Alpha Five: ETF Universe Alpha Streams Competition

Hi Jared.

The competition is a good way to bridge the gap between the quants and the funds: that is much appreciated!

May I ask why there are reqirements such as "Alphas submitted should be long-only"? I think that the requirement to be "long only" does not allow to hedge the portfolio as efficiently as if a long/short portfolio is allowed.

In the announcement email, you mention "daily". Is there a "daily" or greater resolution requirement? Is minute ok?

Hi Jared,

great initiative, let the race begin!

Few questions by the way:

  - Could the winning and/or participating Alphas be licensed by inetersted fund too?

  - A Backtest Handled Error doesn't count as a Runtime Error and respecitve Alpha won't be subject to disqualification, will it be? 

Morning! Glad its exciting! 

Laurent Crouzet - You're right it does bias the strategy but the funds are fairly sophisticated and can handle that. Making it long only eliminates the significant borrowing costs when investing 10-100M short from a prime brokerage. 

SeanBMiller any resolution of data is fine! The algorithm should be active enough to ensure the strategy is responsible for the underlying growth and that if it stops working the institution can quickly determine the "bets" aren't probabilistically accurate anymore.

Mykola Dobrochynskyy absolutely! Alphas entered can also be licensed by other funds -- and you continue to own the IP of the underlying strategy if you win. By entering you give the institution the first right to license the strategy exclusively for 3 months. This seems like a fair compromise and it takes more than 3 months to develop a good track record anyway.

Backtest handled errors are fine. We should determine why they're happening though for overall "resilence". 

We're going to try and refine the alpha submission process in the next month as well to make sure we give you fast-feedback. This could be a great leap forward for QuantConnect and the community, its a big opportunity! Happy hunting!

Hi Jared,

Thanks for quick answers. I've got a few more:

- Is there some way to intercept / debug a Backtest handled error?

-  After a "big" 5 years test I've got a log-message "Algorithm Id:(abcdef) completed in nnnn.nn seconds at yyK data points per second. Processing total of z,zzz,zzz data points.", but the runtime monitor "hangs" at 100% though. Could I assume it as successfullly executed?

I can't say for sure Mykola without understanding the problem more. However, all normal Alpha Streams review applies so if the backtest does not complete (and upload) successfully for some reason we cannot review it. If you have algorithm specific questions shoot us a message via support.

OK, thanks

Very cool competition!

I'm hoping November 1st deadline will encourage the right style of submissions too -- it's not enough time to spend weeks optimizing and tuning a single idea, but hopefully its plenty of time to conceive, test, submit 4-5 "hypothesis-driven" ideas. 

Gonna give it a try! :) 

Are there some examples / documentation how to implement CoarseSelectionFunction / FineSelectionFunction with this 5 Universes?

Hi, 

Thanks Jared. This competition looks interesting. My questions regarding this competition are:

1). Are the leveraged ETFs allowed to use?

2). What is the maximum leverage we are allowed to use?

3). How many algos can we submit? One each category? (totaling 5?) 

4). 'Long only' strategies usually have some hedging strategies against directional risk. Are we allowed to short ETFs to hedge directional risk at particular time periods (not all the time)? If not, can we use inverse ETFs (long inverse ETFs) instead ? 

5). What does exactly mean by 'daily active'? Should we trade at least once a day? or just set up to run our algos daily basis although there are not necessarily trades everyday basis? 

Can you clarify about my questions? Thanks. 

Hi all, 

Can you guys show us an example of how to add one of five universe baskets above in the algo? For example, let's say we want to use 'Energy ETF' basket for our universe with price above $10, AUM above $100M, and average daily volume above 1M. And we want them sorted by volume (or AUM), and then finally get top 20 symbols out of this process.

Can anyone show us an example? Thanks. 

Hi all,

We've just published a short post on how you can use the first of our five predefined ETF universes: Technology ETFs. The different ETF universes we will publish are fixed and must be used in the competition (i.e. you can't build your own or modify the template universes to have your submission qualify). Have a look at it for some guidance, and keep an eye out for posts with demonstration algorithms for the other four competition universes.

Hi guys,

Congratulations on developing a unique attractive initiative that is sure to propel this community forward enormously.

I too have a few questions:

1) assume it is never allowed to go short. So you could not go long Gold and short Platinum, for example

2) do you allow the use of call writing, straddles, calendars and other options strategies within the given sector, in complement to the long stock positions?

3) Is there a time limit to how may parts of the day an active position must be maintained. Let me give you a ridiculous , example to clarify my concernt.  Let's say I have developed the perfect pulse of the market that knows with 100% certainty that all precious metals will go down severely over the period, in a catastrophic  downturn. Since I cannot use short positions, would it be permissible to only play the intermittent upticking reversals, which means being in cash most of the time but occasionally holding long positions intraday. Is there any requirement to hold long positions overnight?

4) Finally, in developing the winning criteria, do you use absolute earnings criteria, or relative onces compared to the index in question. So if tech stocks go up 150% over the 5 year period and you achieve 160% are you favored over an algo in the precious metals sector that outperforms its index, being down only 10% when the underlying sector (or etf basket) is down 40% ?

Thanks

Hey Serge!

1) Long Only -- Correct;

2) Straddles etc would not be possible with the universes provided. Please see the 5 posts here "Competition Template" that Jack made today. 

3) Any holding pattern is fine. The universes are relatively liquid. As you increase your trading frequency it will reduce capacity but the judgment is heavily weighted to Sharpe Ratio.  

4) Each universe is judged independently so it'll just be "benchmarked" against other peers in the universe. The common benchmark for beta calculations is the SPY.

Hi guys, 

one more question, just to be sure. Do I understand correctly that we can trade the five ETF baskets mentioned in your initial post but it's not permitted to trade the single stocks/components which belong to the respective ETF?    

"not permitted to trade the single stocks/components which belong to the respective ETF" - Correct. This is because the larger ETF has greater volume capacity that some of the smaller individual assets and is less susceptible to volatile swings. The underlying assets can be used as signals, as long as only trading on one universe.

Does "active each day" mean that some position turnover must occur each day, or only that the portfolio must be invested for every day of the competition? 

Tough question Kieran, from the institution's point of view; they're looking for the balance between: 

• Sufficient signals to determine the return is a function of the algorithm, not the underlying assets.
• Not too many signals that the strategy potential capacity is reduced.