Now for the finding itself. We've defined the jump, built the control group, and left every failure inside the data - Carillion, Flybe, Afren and all the rest, each held to its last traded price. So once you strip out the hindsight and the survivorship, what does the evidence actually say?
It says two things, and both were sharper than I expected when we started. The first is about where on the StockRank ladder a jump has to land to mean anything. The second is about how big the jump has to be. If both are in place we find something useful - rules that have worked for twelve years, net of costs, and all delistings.
The edge lives in the 90+ decile, not the 80+ quintile
Start with the obvious question. If a stock climbs into the top of the StockRank universe, does it matter how far into the top it climbs? The rank runs 0 to 100. Rank 80 and above is the top quintile - the best one in five. Rank 90 and above is the top decile - the best one in ten.
The instinct is to treat 80 as the line that matters. It's a fine place to search for good shares, and 80+ shares have historically beaten the market handsomely. But a jump into the quintile, measured against the stocks already sitting there, barely moves the needle. The new arrivals didn't beat the incumbents.
The top 90+ decile is a different story. Here is the core table - QVM jumps that finish at rank 90 and above, 12-month mean returns, measured against the incumbents (the stocks already living there in the same weeks, sampled the same way, charged the same costs):
Read down the "vs incumbents" column and the shape leaps out. A +10 to +20 jump beat the incumbents by +5.3%, and that edge is statistically significant - its 95% confidence interval, +0.6% to +10.1%, clears zero, so we can say with reasonable confidence the effect is real and not a fluke of a noisy sample. The +20 to +40 jump did even better at +7.2%, though its interval just brushes zero, so we call it strong but stop short of "formally significant".
The lesson is precise, and it's the single most useful thing in this whole study: the edge is in jumps that finish at 90 or above, not 80 or above. Land in the decile, not just the quintile. Same idea, one rank threshold apart, and the gap between them is most of the alpha.
The sweet spot is +10 to +40
The second variable is the size of the jump itself, and it turns out to matter almost as much as the destination.
Trace the arrivals across the chart and returns climb with the size of the jump - from the feeble +5 to +10 nudge, up through +10 to +20, peaking with the +20 to +40 re-rating that carried the biggest edge in the table. A jump of that size is the market waking up to a real change: a profit upgrade, a step-change in earnings quality, a business that has quietly become a different animal and is only now being repriced for it. Big enough to be a genuine re-rating, not so violent that something is broken.
Because above +40, the whole thing falls apart. That is the wall on the other side of the sweet spot, and it's worth understanding properly rather than just noting it.
So the sweet spot has clear walls on both sides: a +10 to +40 point jump that lands at rank 90 or above. That single sentence is the buy signal the rest of this series is built on.
Why the biggest leaps fall apart
You would expect the very biggest jumps to be the very best signals. More movement, more conviction, more edge. The data says the opposite, and the reason is instructive.
The +40-and-above cohort returned just +8.0%, with an edge over incumbents of +0.7% - which is to say, no edge at all. These stocks did no better than the ones already sitting in the decile. And there are only 36 of them across twelve years, a thin and fragile sample, which itself tells you something: a rank leaping forty-plus points in a single week is an unusual, extreme event, and extreme events in the micro-cap world are often due to operating volatility rather than success.
Think about what it takes to move a composite Quality-Value-Momentum score by forty points at once. Frequently it's not a company ascending - it's a company that had already been poorly ranked due to distressed earnings, now staging an operating bounce off the floor. The tell is in the survivorship: in the +40-and-above bucket, the delisting rate roughly doubles. These are the names most likely to be gone entirely a year later.
That's the difference between a re-rating and a rebound. A +20 to +40 jump is usually a healthy business the market is slowly agreeing to pay more for. A +40-plus jump is too often a falling knife dressed up as a rocket. The strategy caps the jump size at +40 not out of neatness but because beyond it, the edge fades.
Small jumps are noise - and the creep is worse
Here's the part that surprised us most, and it sits at the other end of the scale. A small jump isn't merely a weaker signal. It can be a noisy and negative one.
Look back at the top row of the table: a +5 to +10 jump into the decile returned +11.1%, an edge of just +0.5% - effectively nothing. That's noise, and it's easy enough to dismiss. But now drop down a threshold, to jumps that only make it into the quintile. A +5 to +10 drift over the rank 80 line actually underperformed the incumbents by 1.9%.
That's worth re-reading. A stock easing gently over the 80 line - the classic "it's slowly getting better" story that feels so reasonable to buy - did worse than the stocks already sitting there. The creep isn't a faint signal you can take with a pinch of salt. On this evidence it's an anti-signal.
Why would a small, slow improvement be actively bad? My read is that a gentle drift over the line is usually the tail end of a move that has already played out. The concrete news that powers a real jump - the results beat, the upgrade - simply isn't there. What you're often buying instead is a stock nudging up on nothing in particular, more likely to mean-revert back down than to keep re-rating. The market hasn't just noticed this one; it noticed a while ago, and the easy profits have gone.
That contrast is the whole chapter in one line:
Buy the leap, not the creep.
A +10 to +40 jump into the top decile is a real event with fresh news behind it and months of drift still to come. A +5 to +10 shuffle over the line is noise at best - and a creep into the quintile is, on the numbers, worse than doing nothing.
Means beat medians - which is why it's a basket, not a bet
One honest caveat before we build rules on top of this. Every number in that table is a mean - an average across hundreds of events. And throughout this study, means beat medians everywhere, often by a wide margin. That's telling you exactly where the edge lives: in a long right tail. A minority of big winners drags the average up, while the typical result is far more modest. The edge is real, but it is emphatically not a promise that any given jumper will pay out. It's a base rate, and base rates only reward you across many tries.
You can see this the moment you run the sweet-spot signal as an actual portfolio. Take every +10 to +40 jump into the decile, held to its rank-break exit - and sort every outcome worst to best:
The red on the left is the losers you will hold, and there are plenty of them. The green on the right is the winners. Just 53% of positions made money - barely better than a coin toss. And yet the mean return was +18.8%, while the median was only +1.5%. Read those two numbers together and you understand the strategy completely: the typical stock does almost nothing, but a handful of multi-baggers on the far right carry the entire basket. Miss those, and you have nothing. The average is a real edge, but it's an edge that only exists as a distribution.
This is precisely why the jump effect is likely best earned as a system, rather than as a single conviction bet. If you cherry-pick one jumper, you're taking a roughly even-money bet. Run twenty of them at once, mechanically, and the maths turns in your favour: the losers are capped and diversified away, and you give yourself enough shots to actually catch a few winners that matter. It's also why the signal needs a sell rule as much as a buy rule - to cut the red left tail short before a Carillion or an Afren runs to zero inside your book, while letting the green right tail ride.
Buy rules and portfolio management find enough winners. The sell rules protect you from the ones that get away. You need both, and our full principle - the 90/80 Rule - is next.
Next: Chapter 4 - The 90/80 Rule. The complete entry and exit playbook, and the evidence that an exit on the rank break beats an exit on the calendar.