Entelect Challenge

Entelect has released their new yearly challenge here

It has a new format compared to previous years for beginners and advanced programmers, as well as a GUI challenge for designers. What doy uthink of the this new pool structure? The theme is also quite nicely game orientated, and could be an interesting experiment/learning opportunity.

Just wanted to get some feedback and ideas from MGSA :) I am going to enter the beginner, as even though I am a programmer, I am a beginner in most repsects of the field and particularly in AI. I think I am mainly in it for the learning opportunity. Anyone going to enter? Anyone entered before?

Comments

  • I'm always a little leery of these things when Entellect puts them out. The games never strike me as overly fair and I'm not really sure how you're supposed to really offer critique because they always seem to go "Oh well, it's not supposed to be super serious" when real problems are raised. Yes, sometimes they change rules, so the hope is that it'll always get better next year ;)

    Only had a quick look at the rules at the moment and I'm not convinced this isn't a solved space - especially if the board doesn't change at all between runs... Maybe the pills are randomly positioned, but they haven't said anything about that and that's a big thing that needs to be stated up front. There are also no rules about the poison pills that I could find, so who know how those are created.

    Feels like the first player to move gets a huge advantage, seeing as they can control corner space further from their spawning point, meaning that the player that moves second has less potential territory that could be covered... I'm expecting bots to appear that start using dying strategically as a repositioning device. But yeah, seems like an okay algorithm could pretty easily solve this space entirely.
  • edited
    BTW:
    Each player is given a single poison pill that may be dropped at any point in the game when the player makes a move.
    The poison pill can only be dropped in an empty space, it cannot be dropped on top of a pill.
    A player may only drop a poison pill if there exists one in its inventory.
    I think they need to give the first move alternatively so that both players have the same chance at starting first. First turn advantage is huge when permutations aren't *that* big.
  • dislekcia said:
    I'm always a little leery of these things when Entellect puts them out. The games never strike me as overly fair and I'm not really sure how you're supposed to really offer critique because they always seem to go "Oh well, it's not supposed to be super serious" when real problems are raised. Yes, sometimes they change rules, so the hope is that it'll always get better next year ;)
    I definitely agree to this. It does appear that this they are trying to rectify exactly this issue from previous years, by putting in 2 pools for beginners and advanced entries, they also have opened up the test harness from the start for scrutiny. So more transparency and much, much better explanation of requirements and outcomes, compared to previous years. Will see if it is for the better then?
    dislekcia said:
    Only had a quick look at the rules at the moment and I'm not convinced this isn't a solved space - especially if the board doesn't change at all between runs... Maybe the pills are randomly positioned, but they haven't said anything about that and that's a big thing that needs to be stated up front. There are also no rules about the poison pills that I could find, so who know how those are created.
    There do seem to be plenty of solutions to Pacman mazes out there, with pros and cons. They suffered this issue last year with Battle Tanks where they really didn't explain anything thoroughly and only clarrified properly nearly a week or 2 before the first finals round.
    Based on their wording and provision of an example state and also on the previous comp, the maze appears to be slight random variations on the standard Pacman style of maze.
    Based on the example a maze is filled with pills and so navigating the maze itself to get the most pills is the focus.
    However, it sould be nice if they were far more explicit regarding the game.
    dislekcia said:
    Feels like the first player to move gets a huge advantage, seeing as they can control corner space further from their spawning point, meaning that the player that moves second has less potential territory that could be covered... I'm expecting bots to appear that start using dying strategically as a repositioning device. But yeah, seems like an okay algorithm could pretty easily solve this space entirely.
    I think with the maps being, seemingly, symmetrically generated, and games lasting what would be a few hundred moves, that a first move won't give that much of an advantage, and if it does your program would need to determine if you have first move or not and act accordingly, so you would need a first and second move strategy regardless, i.e. what is the best I can do if I am first, and what is the best I can do if I am not first, kinda like how being white or black in chess affects your opening moves, and though white seemingly having advantage, is not necessarily always the case, just a thought.
    Also, the entire map is fed into the players program each step with no memory retention other than through a file, and the 4 second limit gives rise to clever algorithms and state space solutions. Then again it does seem as though a simpler algorithm could easily traverse such a map easily, but could it do so consistently across multiple games.
    Keep in mind I am not too clued up on AI, just starting to learn about all teh nifty thingamabobs, so I may just be rambling :/

  • Does anybody here want a partner to test against ? I am thinking of doing this as a programming challenge even if I dont enter.
  • Tuism said:
    BTW:

    I think they need to give the first move alternatively so that both players have the same chance at starting first. First turn advantage is huge when permutations aren't *that* big.
    Do you mean 2 opponents play multiple games and alternate. Like to rounds each having a turn to go first, and a third for tie breakers?
  • That would be the most fair. Unless the two algorithms are completely identical which results in a draw if everyone has two games in a round :P
  • Does anybody here want a partner to test against ? I am thinking of doing this as a programming challenge even if I dont enter.
    Perhaps we can use this thread or start a test pool thread? Then if someone wants a "jousting" partner they can find someone there. So I will give you a shout once I have something in the coming weeks maybe, that is if my neanderthal brain can understand the AIs :P
  • I definitely agree to this. It does appear that this they are trying to rectify exactly this issue from previous years, by putting in 2 pools for beginners and advanced entries, they also have opened up the test harness from the start for scrutiny. So more transparency and much, much better explanation of requirements and outcomes, compared to previous years. Will see if it is for the better then?

