Gentlemen,
it's not yet time to congratulate or celebrate.
Keep your eyes wide open..
The leaderboard file not updated yet.

Contest 27
LAST UPDATED
2 days

Hi Douglas,

My pseudonym is Gold Swordfish.
I've been leading the contest 27 for the last month.
I see from my files the sudden jump in your rank from 23 on Aug 17 to 1 on Aug 29.
But annualized return decreased from 16.9% to 14.6%.
How can you explain that jump in rank?

Ryan,

Thank you for the link and congratulations.
You are powerful enough to fight for third prize.
It is much easier to fight for justice in the contests with this support from Willie Trader and others.

Douglas,
You have to refute the Willie Trader findings otherwise Quantopian must recalculate files.

Douglas, what do you think of your algo jump into the 1st place just 2 days before the contest 27 end date?

Do you think it is fair to change the rules and reshuffle the leaderboard standings like that, or you also believe that only new contests that have not started yet may use updated/changed rules? Is it a matter of principal and how clean the competition is run?

We are talking about millions under management and fututure exit event for Quantopian, be it an IPO or a buyout, or something else.

Please share your opinion here or here https://www.quantopian.com/posts/is-there-a-bug-in-calculating-annualized-volatility-results-in-34x-of-max-possible-move-affecting-contest-27-ranking

Best!
Willie

Olive Coyote: thanks a lot, I am working hard to have good out-of-sample procedures, and I'm happy it is working out for me.

Ryan: Thank you for all that helpful information, and for pointing us to that other thread.

Regarding my jump up in the rankings, like Dan pointed out in the other thread, there was a bug in the leaderboard that caused the rankings to be incorrect. I was in first place before the bug was introduced. My top performing entry then suddenly disappreared from the leaderboard for a few weeks. Following the first attempt at a bug-fix from Q staff, there was a strange situation where the metrics for some algorithms (including mine) were much worse on the leaderboard than when viewed on the live trading dashboard and stats page. For example, beta was reported around 0.2 for me on the leaderboard, but 0.01 on the stats page.

I noticed that backtests of copies of my algorithm gave results consistent with the leaderboard. Furthermore, I wrote my own code to compute the metrics, and found results consistent with both the backtests and the live trading stats/dashboard. At this point I was hopeful that the numbers on the leaderboard were incorrect, which indeed turned out to be the case.

I was in first place before the bug was introduced into the leaderboard, and returned to first place after it was resolved. I want to be clear that the bug was in the leaderboard, not in paper trading. The contest rules weren't changed, just the leaderboard was incorrect for a few weeks.

I can speak for myself that I do not post frequently on the forum, because I do not want my name associated with some of the discussions. There have been a lot of strange opinions about what Q staff could, should, or "have" to do. I would like to remind some people that Quantopian is a private company and can do as they wish. We are lucky that Q provides this amazing platform where we have all learned a lot about quantitative finance. Their main responsibilities are to uphold the mandate set by their board of directors, not to run a contest. I am worried that due to unprofessional behaviour by some users, they will decide that the contest is not worth their while and take it away from us.

I personally think of the Q staff as potential business partners, and try to treat them with the same respect I treat my colleagues and collaborators. A few weeks ago, I thought I was going to lose the competition due to a bug in the leaderboard. I assure you I didn't complain. In fact, when I found out about the issue, I told the staff that I would be disappointed, but would obviously accept whatever outcome they determined to be fair.

Besides making oneself happier by refusing to get worked up about these things, there is also a lot to be gained from professionalism, and a lot to lose from the opposite. I can say, from my own business decisions, that if a deal is borderline, I might choose to say "no thanks" if the person I would have to work with is unprofessional.

Finally, there is a lot to be said for confidence in these situations. I assure you, if I don't win contest 27, I will win some other contest -- it is only a matter of time at this point. For example, paper trading was restarted for contest 32, and yet I am back up to third place. Similarly Olive Coyote is back up to #2, and Ryan is also top ten. Victoria: your algorithm looks good. If you don't win this time, you will have a good chance in contest 28. Either way, I wish you the best of luck in future competitions, or in this one if it goes that way.

I am on vacation at the moment and am writing this ridiculously long post using my phone, because I felt the need to defend myself and Q. I will be hiking out into the wilderness soon, where there will be thankfully no cell reception, and certainly no discussions of leaderboards. Thanks again to the various people making constructive comments and positively contributing to the community and competition.

I want to remind you one contest rule:

https://www.quantopian.com/posts/is-the-contest-being-gamed
John Fawcett
Feb 5, 2015
After the submission deadline, you can stop an algo, but you cannot restart it in that contest. For example, if you stopped an algo now, and resubmit it, you would lose the entry in the February contest, but the algo would still be in for March.

In private communication I asked Douglas:

From leaderboard file 1502948978 I see that you have the only one algo in contest 27.
Is that correct?
He did not answered.

My findings conform Willie Trader findings:

According leaderboard file 1502948978 calculated on 02/28/2017 Douglas Staple has only one algo with submission_id 58b38f426dad21000a1d9d6b

According leaderboard file 1504158770 calculated on 08/30/2017 there is no algo with
submission_id 58b38f426dad21000a1d9d6b

According leaderboard file 1504158770 calculated on 08/30/2017 Douglas Staple has two algos with submission_ids: 58b57a5c98847226134617c6 and 58b588bf83494e7b1ed586f9
They appear on leaderbord in violation of John Fawcett Feb 5, 2015 rule.