Portfolio Engineering & Systematic Research We don’t assume a different investment means a different source of return.
Portfolio engineering is the deliberate design of a portfolio so its return drivers, position sizes, exits, hedges and risks work toward one objective, instead of sitting side by side as a collection of investments.
By Mike Shell, Founder & Chief Investment Officer. Updated October 2026.
Different labels can be one bet. Independence is measured, not assumed.
Different rules can still share the same market risk.
Two strategies can own different things and still fail for the same reason.
And two strategies can trade the same market with different rules and behave nothing alike. Neither an edge nor independence can be read from the labels. So every investment system has to answer five questions before it gets capital.
- 01What has to happen for it to make money?The condition that pays: a trend that persists, a leader that keeps leading, a decline it gains from.
- 02What is the evidence, and why should it last?A reason it works, not just a period when it did.
- 03When do we get in, and what proves us wrong?The entry, and the exit that ends the idea, decided before the trade.
- 04How do size and costs change the payoff?A good idea at the wrong size, or after trading costs, can still be a bad investment.
- 05How does it behave next to everything else, especially under stress?If it depends on the same condition as what you already own, it isn’t adding a source of return. It’s adding size.
A promising backtest isn’t an investment edge.
We look for a credible reason a system should make money, then try to break it: other market periods, trading costs, changing correlations and bad scenarios. A system that only looked good in the market it was designed around doesn’t earn a permanent place.
From idea to capital.
- 01A reasonWhy should it make money, and why should that last?
- 02Honest dataWas the information actually known at the time?
- 03New dataDoes it hold up outside the years it was built on?
- 04Costs and stressWhat do trading costs, price gaps and bad markets do to it?
- 05The whole portfolioDoes it add a different source of return, or more of the same?
- 06Watched liveWhat would make us cut it back or stop it?
Reject or reviseMost ideas stop at one of these steps. Saying no is part of the work.
Built to fit the past.
Try enough settings and something will look brilliant on old data. We judge it on data it hasn’t seen.
Using tomorrow’s news.
A test that uses information nobody had at the time will always look better than reality.
Missing the losers.
Data that leaves out the companies and funds that failed makes every strategy look safer.
Settings that worked once.
If small changes to the rules ruin the result, the result was luck.
Everything falling together.
Holdings that move differently in calm markets can fall together in a panic. We test the bad weeks, not the average.
Costs that weren’t counted.
Trading costs, slippage and taxes turn many good-looking ideas into ordinary ones.
A model can be wrong. Our process has to assume it sometimes will be.
What has to happen for it to make money?
Assets are the means, not the idea. What matters is the mechanism: the condition that pays, and the condition that hurts. These are the kinds of systems we research. Whether one belongs in your portfolio depends on what the rest of your money already depends on.
| Mechanism | What has to happen | When it struggles | Where it can apply |
|---|---|---|---|
| Market participation | Prices rise over time. | Long, deep declines. | Stocks, bonds |
| Trend following | Prices make sustained moves, up or down. | Markets chop back and forth. | Stocks, bonds, commodities, currencies |
| Rotation | Leaders keep leading for a while. | Leadership reverses suddenly. | Sectors, countries, asset classes |
| Defensive | A large decline happens. | Markets stay calm and it costs money to hold. | Index hedges, cash |
| Outside specialists | Their edge holds after their fees. | The edge fades, or the fees win. | Liquid alternative managers |
| Rejected | Anything that adds the same risk the portfolio already carries, under a different name. | ||
Correlations change, so we measure them when we decide.
Equal dollars don’t mean equal risk.
A position can look reasonable on its own and still make the portfolio worse. So we count risk the same way across every position: the dollars between today’s price and the planned exit.
| Position | Invested | Distance to exit | Open risk |
|---|---|---|---|
| A | $1,000,000 | 8% | $80,000 |
| B | $1,000,000 | 3% | $30,000 |
Same capital, more than twice the risk. Add up the open risk on every position and you have one number for the whole portfolio. Five positions with $40,000 each is $200,000 open in all.
That isn’t a contractual maximum loss. Prices can gap through exits, correlations can rise and markets can stop trading smoothly. It is a common language for risk, so we can see where it is concentrated and how much more the portfolio can take.
Ask what happens when the normal rules fail.
We study how growth, inflation, interest rates, debt and liquidity have moved markets across history, and how the systems built on those markets behaved. Then we put our assumptions through the scenarios that would hurt most.
Growth turns down.
Earnings fall, and so do the stocks that depended on them.
Inflation rises.
Interest rates follow, and bonds can fall with stocks, as they did in 2022.
Money gets tight.
Lending pulls back and investors sell what they can, not what they want to.
A trend reverses hard.
A move that ran for months gives back weeks of gains in days.
Prices gap overnight.
The market opens far from where it closed, past any planned exit.
Everything correlates.
In a panic, holdings that usually move differently fall together.
Scenarios aren’t forecasts. They show what a portfolio would have to survive, so the decisions can be made before it does.
Decided in advance, not bought in a hurry.
Sometimes the right hedge is simply less exposure. Sometimes it is cash. Sometimes it is a position built to gain in a large decline. The question is always the same: what loss are we trying to change, and what do we give up to change it?
A hedge that costs too much damages compounding. A hedge on the wrong exposure creates false confidence. A hedge that is too small can gain several hundred percent and still do little for the portfolio. So we judge a hedge by what it does to the whole portfolio, not by what it does on its own.
Systematic doesn’t mean unattended.
Rules run the process, so a decision made in a calm market is the one carried out in a falling one. Mike approves each client’s mandate and the positions in it, watches for exceptions, and writes down the reasons for significant decisions.
I’ve never thought of a portfolio as a pie chart. I want to know what makes every position work, what makes it fail, what it can cost us, and what happens when several fail at the same time.
Mike Shell
We use quantitative research and software to test ideas and watch risk. The assumptions, the evidence, the mandate and the decisions are checked by a person, and Mike is responsible for them.
Questions about portfolio engineering.
What is portfolio engineering?
The deliberate design of a portfolio so its return drivers, position sizes, exits, hedges and risks work together toward one objective, instead of sitting side by side as a collection of investments. At Shell Capital Management it is how ASYMMETRY® Managed Portfolios are built; portfolio management is the continuing work of running them.
How is it different from asset allocation?
Asset allocation answers where the money is: so much in stocks, bonds and alternatives. Portfolio engineering asks what the portfolio can do in different conditions: what drives each return, how much each position can lose, how positions behave together, and what happens when they all face the same problem.
Is it the same as portfolio optimization?
No. An optimizer produces a precise answer from historical averages. Engineering assumes returns, volatility and correlations will change, so the design has to keep working when the inputs turn out to be wrong.
Do more holdings mean better diversification?
Not necessarily. Twenty securities that depend on the same condition can be less diversified than five return streams that depend on different ones. The count of holdings isn’t the count of bets.
Can portfolio engineering eliminate drawdowns?
No. No method removes market risk. The purpose is to decide which risks are worth taking, how large they are, how they interact, and what can be done to keep a bad outcome survivable.
Do you show backtest results?
Not on this website. A backtest is a hypothetical result, and it can only be shown with its dates, assumptions, costs and limits explained. We would rather show you how we decide, and talk through the evidence in a private conversation.
How might this fit your capital?
Tell us what the money has to do, and what it already depends on. We’ll tell you what we see, and whether we’re the right fit.
If you write, I’ll read it myself.
Mike Shell, Founder & Chief Investment Officer
Related: ASYMMETRY® Managed Portfolios, Predefined Risk, Adaptive Portfolio Management, Convexity and Return Drivers.