Alpha Streams

Submitting an Alpha


Alphas submitted to be licensed are hosted live in QuantConnect's production environment for free by QuantConnect. Their insights are recorded in real-time and timestamped to the millisecond.

To get an Alpha listed we ask for information about the author; along with meta information about the strategy itself such as a description of its focus. This information will be consumed by marketplace participants to determine whether to license an Alpha.

Once the Alpha is approved and deployed it begins the process of generating a live track record. Insights generated out of sample are marked as such and displayed separately in the marketplace.

To get started; navigate to your fund dashboard to submit an Alpha.

Submission Information

Alpha submissions provide some information funds can use to help them license your strategy. Start by giving your strategy an appropriate name for the Alpha, along with a description of the Alpha strategy. Most Alphas are relatively simple, so a sentence or two is fine here. Alpha pricing is covered in the Pricing an Alpha section.

Don't worry too much about making this perfect; you can update these fields later as your Alpha improves.

From there you should select an icon to represent the Alpha, along with your project and the backtest with the code snapshot you are submitting. For funds to consume the Alphas, they must generate Insight objects, so only Framework or Upgraded Classic algorithms are supported.

Submitted Alphas require a short review process to ensure they are technically sound and follow the community guidelines. We'll explore this in the next section.

Review Process

Alphas are reviewed for seven technical criteria. We aim to not impose any value bias during an Alpha review but primarily to ensure they are mechanically sound. When submitting your Alpha please review it for the following points:

Stateless & Resilient

Mechanisms should be set up to restore internal state in the event of reboots. Alphas will be very long running. Over the course of months or years, the server will need patches and occasional restarting. The alpha should be smart enough to automatically recover in the event of restart using History and WarmUp API methods.

Use Reliable Data Sources

Alphas should use public data, commercial feeds or QuantConnect's data. Importing a personal dataset through DropBox isn't reliable enough to build a long-term track record and ensure the Alpha will continue working as expected. You should use public data sources (e.g. Quandl), commercial data, or QuantConnect supplied data.

Grounded in Reality

Alphas should have clear strategic reasoning underpinning the strategy. A single reason based foundation for the algorithm is necessary to understand the alpha behavior when things go wrong, and it helps improve your alpha application for funds. Obscure correlations or overfitted strategies tend to perform poorly out of sample.

Transparent Sourcing

Authors need a transparent employment history and alphas must not infringe on other IP. We'd like to know the employment history of authors on the platform and ensure the alpha is not infringing on any intellectual property, including your current and past employers. Public or shared content may be used as a foundation and extended. Where you may have licensed the IP from a third party, presenting consents for usage are acceptable.

Edge Case Handling

Alphas need to be able to handle edge cases such as Dividends, Splits, and Delistings. This adds another layer of resiliency and will keep an Alpha running in the event of data changes. This can involve simple control logic, explicitly removing certain affected securities from the Universe, or other methods.

Recommended Best Practices

We won't reject any Alphas based on what is outlined below, but these are good practices to abide by when authoring an Alpha.

Avoid Look-Ahead Bias

Look-ahead bias can negatively affect the performance of an Alpha once it is deployed and will limit its appeal for licensing. Bias is introduced by using information or data in testing that would not have been known or available during the period being analyzed. In order to avoid look-ahead bias in an Alpha, it is vital only to use information that would have been available at the time of the trade.

Limit Selection Bias

Selection bias happens when an Alpha selects specific securities where the data is known and where a subset of the data is systematically excluded due to a particular attribute. Universe Selection aids in preventing selection bias by blindly choosing assets with limited filtering and without foreknowledge of whether the security will be delisted, declare bankruptcy, split, etc. Fixed universes are not necessarily biased, but authors must exercise caution. Alphas whose Ingiths are generalizable across groups of securities and market conditions will generate more appeal than those with minimal scope.

Although not a technical requirement for submission, Alphas should be simplified and generalized as much as possible by omitting specific APIs that LEAN offers.

Reality Modeling Code to Omit
Fee ModelsBrokerage Models
Fill ModelsSlippage Models
Buying Power ModelsSettlement Models

All of these APIs add functionality for individual traders or researchers, but funds have specific methods and models they'll want to use. Omitting these will remove some constraints upon Insight emission and provide an unfiltered view of how the Alpha model performs.

You can also see our Tutorials and Videos. You can also get in touch with us via Chat.

Did you find this page Helpful ?