Posts made by support

@sun-73 @RoyPalo, Hi,

Q23 Leaderboard was updated several days ago, all eligible submissions are there now, sorry for late notice. Please let us know if you find any submission that is missing.

@roypalo Hi,

can you please send the submission id?

Dear @sun-73,

All strategies that are sent before contest deadline will be processed and eligible for participating in the contest, even if they finish with processing later.

The limit of 15 strategies per contest remains the same, but that doesn't mean one user can't submit more than 50 strategies. He just has to pick which 15 of them to send to contest, and if he doesn't, then the 15 best sorted by Sharpe ratio will be considered as participating in the contest.

The strategies of other users, or their number, will not make your strategies exceed limits and filter them out. If you suspect that some strategy shouldn't be filtered out by time limit constraints, you are free to submit the same strategy again. Maybe it will work the next time, and if not, then there is probably some issue with the strategy itself.

Regards

Dear @gjhernandezp,

That strategy is filtered out because it exceeds limits (memory or time) that a strategy has. Your strategy is using machine learning, so that is risky section and could lead to something like this. The success status is irrelevant, it is probably some edge case. The only thing that matters is the status of the strategy which is filtered.

Best regards

Dear @antinomy,

There has been some lag in processing, thanks for pointing it out.

Regards

Dear @antinomy,

We didn't want to completely reprocess your strategy, but since it was filtered out before we had to manually return it to eligible candidates list.

Regarding your second question, we have updated our correlation service, and you could try submitting again and see if it will be filtered out.

Best regards

Dear @antinomy,

This issue has been resolved, and your strategy is now eligible to participate in the future contest. Thank you for bringing this up.

Best regards

@antinomy Hi,

sorry for late notice, we are checking what is the reason of such behavior, we will let you know soon.

Best regards,

@sun-73 Hi,

sorry for late answer, this has been fixed, please use the latest qnt library version. Thanks, that's correct, there is a typo in check() function, it will be changed in next update - cut_big_positions() is function from qnt.exposure module.

Best regards

@theflyingdutchman Hi, sorry for the delay, yes, all fine, more details by e-mail

@antinomy Hi,

You are absolutely correct, we will fix check() function ASAP. It should only cut weights which exceed 0.1 allocation by asset, and normalize the sum of allocation to maximum 1, on every timestamp. If sum was < 1, and weight of an asset < 0.1, the output remains the same.
Thanks a lot for pointing this out with examples.

@angusslq Hi,
yes, it is expected and will be updated soon. Since it is the end of Q22 contest period, this update will show preliminary results of Q22. Thanks for your patience.

@sun-73 Hello,

Thank you for pointing out the bug — we really appreciate it. The issue with the "Show only my results" checkbox on the Q23 Global Leaderboard has been fixed. You can now check that everything is working correctly on the site.

Please note that you might need to clear your browser cache for the changes to take effect.

If you have any suggestions regarding the interface or other features, we’d be happy to hear them!

Best regards,
Support Team

@illustrious-felice Hi,

sorry for late answer, please check the correctness of dataset used for defining weights in strategy. Be sure that 'stocks_s&p500' dataset is used and not 'stocks_nasdaq100' for current competition. To ensure strategy trades only liquid assets in certain time period, multiply the output from your strategy function with 'is_liquid' field from correct dataset, or simply use clean() function from qnt.output:

import qnt.data as qndata
import qnt.output as qnout

def strategy(data):
    .....
    # liquid = data.sel(field='is_liquid')
    # weights = weights * liquid
    return weights


data = qndata.stocks_load_spx_data(min_date='2005-01-01')
weights = strategy(data)
weights = qnout.clean(weights, data, kind='stocks_s&p500')

qnout.write(weights)

Also, keep in mind that submission will not be eligible for contest if stocks universe (in this case "top 7 magnificent") is hand picked (manually defined).

Best regards,

@magenta-grimer Hi, the Q22 basic template is a good starting point.

@nosaai Hi,

which version of qnt library is used? We introduced that function about a year ago, with S&P500 stocks dataset, maybe try with the most recent qnt version. If the issue persists, please let us know.

Regards

@alexeigor Hi,

sorry, it should work now.

Best regards

Hi,
thanks for pointing this out, we are working on refactoring the code, qnt is reverted to previous version. Sorry for late answer.

Dear @omohyoid,

That would not be a quantitative approach, hence it is not allowed based on the current contest rules.

Best regards

Dear @angusslq,

The ranking is calculated based on contest sharpe ratio only. The situation that you saw at that time was only temporary because at that point the calculations for new day were ongoing, so some strategies were processed before others, and their contest sharpe was updated, but the ranking is only updated after every single strategy from that contest has been processed and updated.

Regards