A finding becomes useful only when it tells you what to do - both when to buy and when to sell. The chapters so far have built the entry side: a +10 to +40 point QVM jump that lands at a StockRank of 90 or above. This chapter completes the playbook and, just as importantly, proves the exit. "When to sell" is the half most people skip. It turns out to be where a surprising amount of the return actually lives.
The rule in full
Here is the whole playbook, stated as a weekly, mechanical routine:
Buy a +10 to +40 point QVM jump that lands at a StockRank of 90 or above.Hold while the rank stays at 80 or above.Sell on the first weekly close back below 80.
That's the 90/80 Rule. Buy into the top decile, hold through the top quintile, exit when the rank breaks the floor. In practice you run one hard long-stop alongside it - if a name has neither broken 80 nor gone anywhere after two years, it's wise to let it go and recycle capital.
This was an idea we borrowed from an academic's work. Christoph Reschenhofer's earlier work gave us the idea for a 90/70 rule - buy at 90, sell below 70. We re-ran it on our own data, tested where the sell floor should sit, and found that tightening the line from 70 to 80 improved the risk-adjusted result. The prior was sound, but we sharpened it on our own data. Credit where it's due, and evidence where it counts.
The rank beats the clock
The most interesting answer from the whole study wasn't where to buy, it was how to sell. Should you hold for a fixed period - say six months - or should you let the rank itself tell you when to leave?
We tested both, net of full costs, on the same entries and the same stocks. The only thing that changed was the sell discipline. Here is the comparison, run on the sub-£350m core with a £10m tradeability floor:
Read the top and third rows together, because that's the finding. The 90/80 Rule - selling on the first weekly close below 80 - delivered +16.5% CAGR at a Sharpe of 1.25, on a median hold of about 277 days. A simple fixed six-month hold managed +12.0% CAGR at a Sharpe of 0.79.
Letting the rank decide your exit added about 4% a year of compound return and lifted the Sharpe from 0.79 to 1.25. That is an enormous difference for a change that costs nothing to implement. Same buys, same universe, same costs - only the reason you sell is different. The calendar sells on a date it picked in advance and knows nothing. The rank sells when the story has changed - the same principle behind using the StockRanks as buy and sell thresholds across a whole portfolio.
Notice too that the calendar isn't useless - holding twelve months rather than six lifts the return to +13.5% because it stops you cutting winners in the middle of their drift. But even a fixed holding period loses to the rank, because a fixed period can't tell the difference between a name that's still ranking well and one that's falling apart. The rank can.
Why the rank knows first
This is the second rule of thumb in the series, and it's worth saying plainly: the rank knows before the calendar does.
Remember what a StockRank actually is. It's a live composite of Quality, Value and Momentum, updating as the underlying data updates. When a thesis breaks, it rarely breaks all at once in the share price. Earnings start to miss. Brokers trim their forecasts. Momentum rolls over. Each of those feeds the rank, so the composite begins sliding through 80 while the price is still only halfway to admitting the damage. A weekly close below the quintile line is the market's polite first warning that the fundamentals have turned - issued, on average, before the full fall shows up in the price.
The same adaptiveness works in your favour on the winners. A median hold of 277 days is longer than a six-month rule and shorter than a rigid year, because the rank lets a name that keeps earning its place keep its slot. Yü Group, which we'll meet later, held the top decile for roughly a year while it re-rated - the exit didn't touch it until the rank finally slipped, deep into the best part of the move. Cut the losers early, let the winners ride: the rank does both, not because it's clever, but because it's reading real fundamentals.
The departures carry information
We ran a separate study on the falls - stocks leaving the decile rather than entering it. Names that fell 5 to 20 points below the 80 line went on to underperform the ones that held their rank by 3% to 4% over the following year.
That symmetry is the point. In Chapter 3 we saw that arrivals carry information - a genuine jump into the decile is a signal worth acting on. This is the mirror image: departures carry information too. A rank breaking 80 isn't a tidy stop we invented to feel disciplined. It marks a real, measurable divergence in what happens next. The exit earns its keep because the market is telling you something on the way out, exactly as it told you something on the way in.
A plateau, not a knife-edge
We swept the sell floor at every rank from 50 to 90, and every floor between 70 and 90 earned roughly 15% to 17% net. There's no cliff at 80 where the result suddenly appears and vanishes either side. It's a broad, gently sloping plateau. Reschenhofer's 90/70 works. Tightening to 80 works a touch better. You are standing on level ground, not balancing on a pin.
Why does that matter so much? Because a rule that only works at exactly one threshold is usually an accident of the data - a number the backtest fitted to noise. A plateau is the opposite. It says the effect is robust to the arbitrary choices you make around it, which is the signature of something real. And it held up under the harder test too: the ordering survived in both halves of a 2014/2020 split, so it isn't an artefact of one lucky era or one benign market.
Don't be tempted by a 90/90 rule. We found that collapsed the average holding period from roughly 9 months (for 90/80) to 3 months. That's impractical for any sane private investor, due to the whipsaws and dealing activity.
For those who'd rather trade less
The standard rule sells on any close below 80, which can mean trading in and out of a wobbly name more than you'd like. If you'd rather touch the portfolio less often - a taxable account, a smaller book, or simply a temperament that hates churn - there's a lower-turnover variant that only acts on a decisive break.
It sells only on a sharp fall: a drop of 10 or more rank points that takes a name out of the top decile and below 85. Wait for the rank to fall hard, in other words, rather than reacting to every dip across the line. That version earned a Sharpe of 1.26 with a shallower -18% maximum drawdown and roughly 12-month holds. Fewer trades, a slightly gentler ride, essentially the same reward. It's a reminder that the rule has a little give in it - the exit is a principle you can tune to how you actually want to invest, not a single brittle setting.
One temptation to resist
The chart above does offer one genuine trap. If tightening the sell floor helps a little, why not tighten the buy bar too - only take the very best jumps, the ones that land at 95 or above?
On paper it looks spectacular. The headline CAGR jumps to a flattering, eye-watering number. But it's a mirage. That strategy runs on a median of just nine concurrent positions. Nine. That's not a portfolio, it's a handful of bets, and its "outperformance" is single-stock luck - a worse Sharpe once you account for the risk, and results that fall apart the moment you split the sample periods in two. Don't chase 95. The robust, repeatable edge is the broad 90/80 Rule, run as a diversified basket, where no single name can make or break you.
So we have a complete rule now, with an exit proven to beat the calendar, a comforting plateau of robustness beneath it, and one seductive shortcut safely disarmed. But a rule that works on average across the whole market can still be very uneven under the surface. The next question is where this edge actually lives.
Next: Chapter 5 - Where the jump effect lives. Why this is a small-cap phenomenon, which sectors carry it, and which factors do the work.