Best posts made by Sun-73

Hi there,

Since we're approaching the end of the submission phase for the Q16 contest, is it possible for us quants to already know a little bit about the upcoming competitions?

Will the Q17 contest be designed similarly to Q16? In other words, will Q17 be based only on crypto trading (or will it allow for futures trading too)? Only long-positions allowed (or short-positions too)?

I would suggest (if feasible) to put in place a new contest allowing to trade both cryptos and futures, besides trading long and short in any asset, but limiting the overall leverage of the portfolio. For example, adopting a 130/30 portfolio structure, which is a structure with a net long position, but permits leverage with long and short positions up to a maximum amount of leverage. In the case of a 130/30 portfolio, the leverage is 30%.

This broader and combined setup would allow us to design enhanced portfolios and might (potentially) increase the out-of-sample Sharpe ratios. Thank you!

Best regards,
Sun73

Hi there,

The set of 15 selected systems displayed on page: https://quantiacs.com/personalpage/submissions/tab_in_contest
or on page: https://quantiacs.com/leaderboard
is not the same that I have selected (before October 31, 2021) on page: https://quantiacs.com/personalpage/submissions/tab_candidates
to be my final set of systems to compete on the Q16 contest (for example: Sun73_Q16_4f).

Also, there are systems not listed anymore on the latter page (such as: Sun73_Q16_3a) which I have previously selected to compete on Q16.

Could you please check this (selection) issue? Many thanks!

Kind regards,
Sun73

Hi there,

Another question about the new Bitcoin spot data: Should we use the code lines below to import the data? Or there is a new way to do that?

def load_data(period):
data = qndata.cryptofutures.load_data(tail = period, dims=("time", "field", "asset"))
return data

In respect to the competition "type", will it be called "cryptofutures" or there is a new name for it?

Many thanks again!

Hi @support,

Thank you for the prompt response. My set of Q16 systems displayed on the website https://quantiacs.com/leaderboard/16 is the following: Sun73_Q16_1c , Sun73_Q16_1d , Sun73_Q16_1e , Sun73_Q16_2b , Sun73_Q16_2b_v2 , Sun73_Q16_2b_v3 , Sun73_Q16_2c , Sun73_Q16_2e , Sun73_Q16_2f , Sun73_Q16_3a , Sun73_Q16_3b , Sun73_Q16_3c , Sun73_Q16_3d , Sun73_Q16_3g , Sun73_Q16_3h.

However, my chosen set of selected systems is the following: Sun73_Q16_2c , Sun73_Q16_2e , Sun73_Q16_2f , Sun73_Q16_3a , Sun73_Q16_3b , Sun73_Q16_3c , Sun73_Q16_3d , Sun73_Q16_3e , Sun73_Q16_3f , Sun73_Q16_3g , Sun73_Q16_3h , Sun73_Q16_3i , Sun73_Q16_3j , Sun73_Q16_4d , Sun73_Q16_4f.

All systems above were submitted before the Q16 deadline. Since there is an issue with the auto/manual selection, could you please select the 15 systems above for the Q16 contest? Many thanks once again!

Regards,
Sun73

Hi there,

The new function backtest_ml() implemented in the example: "Machine Learning with a Voting Classifier" is a great way of getting a fast feedback from a given model.

Is it possible to use it with multiple datasets (for example, using "futures" and "commodity" data, as done in the template: "Futures - IMF Commodity") ?

Thank you!

@support Yes, I can load now the stocks data. Thank you once again!

@support Great! This route opens new possibilities in terms of model design. Thanks a lot!

Hi there,

I submitted a strategy for crypto-futures, called Sun73_Q15_BTC_1f that was not accepted in the contest due to a Sharpe ratio lower than 1.

However, the in-sample Sharpe ratio obtained with the JupyterLab does not match the one displayed online. Both prints are shown below.

It seems that the online system almost does not trade Bitcoin after March 2017 (please see the first graph below), whereas the original system from JupyterLab trades Bitcoin on a regular basis since 2014 (as shown in last graph).

Could you please verify this mismatch?

Thanks a lot!!

@support Thank you for the prompt response. I appreciate if you could reprocess the strategy Sun73_Q15_BTC_1f after solving this cache issue. You guys are the best!

@qida1995 Hi qida1995, please take a look at the documentation available in the Quantiacs' website (https://quantiacs.com/documentation/en/), which is very useful not only for beginners but also for experienced users. Best of luck!

Hi @support,

I modified the retraining interval to 1 day and it worked. Thank you for the help.

You guys rock!

Hi @support,

The SR of the system "Sun73_Q17_1a" is above 1 in JupyterLab (SR = 1.08), but almost half of it (SR=0.50) after submission to the Q17 contest.

Do you have any clue on that? I am using a modified version of the example "Q17 Supervised Learning".

Many thanks!

I am using the "in-sample Sharpe" in the comparison above.

Hi @support, a suggestion to reduce the overall processing time is to adjust the "Running Time" rule: instead of 5 minutes of time per each point in time, I would limit it to 1 minute per observation.

Besides reducing submission processing time for ALL users, it would stimulate us to produce parsimonious systems (more focused on results), instead of building huge models with tons of data that generate a severe computational burden without effective gains.

This would also guide users to shift the discussion from "big data" to "good data", in the sense that finding high-quality data is much more important than the quantity of data itself (Ng, A., 2015. Advice for applying machine learning).

Hi @support, last week, a set of previously submitted (and accepted) strategies were moved from the Q22 competition (from the “Candidates” area, in the website, to the “Filtered” area).

Some strategies were submitted more recently while others several months ago.

In the “Filtered” area, there is no comment attached to each strategy to justify the mentioned exclusion from Q22.

The strategies are the following: Sun73_Q22_6a, Sun73_Q22_5a_v3, Sun73_Q22_4a, Sun73_Q22_2e, Sun73_Q22_2d, Sun73_Q22_2c

Could you please check this issue? I noticed that it also happened to other users as well.

Many thanks!

Hi @angusslq, the same thing happened to me (ex: with my strategy Sun73_Q22_6a).

Hi @support, can you please check this issue for us? Thank you!

Hi @support, thank you! The box is now working fine!

I have a simple suggestion regarding the charts created for each strategy:

Besides the equity curve (strategy, SPX and Nasdaq100), Logarithmic scale, Long/Short, Underwater, and Bias, it would be nice to see an additonal chart (with two lines) showing the number of assets in each day that have a positive weight (long position), as well as the number of assets in each day that have a negative weight (short position).

This will help us developing the strategies, since we can better track the outcomes generated by different algorithms being constructed.

Also, this can help you identify asset hand-picking with buy-and-hold positions.

Thanks!

Hi there,

Is the sample period used to evaluate the new dataset (BTC spot) the same (that is, since 2014-01-01)?

Thanks!