Every factor strategy has a bad decade in it somewhere. Value spent 2015 to 2020 losing to growth across developed markets. Momentum crashes at market inflections. Low volatility gets left behind in violent recoveries. None of these are flaws in the construction — they are the strategies working as designed in an environment that does not suit them.
The useful observation is that the bad periods do not coincide.
Factor cyclicality is the whole argument
Consider what happens at a market bottom.
Momentum is at its worst. A momentum book at the bottom of a bear market is holding whatever was defensive on the way down and is underweight everything beaten up. When the market turns, the beaten-up names rebound hardest and the defensive names lag. Daniel and Moskowitz call this a momentum crash, and it is concentrated in exactly the weeks after a bottom.
Value is at its best. Value is overweight the beaten-up names — that is what makes them cheap — and captures the rebound directly.
Low volatility is at its worst, for the same reason as momentum: structurally underweight the high-beta cyclicals that lead a recovery.
Now consider a slow grinding bear market. Momentum rotates defensive and holds up relatively well. Low volatility does what it says. Value gets destroyed, because cheap gets cheaper and some of it does not come back.
The factors are not uncorrelated in returns — they are all long equity, so they all fall together in a crash. What is uncorrelated is relative performance, and specifically the timing of their worst stretches. That is enough for the blend to have a shallower worst drawdown than any individual leg.
The behavioural argument matters more than the statistical one
The statistical case for blending is that the combination has a higher Sharpe ratio than the average of its parts. That is true and it is the smaller half of the argument.
The larger half is that the strategy you can actually hold is the one whose worst period you can survive.
A single-factor investor who deploys momentum and then lives through a momentum crash sees their strategy underperform a plain index fund for eighteen months while doing exactly what it was designed to do. Most people abandon it. They abandon it near the bottom of the relative drawdown, lock in the underperformance, and switch to whatever has been working — which is usually the factor about to have its own bad stretch.
A blend never gives the investor the same intensity of that experience, because at any moment one or two of the sleeves are working. It underperforms less badly, less often, for shorter. That is a lower-variance experience, and it is what turns a strategy on paper into a position someone actually holds for a decade.
This is why the case for multi-factor is stronger for an individual investor managing their own money than for an institution with a mandate that will not let them capitulate.
How the blend is built
A multi-factor portfolio can be constructed two ways, and they are not equivalent.
Integrated (composite score). Compute a momentum z-score, a quality z-score and a low-volatility z-score for every stock, average them into one number, and hold the top N by that composite.
Mixed (sleeve blend). Run three independent strategies, each selecting and weighting its own names, then combine the resulting portfolios by allocation.
The integrated approach can produce holdings that are mediocre on every factor and excellent at none — a stock ranking in the 60th percentile on all three scores the same as one ranking 95th, 95th and 10th. Sometimes that is what you want. Often it is not, because factor premia are concentrated in the tails and a portfolio of universally-average stocks captures none of them.
The sleeve approach keeps each factor’s selection pure. The momentum sleeve holds genuine momentum names, chosen by momentum, weighted by momentum logic. When two sleeves independently pick the same stock, its weights add — so agreement across factors produces concentration organically rather than by construction.
The template uses the sleeve approach, with an allocator node that normalises each sleeve, scales it by its allocation, and unions the results.
Choosing allocations
The allocations here are 40% momentum, 35% quality, 25% low volatility, and it is worth being honest that any set of numbers in this region is defensible. The evidence does not support fine-tuning:
- Equal weighting across sleeves is the most robust choice and hardest to overfit.
- Tilting toward momentum reflects that it has historically had the largest premium and the highest turnover, so it contributes most per unit of allocation.
- Tilting toward low volatility reduces overall portfolio risk more than the other two.
What you should not do is optimise these weights on historical returns. Three parameters fitted to a fifteen-year sample will find the combination that best explains that particular sample, which is the definition of overfitting. If you change them, change them for a reason you can state before looking at the result.
Each sleeve keeps its own sizing
A detail that is easy to get wrong: each sleeve applies its own weighting scheme before the blend.
The low-volatility sleeve uses inverse-volatility weighting, because that is intrinsic to how low volatility works as a strategy — equal-weighting a low-vol basket hands the largest risk contribution to its most volatile member. The momentum and quality sleeves use equal weighting.
If you instead selected all names from all sleeves and then applied one weighting scheme across the union, you would have destroyed the low-volatility sleeve’s construction. The sleeve’s weighting is part of the strategy, not a post-processing step.
What the blend does not fix
It is still fully long equity. All three sleeves are long-only equity strategies. In a market-wide crash, all three fall. The blend reduces relative drawdown against a benchmark, not absolute drawdown against zero. If you want protection from the latter you need a market-level filter or a different asset class, not more factors.
Overlap concentrates risk quietly. When momentum and quality independently select the same stock, it receives both weights. That is usually desirable, but it means the effective position in a name both factors like can be considerably larger than a naive read of “twenty names per sleeve” suggests. A position cap is worth adding if you scale this up.
Costs multiply. Three sleeves rebalancing quarterly generate more turnover than one, and any name entering or leaving any sleeve trades. The blend is not free.
Try it
The Multi-Factor Blend template runs three sleeves on the Nifty 200 — 12-1 momentum, return on equity, and realised volatility — each selecting twenty names with its own weighting, blended 40/35/25 and rebalanced quarterly.
The comparison that makes the point: run it against the standalone Momentum 30, Quality 30 and Low Volatility 30 templates over the same window. The blend will rarely have the best return. It should have the shallowest worst drawdown — and if you are managing your own money, that is the number that decides whether you are still invested in year seven.
Further reading
Each sleeve has its own post, and the construction detail inside a sleeve matters more than the blend weights across them:
- Momentum investing in India
- The quality factor: screening for businesses, not charts
- The low volatility anomaly
- Value screens that survive contact with Indian data
Glossary: factor investing, rank correlation, z-score, max drawdown.