Factor Relevance and Timing Model
Test factor relevance, momentum, and mean reversion before allocating
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 98%
The model evaluates factors along three connected dimensions: relevance, hit rate, and timing. First, determine whether a factor consistently explains why groups of stocks perform differently; arbitrary classifications fail this test. Second, replace certainty with a probabilistic hit rate, recognizing that even the strongest factors may win only six or seven times out of ten. Third, examine factor momentum and mean reversion together. Performance during the previous year carries the strongest persistence, performance over two to three years remains informative but weaker, and periods beyond roughly three years may begin to support reversal. The resulting allocation favors factors that remain economically meaningful while balancing recent momentum against the possibility that an extended cycle is nearing its end.
Origin
Harin de Silva developed this perspective through decades of factor research across US, Japanese, European, and emerging-market equities. Extracted from Masters in Business.
Core principles
- 01A useful factor must explain differences in stock-group performance.
- 02Historical average returns do not guarantee near-term outperformance.
- 03Even strong factors fail frequently and for extended periods.
- 04Recent factor performance tends to persist before eventually reverting.
- 05Factor allocations should reflect both momentum and mean-reversion signals.
How to run it
- 1
Establish relevance
Test whether the factor meaningfully distinguishes outperforming stocks from underperforming stocks across different environments. Reject classifications that produce random or unstable differences.
Pro tip Ask whether the factor helps describe what is happening in the market even when its return sign is negative.
Watch out Do not confuse a definable stock characteristic with an economically meaningful factor.
- 2
Calculate the hit rate
Measure how often the factor delivers its expected result rather than relying only on its average historical premium. Use this probability to set realistic expectations.
Pro tip Model consecutive losses even when the factor wins 60% or 70% of the time.
Watch out A positive long-run average can conceal long and painful periods of underperformance.
- 3
Measure recent momentum
Review the factor's performance over the preceding year, where persistence appears strongest. Give successful factors more weight when the evidence remains robust.
Pro tip Use multiple observations rather than reacting to one strong month.
Watch out Do not assume factor momentum is the same thing as the conventional stock-price momentum factor.
- 4
Review the intermediate cycle
Examine the prior two to three years to determine whether momentum remains established or is weakening. Treat this evidence as useful but less powerful than the one-year signal.
Pro tip Compare the pattern across markets to test whether it reflects broad behavior rather than local noise.
Watch out A manager's recent success may simply reflect prolonged exposure to a favored factor.
- 5
Assess mean reversion
After roughly three years, test whether valuation, positioning, and extended performance support a reversal. Combine this assessment with recent momentum rather than using it alone.
Pro tip Look for factors that have begun improving after a prolonged period out of favor.
Watch out Being historically cheap or overdue does not reveal exactly when reversal will occur.
- 6
Allocate with loss tolerance
Set exposure based on signal strength and the investor's ability to endure repeated misses. Maintain a horizon long enough for the factor thesis to play out.
Pro tip Stress-test the allocation against several consecutive years of disappointing results.
Watch out Do not promise that a diversified bundle of factors will beat the market in every cycle.
In the wild
A portfolio team observes that value has underperformed for several years but has recently delivered four to six months of stronger results. The team treats the long slump as mean-reversion support and the recent improvement as an emerging momentum signal, then introduces a measured value tilt rather than making an all-or-nothing bet.
→ The allocation reflects both an improving trend and the uncertainty surrounding its duration.
An allocator finds that a celebrated growth manager's returns closely track earnings-acceleration stocks. Instead of attributing all performance to skill, the allocator measures the factor's cycle, separates factor exposure from security selection, and adjusts expectations for what may happen when that style falls out of favor.
→ Manager skill is evaluated separately from a favorable factor tailwind.
Common mistakes
Treating averages as promises
Investors see a positive long-run factor premium and assume it should remain positive over their own investment horizon. This ignores hit rates and sequences of losses.
Attributing factor returns to genius
A manager may look uniquely skilled when a concentrated factor exposure has simply remained in favor. Diagnose the portfolio's underlying factor loadings before judging skill.
Using momentum without reversal risk
Recent winners can continue working, but persistence weakens over time and may eventually reverse. Momentum and mean reversion must be considered together.
Is it for you?
Best for
It is best for quantitative investors, asset allocators, and manager researchers making medium- or long-horizon factor decisions.
Not ideal for
It is not ideal for investors seeking precise short-term market-timing signals or guaranteed factor returns.
From the transcript
“So I really look at it from two dimensions. One is the does the factor matter?”
“My conjecture is that even the best factors will outperform six or seven out of out of ten times.”
“The persistence is very strong over one year, less strong over two to three years. And then after three years you actually start to see…”
From the episode
Harin de Silva on Portfolio Management (Podcast)