Digital Green Crop Yield Estimate Challenge
'%3e%3ccircle r='20' fill='%23008'/%3e%3ccircle r='17.5' fill='%23fff'/%3e%3ccircle r='3.5' fill='%23008'/%3e%3cg id='d'%3e%3cg id='c'%3e%3cg id='b'%3e%3cg id='a' fill='%23008'%3e%3ccircle r='.875' transform='rotate(7.5 -8.75 133.5)'/%3e%3cpath d='M0 17.5.6 7 0 2l-.6 5L0 17.5z'/%3e%3c/g%3e%3cuse xlink:href='%23a' transform='rotate(15)'/%3e%3c/g%3e%3cuse xlink:href='%23b' transform='rotate(30)'/%3e%3c/g%3e%3cuse xlink:href='%23c' transform='rotate(60)'/%3e%3c/g%3e%3cuse xlink:href='%23d' transform='rotate(120)'/%3e%3cuse xlink:href='%23d' transform='rotate(-120)'/%3e%3c/g%3e%3c/svg%3e)
Helping India
€9 400 EUR
Completed (over 2 years ago)
Prediction
1368 joined
678 active
Start
Sep 04, 23
Close
Dec 03, 23
Reveal
Dec 03, 23
'%3e%3cpath fill='%23001489' d='M0 0v6h9V0z'/%3e%3cpath fill='%23e03c31' d='M0 0v3h9V0z'/%3e%3cg stroke='%23fff' stroke-width='2'%3e%3cpath id='d' d='m0 0 4.5 3L0 6m4.5-3H9'/%3e%3cuse xlink:href='%23b' stroke='%23ffb81c' clip-path='url(%23c)'/%3e%3c/g%3e%3cuse xlink:href='%23d' fill='none' stroke='%23007749' stroke-width='1.2'/%3e%3c/g%3e%3c/svg%3e){kind=small}
Updated leaderboard
Platform · 6 Dec 2023, 13:37 · 12
Dear Zindians,
At the close of the competition a bug was discovered.
If you did not do submission selection then your private score corresponding to your best public score was chosen. Which is incorrect as the rules state:
If you do not make a selection your 2 best public leaderboard submissions will be used to score on the private leaderboard.
This bug has been corrected, the leaderboard has been re-ranked accordingly.
Emails to the top 10 will go out this afternoon requesting that your code be submitted by Friday afternoon.
Thank you for your patience,
Amy
Once again your scoreboard is wrong. In case of me you choose my best private score 110.1269811 (public score=431.1896058).
But based on the two best public scores my private score
Public score = 266.2481762 Private Score= 223.7178574
Publics score = 292.6099404 Private Score= 172.3468881
Please carefully check the calculation. Because everyone has given too much time in this competition.
So instead of 2 best public leaderboard submissions, 2 best private leaderboard have been selected..Was wondering why theshakeup was so huge
yeah, I was wondering the same
It seems that this update has placed some users' best private scores on the leaderboard. If you check the front page, you will see several new faces with public scores ranging from 420 to 500. Before the update, these same users had lower public scores (around 160) but worse private scores.
Hopefully @amyflorida626 and team , corrects this second bug in good time before sending out the emails to the winners.
Hi @Zindi, I think the update may have introduced a new bug. As others have mentioned below, I think the new model has used the submissions with the best private scores for those who did not choose two submissions. This is clearly wrong, as it should be their submissions with the best public score that is chosen, as per your stated rules.
Cleary choosing the model with the best private score is a massive advantage to teams that just submitted a bunch of results, since it chooses results based on something they would have had no knowledge of during submission time.
This can easily be seen by looking at the scores that were changed. There were people who originally had submissions with a better public leaderboard score and a lower private leaderboard who now have submissions that have a a better private leaderboard score and a lower public leaderboard score, which makes no sense if their best public leaderboard should have been chosen.
@JuliusFx@yanteixeira@Sourabh Exactly what I have been talking about in my previous post guys. I think this is pretty straightforward, according to zindi rules, only your best two public scores must be selected for any participant even though the participant didn't select his or her best two public subs scores, just like on kaggle. That's my thorough understanding of this. I stand to be corrected though. This has nothing to do with our private scores because no one has access to this not until competition ends.
My candid take.
Yes if the best private scores are automatically selected, it's effectively the same thing as allowing you to submit all of your predictions instead of just the best two.
This is also not great for Digital Green, since it means that the top models now are not necessarily the best ones, just ones that may have gotten closer to the top based on random chance in their private score.
I'm sure @Zindi will correct the issue once they see this thread.
sure micadee
The main issue for this competition is the outliers from the public test. So, competitors are misleading because if you fit the private score better, the public score worsens. This explains why some of the best private scores were not chosen for submission and suddenly became a top solution in the leaderboard. This happens to me as well. My selections for submissions were not the best private scores, but I made my own choices and had to accept them.
I think many of you may share the same problem. It hurts true competitors who choose their submission carefully and it's a "massive advantage to teams that just submitted a bunch of results". Hopefully, Zindi can fix this and make it a fair competition.
Dear all, we will look into this over the next 24 hours and get back to you.
Please update the leaderboard.