Asymmetric Long-Short Signal Design
Use different empirical signals to identify compounders and melting ice cubes
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 99%
Asymmetric Long-Short Signal Design begins with the premise that winners and losers emerge through different mechanisms. To find compounders, Parker tested measures including gross-margin expansion, net-income growth, earnings growth, revisions, and price momentum; consistent gross-margin expansion showed the strongest subsequent stock performance. The short-side inverse was not margin contraction. The strongest melting-ice-cube candidates combined large accruals—a disconnect between earnings and cash flow driven by capital expenditure, inventory, or intangibles—with poor price performance relative to industry peers. Additional share loss and margin-contraction variables did not improve the result. The reusable lesson is to define each investment archetype independently, empirically test its candidate signals, and retain only variables that improve future return separation rather than imposing intuitive symmetry.
Origin
Parker developed the screens at Trivariate Research after clients repeatedly asked for compounder longs and melting-ice-cube shorts.
Core principles
- 01Long and short selection are not mirror-image problems
- 02Consistent gross-margin expansion can identify durable compounders
- 03Margin contraction alone does not identify the best shorts
- 04Accruals reveal gaps between earnings and cash flow
- 05Weak relative momentum confirms that deterioration is reaching the market
How to run it
- 1
Define each archetype
Describe what a successful long and short should represent without assuming they are opposites. Translate each concept into measurable candidate variables.
Pro tip Use investor language such as compounder only as a starting hypothesis.
Watch out A vague label can produce a screen that merely confirms intuition.
- 2
Test the long signals
Evaluate persistence in gross margins, net income, earnings, revisions, and momentum against subsequent returns. Rank signals by their actual return separation.
Pro tip Test both historical consistency and forward forecasts.
Watch out Do not select a signal because it sounds fundamental.
- 3
Build the compounder basket
Select businesses with consistent historical gross-margin expansion and forecast expansion ahead. Recheck whether inflation or input costs threaten that trajectory.
Pro tip Gross margins can expose pricing power more clearly than revenue growth.
Watch out Forecast margin expansion may already be fully discounted.
- 4
Test short signals independently
Evaluate accruals, cash conversion, momentum, share loss, and margin contraction without forcing the long model into reverse. Identify the combination with the strongest historical underperformance.
Pro tip Measure performance relative to industry peers.
Watch out A weak business can remain a poor short while momentum is positive.
- 5
Diagnose accruals
Determine whether the earnings-cash-flow gap comes from capital expenditure, inventory accumulation, or acquired intangibles. Assess whether those investments are likely to earn acceptable returns.
Pro tip Study several quarters to distinguish a project cycle from persistent deterioration.
Watch out Not every large investment is value destructive.
- 6
Remove non-contributing variables
Retain only signals that improve the spread between future winners and losers. Avoid complexity that adds explanation but no predictive value.
Pro tip Compare incremental performance after adding each variable.
Watch out More conditions can overfit historical data.
In the wild
Trivariate tested consistent gross-margin expansion, net-income expansion, earnings growth, revisions, and momentum. Gross-margin expansion had the strongest association with subsequent stock performance, so the firm built a basket using historical consistency and forecast expansion.
→ The long screen targeted companies demonstrating durable pricing power and improving unit economics.
The short screen combined the largest accruals over the prior three quarters with underperformance versus industry peers. Adding share loss and margin contraction did not improve the historical result.
→ The model focused on deteriorating cash quality confirmed by weak relative price behavior.
Common mistakes
Mirroring the long model
The inverse of a productive long signal may not identify the strongest shorts.
Ignoring cash conversion
Reported earnings can conceal inventory builds, heavy capital spending, or uncertain acquired intangibles.
Adding intuitive but useless factors
Share loss or margin contraction should be excluded when they fail to improve predictive separation.
Is it for you?
Best for
It is best for systematic or hybrid investors constructing long-short equity baskets.
Not ideal for
It is not ideal when financial histories are too short or unreliable to measure persistence and accruals.
From the transcript
“identifying longs and identifying shorts use different signals.”
“And the answer was gross margin expansion.”
“The two things that matter the most were accruals, which would be disconnects me and anuras in cash flow, which were driven by cappex inventory…”
From the episode
Adam Parker on Strategies and Valuations
Adam Parker