    There do seem to be plenty of solutions to Pacman mazes out there, with pros and cons. They suffered this issue last year with Battle Tanks where they really didn't explain anything thoroughly and only clarrified properly nearly a week or 2 before the first finals round.
    Based on their wording and provision of an example state and also on the previous comp, the maze appears to be slight random variations on the standard Pacman style of maze.
    Based on the example a maze is filled with pills and so navigating the maze itself to get the most pills is the focus.
    However, it sould be nice if they were far more explicit regarding the game.
    Well, it's a hiring exercise for Entellect, isn't it? ;) I'm sure they set up the beginner/pro tiers in order to not scare away potentially hireable junior programmers who might never manage to compete with the people who keep winning/entering these things and Entellect *doesn't* want to offer jobs to.

    I suspect that the re-organisation/rule clarification part of the competition is probably useful in how they vet possible hires. You want devs that can handle scope changes on short notice, after all... (Not saying it happens on purpose, but seem to keep needing to do that each year)
    I think with the maps being, seemingly, symmetrically generated, and games lasting what would be a few hundred moves, that a first move won't give that much of an advantage, and if it does your program would need to determine if you have first move or not and act accordingly, so you would need a first and second move strategy regardless, i.e. what is the best I can do if I am first, and what is the best I can do if I am not first, kinda like how being white or black in chess affects your opening moves, and though white seemingly having advantage, is not necessarily always the case, just a thought.
    Also, the entire map is fed into the players program each step with no memory retention other than through a file, and the 4 second limit gives rise to clever algorithms and state space solutions. Then again it does seem as though a simpler algorithm could easily traverse such a map easily, but could it do so consistently across multiple games.
    Keep in mind I am not too clued up on AI, just starting to learn about all teh nifty thingamabobs, so I may just be rambling :/
    Well, I haven't had a long think about it, but the whole way that players can be repositioned by being moved onto by other players means that the player with the first move has the ability to get to any key points faster than the other. As soon as that happens, you have at best a system in which the "weaker" player can draw if the map has absolutely no choke points that have more resources on the other side of them. If there are value choke points, then the player with an advantage can keep it indefinitely by simply moving to block off any chance at the weaker player reaching more pellets.

    I'm not sure what the impact of the "teleport to the center of the map in 2 moves" pills is on this but I assume that it's probably not enough to mitigate the first mover advantage... Instinct says that it'll either become a race to the furthest point on the map and then teleport to the middle before the other player so you can take the pills there first, even a single pill advantage will win the game provided you can secure 2 of the big power pellets. I also suspect that the middle/respawn rules will turn into a bit of a clusterfuck of constant teleporting/positional locking-in once bots start using them.
  • edited
    Hmm. I want to enter this but I would just like the money - don't really want to be hired because I don't mind my current job and i don't want to move away from where i live so maybe I shouldn't enter? I'm just thinking out loud...

    lol... I make it sound like I would absolutely win :)
  • AntiVoid said:
    Hmm. I want to enter this but I would just like the money - don't really want to be hired because I don't mind my current job and i don't want to move away from where i live so maybe I shouldn't enter? I'm just thinking out loud...

    lol... I make it sound like I would absolutely win :)
    You don't win a job, you win the money ;) But in entering an AI in a language that the want skills for, they're seeing how you work and what you're good at doing. So it's a smart way to pay less in recruitment fees, right?

    That's probably the main reason for the second category: To make people feel more like they have a chance at winning so they'll enter.
  • Hi Guys,

    RE: the test partners, we'll be hosting some events at Entelect over the next few months where you can come in and code with other R100K challengers and test your bots out against each other. More details to come in the next few weeks.

    To answer dislekcia and AntiVoid's concerns, no - you will not be hired on the outcome of the challenge and you have absolutely no obligations other than taking home your prize! We are a software company and of course we are always on the lookout for great talent, but the R100K Challenge forms a small part of a bigger picture. We're trying to inject a bit of life into the local software industry, connect individuals through competitions, training forums, community upliftment and education. We do the same for our guys internally, but without industry support of these initiatives in the public domain they just wouldn't happen.

    If you have any questions you'd like to put to the Entelect team, you're welcome to join us at our official discussion board over here:

    Tim
  • Sounds good. Think I'll enter :)
  • @timkroon, thanks for the further information :)
  • Glad to hear it, looking forward to your entry @AntiVoid and @Thelangfordian!
  • edited
    Now there's a video for this



    Or at least, it's the first time it popped up as an advert for me on Youtube
Sign In or Register to comment.