AiLA’s Product Design - A Simple and Systematic Approach

AiLA’s Design Team

Designers

Arun
  • Prior Experience in Shinsei Bank, ICICI
  • Graduated from IIM, Calcutta, India
Brandon
  • Experience at Barclays and Standard Chartered Bank
  • Graduated from London School of Economics & Political Science, UK
Emmanuel
  • Prior quant experience genetic research in Japan
  • Graduated with PhD from Harvard, USA
Kenneth*
  • Prior experience at Singapore Refining
  • Graduated from the University of British Columbia, Canada
Pravin
  • Prior trading experience at ADM, Citigroup and Deutsche Bank
  • Graduated from Harvard, USA
Stefan
  • Prior research experience at Winton Asset Management
  • Graduated with PhD from Lund University, Sweden

Advisors/Directors

Anthony Kettinger
  • Experience at Cargill and ADM
  • Graduated from University of Wisconsin
Francisco Blanch
  • Experience at BAML and Goldman Sachs
  • Graduated from Harvard
Kurt Chapman
  • Experience at Morgan Stanley and Mercuria Trading
  • Graduated from Harvard
Noriaki Sakamoto
  • Experience at McKinsey and Co. and METI-Japan
  • Graduated from Columbia

AiLA’s Systematic Approach to Design

Public Data & Open Source Models BUT Interpretation of Risk and Returns with a commodity lens

QUANTITATIVE METHODS
  • Treat every futures expiry on an asset separately within the same curve. This gives different returns within the same commodity market. Example: WTI Jun and WTI Dec have separate return streams thus over all WTI returns from 12 months on the curve yields low volatility in returns
  • Embed risk parameters within the performance such that if returns are negative then exit and re-allocate when opportunities arise. This results in opportunistic allocation instead of constantly allocated
  • Avoid forecasting of prices. Instead focus on timing of allocation which includes when to allocate and exit risk.
FUNDAMENTAL BACKGROUND
  • Focus on commodities where team has a strong background from fundamental trading perspective
  • Use data which are publicly available but transform into meaningful factors. Example: Use the distance between first notice date and expiry date of the contract to asses liquidity drops
  • Focus on term structure of the asset as an information highway. Example why is March/April RBOB or July/Nov Beans more important than other spread months in RBOB and Soybeans
  • Respect the behaviour of assets across different months to extract seasonality and the drivers for seasonality

DOWNSTREAM PORTFOLIO DESIGN WITH FOCUS ON RISK MANAGEMENT 35%

MODELING AND QUANTITATIVE METHODS 15%

UPSTREAM DESIGN FUNDAMENTAL BACKGROUND 45%

DATA 5%

Relative Importance of building blocks

BACKGROUND

Amplification of Small Edge Opportunities

Ambition
  • Like many systematic investment strategies,
  • Small individual profit/loss edge opportunities,
  • Being amplified to relevant long-term profits,
  • By independently repeated many times.
Application
  • Predict small edge commodity opportunities,
  • Repeated across many individual commodity markets as well as futures contracts.
  • Independent repetitions also from short holding periods, however, balanced w.r.t. transaction costs.
For example
  • Return distribution for an individual asset and a single day has tiny PnL edge (2.5 bps) compared to the typical return size (1%).
  • Given 100 daily returns typically 51 are positive and 49 are negative.
  • However, same opportunity repeated every day across 25 independent assets result in a Sharpe ratio ~ 2, i.e. 20% annual return at 10% risk.

SMALL EDGE OPPORTUNITIES

AiLA’s Approach to Data Features

Ambition
  • Commodity returns expected to be driven by different micro as well as macro factors.
  • Factors partly represented in market as well as fundamental data.
  • Try to identify dominant factors in order to motivate allocation opportunities.
Application
  • Construct data features as proxy for price factors.
  • Train model to predict allocation opportunities, i.e. favorable risk/reward forecast associated with characteristic factor signature in features data.
  • Price typically impacted by multiple factors and “noise” yielding only small PnL edge opportunities.
For example
  • During Sep and Oct in 2017 the RBOB price was partly impacted by a supply (micro) factor.
  • The refinery disruption by hurricane Harvey was anticipated to decrease short-term supply.
  • The tighter supply was reflected in the front contract spreads, i.e. part of micro features.

UN-CORRELATED COMBINATIONS

Combination of Un-Correlated Opportunities

Ambition
  • Equivalent opportunities should contribute equally to the portfolio return.
  • Dependent opportunities should not result in duplicated allocations to “same” opportunity.
  • Better opportunities should contribute more to portfolio return, but difficult to estimate.
Application
  • Equalize allocations w.r.t. risk, i.e., prevent concentration of risk to more volatile assets.
  • Consider allocations together with market return correlation, i.e., prevent concentration of risk to similar positions in highly correlated assets, such as same commodity contracts.
  • Typically treat individual opportunity predictions as equivalent.
For example
  • Many independent opportunities result in higher risk adjusted portfolio returns.
  • However, increased market correlation result in lower number of independent opportunities, for a given number of actual assets allocations, as well as increased portfolio volatility.
  • Micro factors typically more important than macro for individual allocation decisions.
  • Therefore, strategy expected to perform better during periods dominated by idiosyncratic micro factors, rather than common macro factors, such as interest rates or the USD.