All Collections
Factor Drift Methodology
Factor Drift Methodology

Automatic calculations display how a factor's current risk & exposure sizes up against its historic values

Written by Jeremy Mulder
Updated over a week ago

The Omega Point application provides portfolio analysis tools that include describing how a portfolio's factor risk & factor exposure have deviated from their historic values. This is described as a factor's drift and can be found in the right-hand column when using the Analyze: Risk and the Analyze: Exposure modules

Drift is categorized by the following percentile buckets:

Very High: 95% - 100%
High: 75% - 95%
Neutral: 25% - 75%
Low: 5% - 25%
Very Low: 0% - 5%

Image of line graph displaying Style Exposures, with Apr 26, 2021 selected and showing that day's portfolio exposures for each style factor
  1. Drift is calculated for the specified date range
    — when the date range value is less than 3 months, the app automatically uses as many dates as possible to get a statistically significant drift calculation. 

  2. Selecting a factor category will display historical factor exposure values [or a factor's risk decomposition values, when using Analyze: Risk].

  3. Dynamically interact with the graph, and introspect any particular date (dashed line), then...

  4. Display that date's weighted-average factor exposures in the right-hand column, and it's drift (red, blue arrows).

Drift Methodology

  1. Find the mean and calculate the standard deviation over the full date range*

  2. Find the difference from today's value from the mean, divided by the standard deviation

    driftValue = (todaysValue - mean) / deviation

  3. The driftValue is then bucketed into it's respective percentile by the following categories

    driftValue < -1.645σ = Very Low
    driftValue < -0.674σ = Low
    driftValue < 0.674σ = Neutral
    driftValue < 1.645σ = High
    driftValue > 1.645σ = Very High
    The range [-0.674σ, 0.674σ] represents the range of z-scores that can be mapped for the middle 50% of any set, i.e.
    Neutral: 25% - 75%

    The range [0.674σ, 1.645σ] & [-0.674σ, -1.645σ] corresponds to the next 20% of values falling into their respective buckets
    High: 75% - 95%
    Low: 5% - 25%

    And values that fall above or below +/- 1.645σ correspond to the last 5% of values
    Very High: 95% - 100%
    Very Low: 0% - 5%

Did this answer your question?