Build your own boss results!

Build your own boss results!
0

#21

I don’t think the boss has been up long enough for new versions bent on “countering” Aeldrexan. Even if you came up with a version that consistently beats him in about a day or two, it’d still be climbing elo trying to get to the top players. It’ll be interesting to see how well that type of algo does in the long run with more people able to play against it.


#22

Thanks @KauffK, indeed I’m pretty proud of this result :slight_smile:. (Actually I don’t make a full use of the simulator, I only use 2% of the computing time because I still don’t know how to efficiently use the rest: free dynamic defense is still a huge mystery for me -> huge respect for @Aeldrexan )

@Ryan_Draves actually it is quite possible: I uploaded tiny_Oracle1.2C 48h ago and it is already at the top. The climbing can be quite fast.

Only @Aeldrexan will be able to confirm if more algos countering his have been uploaded since the competition but I uploaded 5 different variations of tiny_Oracle and they can all beat Aelgoo54_CARDS. So it could explain partially the drop of its elo.

Edit: If you look here https://876584635678890.github.io/?time=2018-11-15T18:43:05.915Zto2018-11-23T18:43:05.915Z, actually the whole leaderboard took a drop


#23

In the last 48 hours, Aelgoo54 had 10 loses, 4 of them comes from the different versions of tiny_Oracle, so I think this is one the main reason of Aelgoo54’s drop.
I think the overall drop of the leaderboard is caused by the fact that many strong algo are being updated (Aelgoo, backtostart/tiny_Oracle, Cthaeh, …) causing an elo deflation (I believe elo inflation/deflation is determined by the average difference between starting elo value and elo value when the algo is removed)


#24

I wonder if the winner lf BYOB should get to pick an algo or two (a week after the tournament say) and have those algos show up as bosses on their account. This would compensate for the information leakage of having the winner’s algo show up as everyone’s boss and would help the winner iterate and improve.


#25

Thank for your proposition, but there is no need for such kind of compensation, I knew that winning BYOB would lead to this disadvantage, and I still joined the competition.

I think however it would be great for the game if every player could fight any other algo in unrated matches, about the same way as bosses, it would really help to quickly improve our algos.


#26

I agree, as long as everyone keep uploading their algos. The only way to have a true feedback on your algos now is to upload them and leave them to climb the leaderboard. If you can have direct feedback by sparring against the algos you choose, you can delete your algo as soon as you recieved a sufficient feedback and work again on your algo. In the end, virtually no one will have seen your algo.

Someone could then implement a great algo in pseudo secret and reveal it at the end of the competition.

Something else I can think of is the computing power it will need. The more sparring matches are played the less ranking matches you will have because the servers will be busy. (This is an assumption, C1 people can confirm this?).


#27

I think this is by design. There are definitely times when I make a change trying to improve on a specific matchup and being able to challenge that algo directly would speed up the feedback loop, but it reduces the global feedback given to all the other algos I face as I climb up the ladder.

I’m hesitant to allow anyone to challenge anyone unranked, for some of the reasons already shared, but if it were to happen I feel strongly that both players need to get that match history. Right now @Aeldrexan is getting zero benefit in all the matches played against his boss algo because he can’t see reason why his algo lost.