, throwing extra punches in boxing or abusing specifics of an FPS game engine.
This is important to keep in mind because no matter how much organizers try, somebody will always find a loophole and so itās up to the community to step up and call people out on their anti-competitive behavior.
Iāve discussed this situation with Recar in chat first and have asked multiple times for him to state his case here, but he declined (wonder why 
) so Iām going to paraphrase for him.
His claim is that he sees nothing wrong with it and is doing it because he just wants to win very bad and if people are willing to show their game changing features early on then they just donāt want to win as much as him. Aside from that, he dodged any question on fairness and asymmetry of testing opportunities, including the most obvious āWhat if everybody starts doing it?ā.
Since itās not breaking any rules he is of course entitled to uphold that opinion just as I am entitled to consider it short-sighted, anti-competitive and, most importantly, anti-fun and publicly call him out on his behavior.
I suppose everyone already knew why Recar hide his AI
The problem is that CodinGame response is always the same:
Well, we have some easy ways. Like adding a feature to fight against an old agentId (And yes it is simple because it is possible right know if you know how to tweak the IDE).
But CodinGame never have time for this kind of things So i suppose weāll have to deal with it.
But thatās my point - there will always be a way around the rules.
Allow agent id => Timeout the bot after a certain time.
Start the bot with locked time setting => Skip submitting altogether and just test in IDE.
Isnāt that the point of a contest? Everybody does his thing and at the end we look at who did it best. Everybody tries to keep his best idea for himself, to win the contest. Doesnāt that keep the suspense till the end? Isnāt that what makes it exciting?
Also whatās with the people who submit nothing till the very end? Isnāt that much worse? They donāt even show you a score while the contest is running. Do you want to call them out too?
Nobody should have to defend themselves for staying within the rules.
2 Likes
Hear, hear! At the risk of repeating myself, it is not the playerās fault for playing within the rules, but the organization/community for allowing it or not enforcing it. We have already discussed a very simple plan of action, yet it is still not in place! Allow me to quote myself:
Number 1 and 2 can be already done by the community, namely by sharing agent ids in forum/chat (no website even needed though itās ideal), and exposing the agent ID feature in CG spunk (should be relatively easy to do). Number 3 is probably a tweak on the backend that would take a few hours to do max.
So why isnāt this already in place? If the issue is so important, why isnāt CG/the community putting the (relatively small) amount of work to make it less of a deal? Those are the real questions, and will yield a better answer than shaming participants for not playing within arbitrary/unwritten rules.
No offense to anyone.
EDIT: Before anyone points it out, I realize I am contradicting myself on one point. I said before āit would be a lot of effort to do that just to prevent AI hidingā, but in fact after thinking about it, it really isnāt that much effort to put in place, if the issue is that much of a deal.
How does that prevent keeping your AI in the dark and running tests via CG Spunk against people that do submit their best?
Measures have already been put into place to limit the use of CG Spunk (once every 15 seconds). The idea is not to make AI hiding completely impossible (as it will be hard or detrimential to other players), but raise its cost so itās not as good of an option anymore. The harder it is to get results out of it, the less attractive it will be.
So no valid solution that directly attacks the problem has been found. In the end best bot is still the one who deserves to win.
Last minute fixes can still be considered hiding.
Finishing your offline training late in a contest might also be confused with hiding.
You could get replay data and always avoid the IDE until last moment. Every contest becomes a sprint with long wait time. Tampering with the IDE will only mess with the newbie bot hiders.
If you dislike AI hiding what you can do imo is contribute with your best possible bot. I know it sounds counter intuitive, but if the top can improve against each other fast enough it could make hiding pointless. ācouldā and not would, because the hider can still be the better player or simply have a tiny bit of extra insight nobody else does.
4 Likes
Alright, time for an unpopular opinion.
First point:
AI āhidingā is an inherent part of the game.
Someone can code a bot completely offline and improve solely through self-play.
Second point:
Itās completely fair. Everyone can do it and everyone has access to the same
information that is present in the arena.
Third point:
Itās a competition. People are trying to win, right? To me the fun in a contest is
seeing other people go above and beyond to deliver the greatest bot. When I saw
that reCurse submitted on the Amadeus challenge I found that very exciting.
When RoboStac almost beat me on the Thales competition by circumventing the meta, I thought that was
amazing. I have great respect for other competitors.
What I however donāt respect are people who voluntarily participate in a competition and then are
upset when they see that other people are competitive. I donāt mean to offend anyone,
but thatās whatās happening from my point of view.
4 Likes
There have been multiple discussions on this very topic, all of which disagree with you on that point.
The only people it is fair to are the one who also hide, it is unfair to the 99% other players who keep their best bot in the arena during the whole contest.
You can train against them, they canāt train against you. How can you think this is fair ?
I understand why you do that as it provides a real advantage and youāre clearly here to win (I respect that), but thatās a really lame way to win imo (cf. my sucker punch reference in the PM topic).
Also, donāt forget that the prerequisite for you to win contest by hiding is that some people among the top play the ālegitā game and push their AI so you can train against them.
āDo the sameā will you say. This is clearly not what this platform has been designed for. Imagine if everyone does thisā¦ How sad.
As for the solution, itās already been proposed to do the same as other platform, make every IDE play public, so that if someone train against your bot in the arena, you can watch the replays.
EDIT : Also crying about the unfairness of the multi ranking on CSB while abusing the very same system in contests is kind of funny imo
EDIT 2 : To clarify : If your definition of fair is that everyone can do it, then yes itās fair (my definition differs a bit).
But unless CG fixes the issue, the fact that hiding becomes more and more common is an incentive to hide for everyone.
As a result, contests will become offline-self-training contests and the level of the competition will significantly drop, which is the opposite of what you seem to cherish.
Iām not blaming you, Iām just sad that youāre promoting something that could endanger the quality of contests as we currently know them, as CG will most likely never fix this issue.
8 Likes
Heās not promoting it though, simply stating why it is the most competitive option on CG. The problem you describe is not his problem, and if CG does nothing about it, itās on them.
3 Likes
Winning two contests with AI hiding is just proving to everyone that this an easier/legit alternative way to achieve victory.
I call that promotion.
1 Like
I call that winning.
4 Likes
winning++;
5 Likes
My condolences.
Last submit: 4/11/21 11:58 PM
You had a whole minute to CGBenchmark it.
1 Like