Quantitatively test individual signals for alpha and risk
Alpha Tests allow you to quantitatively test individual signals for alpha and risk, providing a quick and easy way to run such analysis in bulk.
Click here to open the Alpha Tester, or from the Exabel menu bar, click "Strategies" → "Signal alpha tester".
Alpha Tests allow you to individually test the signals that you build in Signal Explorer. They are highly complementary to Portfolio Strategies, which allow you to create strategies using multiple signals and provide additional optimization and risk control options.
Under the hood, Alpha Tests and Portfolio Strategies both use the same backtesting engine.
Alpha Tests in workflows
In a quant workflow, users might use Alpha Tests as part of an rapid iterative loop of develop signal → alpha testing → optimize signal → ... Optimized signals can then be exported for use with in-house backtesters & optimizers, or put into production with Exabel Portfolio Strategies for live trading.
Alpha Tests may also be interesting to fundamental / quantamental users, in providing a high-level view of potential information value in an evaluation dataset, as applied to your investment universe, prior to fundamental deep-dives into the data.
Running Alpha Tests
From the Alpha Test page, click on "Create alpha test":
From this Alpha Test wizard, you can run multiple Alpha Tests at a time by choosing one or more signals and company universes; each company universe is specified as a tag or screen. An alpha test is run for each combination of signal + tag/screen. For example, if you choose 2 tags and 2 signals, 4 alpha tests will be run.
Only up to 6 alpha tests may be run at a time
We are working to raise this limit.
There are additional options with which to configure your alpha tests:
- Normalization: choose a method with which to normalize the alpha signal being tested
- Time period: the time period over which to backtest the signal
- Rebalancing frequency: how often the portfolio optimizer will rebalance the portfolio
- Trading cost: the assumed cost of buying/selling positions
- Daily stock movement threshold: filter out stocks that have daily movements above this threshold, to remove the impact of outlier / extremely volatile stocks
- Factor attribution: attribute style, industry & country returns using Exabel's factor model, to isolate the signal's factor vs specific return
- Point-in-time evaluation: whether to evaluate the signals based on their known values as of each rebalance date. For more, see our release announcement.
Alpha Test results
Alpha Tests provide a detailed analysis of a signal's alpha and risk characteristics.
The summary at the top displays the alpha test configuration, key return & risk metrics, and charts of returns, turnover and position:
The next section shows return attributions - breaking out factor vs specific return, and returns by industry and country. Click on › next to Style, Industry and Country for more detailed attribution breakdowns.
Finally, it is possible to analyze the individual positions taken by the Alpha Test, top one-day contributions, and signal coverage within the universe.
Updated 9 months ago