Rebalancing Context

Background

• The AILA strategies by default rebalance daily at close.
• The individual AILA allocation models can vary in terms of holding period from a few days up to a month.
• However, typically an asset model would target a holding in the order of 10 days.
• In this short study we address what performance impact could be expected from scenarios different to the daily rebalancing process.

Weekly Weights

• Update target weight only one time per week.
• Except if DTE < limit value, then exit position.
• Only update target weight if change larger than threshold value.
• Respecting all rebalancing, position and portfolio caps.
• Actual weight executed at close on the next business day.

Weekly Rebalancing

Expected Impact

• For a diversified AILA portfolio, e.g. here 158 assets across Ags, Metals and Energy Commodities.
• Rebalancing only one time per week has a substantial impact on the performance compared with daily rebalancing.
• The impact is reduced to a large extent by increasing the rebalancing frequency to three business days per week.
• Typically, the weekly weights do not increase the periods with non-zero holdings, e.g. weekly and daily weights hold positions roughly over the same period.
• However, weekly weights does not change direction as quickly as daily weights, e.g. from net long to short, or vice versa, in a given commodity market.

Expected Impact

• Investigate sign-difference impact in two different ways.
• Adapt the weekly functionality to allow daily rebalancing in case the sign change, until the end of the given week (flip-sign daily).
• Flip-sign results indicate significant fraction of effect attributed to opposite signs.
• Breakdown weekly/daily P&L in three type of categories:
• 1Z: days where one of the weekly or daily weights are zero (only one have position).
• OS: days where weights are non-zero with opposite sign.
• SS: days where weights are non-zero with same sign
• For OS weight days, the daily weight has a large advantage in average.
• However, SS weight days are 10x as common, so smaller advantage of SS and 1Z days also becomes relevant overall.​

Delayed Rebalancing

Expected Impact

• The impact from delaying the execution by one day is also indicated to be sizeable.
• The default process would be to execute the weights at close on the following day w.r.t. when the weights were calculated (t+1), and therefore the additional delay would correspond to t+2 days.
• The reduction of the Sharpe ratio by about 0.5 in the test portfolio show (top right) is potentially at the high end of the range, however, still comparable to the effect seen for other AILA portfolios (bottom right).
• The result seems reasonable given that the first day of P&L is lost in a portfolio where many of the asset models tend to have total holding periods of 10 days or less.