A number of users have shown interest in a more equitable Global competition prize distribution. The team is officially considering changing the distribution of prizes at this point. I will be hosting a complete discussion about the proposed changes here. We will be heavily considering the opinions of “Top Players” as they will be most affected by the change.
Proposal: Distribute rewards among the top 10, instead of all for one
Benefit: The ladder is inherently volatile, and it is hard to say one algo is definitively more robust than another. It is possible to hold the #1 spot for a while, but win nothing if you lose in the final hour in the current system.
Downside: The main downside I have seen so far is that players are concerned about multi-accounting. We will be individually verifying the identities of our top 10 users before we distribute prizes. We ultimately believe it will be very difficult for one person to make multiple algos.
Teams are allowed but must submit on a single account.
If anyone has any opinions on this, let us know, and answer this poll.
We should distribute prizes to reward everyone in the top 10 in some way
The prize pool should go to #1 only, as originally planned
Distribution proposal: “We need something to stream”
For starters, everyone in the top 10 gets 300$, for 3k total.
A round robin with every player in the top 5 facing every other top 5 player, for a total of 10 games. Every game would reward the winner 600$, for 6k total. The last 1k will be split among the player(s) who won the most games in the round robin.
This is one proposal to get us started. We will have others from the team, and will be taking yours into consideration
The proposal you gave is already a major improvement! It also creates the opportunity for a nice stream with interesting content hosting the true finals. To everybody who didn’t make the top 10, but are interested, there is the opportunity to see the algorithms in high level action. I would then suggest to ask the top 5 players to hop on the stream, and explain the idea behind their algos.
Maybe would suggest I slight gradient between spot 10 and spot 6 though. I’d like to see everybody fighting for one spot higher, and squeeze everything out of their strategies.
I am 100% in support of this idea, and not just because of my recent fall from grace.
I have always been slightly bothered by a winner-takes-all approach for this game, considering the non-transitive victory relationships between algos, and the high diversity and volatility of the leaderboard.
For me at least, this prize distribution feels much better and greatly increases my motivation to pour more time into the game (as opposed to, I dunno, getting my degree or something silly like that ).
Multi-accounting has been bothering me for a while, since it can be deployed in a variety of ways to tamper with the leaderboard. The idea of interviewing the highest prize winners and other account-verifying methods puts most of these concerns to rest. The only abuse of multi-accounting I can think of at the moment is if you wanted to jump your algo up a few spots in the top 10, so you make many other accounts and flood them all with single-purpose hard counters to try and disrupt the algos above yours without actually beating them with your main algo, and without your other accounts actually winning. However, this could probably be detected depending on the similarity of the “silver bullet” algos and on what account verification methods are used.
I’m in full support of the top ten splitting the monetary reward. Would there be non-monetary rewards for players that are in say top 100 (top .1%) such as a certificate or something so we have something to show for doing well?
Just checking info on prize distributions, top 10 seems a good balance.
I think someone submitting multiple accounts is not really a risk, as anyone talented enough to be in the top 10 with the current job market for ML programmers wouldn’t have any reason to cheat and damage their CV.
If someone adds a last minute trick play to reach #1 perhaps they don’t deserve to win the prize money away from other top players who’s been in the ladder exposing his strategy for a long time hmm (full disclosure been ranking in top 15 today so my vote is biased ha)
Is there any discussion around expanding qualification from top 10 to maybe top 15 or 20. Its been my experience as someone in the lower tier of the top players that there is a lot of volatility outside of the top 5 or so spots. And there are great algos outside of the top 10, rubercuber won the last competition while being ranked 16th if im not mistaken.
).