Fixed-Risk Products

Motivation

  • The fixed-risk (FRISK) portfolio provides an alternative approach to the other AiLA products, by being defined in terms of asset risk rather than notional allocations.
  • The intention is to allow for input quantities that are, central in the context of risk-adjusted performance and which prioritize the risk profile of the portfolio rather than different allocation criteria.
  • In a scenario with independent assets which all have equal E [ r % / 𝜎 % ] classical arguments, such as the principles of Kelly betting, would suggest to keep a similar risk assigned to each asset.
  • However, often the assets are not independent nor the betting edge equal, and for this reason a number of static parameters are available for each asset. This initial setup will possibly develop further.
  • It should be noted that due to the non-linear relation between risk and allocation, an evenly distributed portfolio in terms of risk may appear much less even when allocation weights are summed and compared.

Parameters (per assets)

  • Asset Risk: annual risk target of individual asset when being allocated, as percent of total capital.
  • Asset Group: sector or risk group name associated with asset, used to calculate risk scale factor, i.e. option to split assets into separately treated groups.
  • L/S Ratio: ratio of long and short signal weights, i.e. for a given asset 0.5 result in short weights being half of the long weights, if all else is being equal.
  • Position: Allow only long (L), short (S) or both (LS).

Portfolio Risk

  • The fixed risk target assigned per asset, result in an asset group as well as portfolio risk directly related to the number of asset allocations.
  • Asset groups are treated independently. However within an asset group, the variance is estimated using a risk model, and the group allocations are scaled down in case of excess variance from highly correlated assets.
  • Example: Metal Asset Group
    • The risk scaling reduce the allocation weights at times with many similar assets, e.g. several gold contracts allocated at the same time.
    • This results in a more uniform portfolio allocation as well as risk profile over time.
    • The scale factor maintain a maximum risk equal to that of an independent set of assets, which results in a group risk similar to that expected from the number of allocated assets.
  • Good correspondence between risk assigned and achieved, however, number of allocations can vary significantly over time.

Final Properties

  • The portfolio risk intentionally increase with the number of prevailing allocation signals, i.e. analogues to the number of equivalent bets, whereas the group risk scaling address the possibility of highly correlated assets, i.e. analogues to several bets effectively being the same bet.
  • However, periods with unusually large number of allocation signals can lead to significantly higher leverage, and therefore the portfolio allocations are scaled in order not to exceed a specified max value of capital allocated.
  • After the FRISK portfolio is constructed, the same execution logic as used by the automated index methodology is applied, where liquidity constraints etc. will impact the results and limit the risk of certain assets.

Conclusions

  • The so-called AiLA fixed-risk (FRISK) portfolio methodology allow for a simple framework to define a product based on risk as an input, rather than being a consequence from different allocation criteria.
  • The current FRISK framework has only a limited amount of automated functionality, and therefore require a clearer a priori view on the risk profile of the product.
  • On the other hand, it offers high granularity in terms of its asset configuration, allowing for a detailed definition established on principles and results obtained externally of the product code.
  • The FRISK portfolio framework might therefore be well suited for either exploring alternative definitions than those possible within the traditional AiLA products, or more specific or evolving products.