Methodology
Last updated: 9 August 2026
How The Microcap Minute Global decides what appears in an issue, written out so a reader can check the work rather than take it on trust.
There are two steps and only one of them is mechanical. That distinction runs through everything below, so it is worth stating first: the tests applied to a company are fixed and repeatable, and the choice of what goes into those tests is a human reading of a chart.
What is being measured
One thing: how a sector, a fund or a company is moving against the S&P 500.
Not whether it is cheap. Not whether the business is good. Not what it will do next. A company can appear in an issue while losing money, and several have.
Everything else in an issue, the margins, the returns on capital, the growth, is reported so a reader can judge the business for themselves. It plays no part in selection.
Three layers
An issue works downward: the eleven US sectors, then the industry funds inside them, then the companies.
1. Sectors
Every week we read a Relative Rotation Graph of the eleven Select Sector SPDR funds against the S&P 500, on charts from StockCharts.com.
How a rotation graph is read
A rotation graph places two measurements on one picture.
Across is strength against the benchmark. Up is whether that strength is building or fading. Both are centred on 100, which means behaving exactly like the index. So the picture divides into four:
strength buildingstrength fadingstronger than the indexLeadingWeakeningweaker than the indexImprovingLagging
Things tend to travel clockwise through those four over time, though nothing obliges them to, and plenty stall or reverse halfway.
A tail is the path over recent weeks, and the arrowhead is the only thing telling you which end is now. Without it a tail is ambiguous in both directions, and reading one backwards turns a name rotating out of Leading into an apparent leader.
Four principles, each learned by getting it wrong first:
Direction beats position. A name sitting in Leading with its tail turning down is rotating out of it, not leading. The quadrant label alone is not the signal.
Distance from the centre is not quality. For something on the weak side, a bigger gap mostly means it sits further left, which is further from Leading, not better. Sorting by how far a thing has travelled rewards the wrong thing.
A short tail says nothing. Something sitting near the centre is behaving like the index, and length is what separates a real rotation from noise.
The benchmark must be the same everywhere. Measured against a sector fund instead of the index, the same companies on the same day can read Leading or Weakening. Every graph we read is against the S&P 500, without exception.
It is a screener, not a timing tool. It says where attention is going, not when to act, and nothing in an issue is an instruction to do anything.
The rule we read it by
On every ticker we look for a long tail. Trajectory matters more than position, so we favour names moving north east, in the Leading or Improving quadrant.
A name already leading but turning down is not a candidate. That exclusion matters as much as the rule itself.
2. Industry funds
Beneath the sectors sit thirty two industry funds, each holding one line of business: semiconductors, regional banks, biotechnology, airlines, gold miners and so on.
They are read the same way, on the same graph, against the same benchmark. They are judged on their own record, not on whether the sector containing them was chosen, because a line of business can gain ground while the sector around it loses it.
Up to nine appear in an issue. Every one is then put through the same ten tests the companies face.
No company in an issue is taken from these funds. They are a reading in their own right, not a route to a name.
3. Companies
Inside each chosen sector we read the graph again on the companies themselves, and note the ones turning up against the index.
Only companies that already pass the ten tests below are plotted, so the chart carries around twenty readable names rather than eighty that could not be used anyway.
The names that survive are then sorted into the businesses they are actually in, and the strongest of each is kept. Without that step a list can run five regional banks and three laboratory companies, which is one story told eight times rather than eight companies.
The ten tests
Every company and every fund must clear all ten. No score, no weighting, no exceptions.
keep = all([
price > ma_50, # above the 50 day average
ma_50 > ma_150, # the 50 above the 150
ma_150 > ma_200, # the 150 above the 200
rsi_weekly >= 50,
rsi_monthly >= 50,
rsi_daily <= 85, # not already overbought
price >= low_52w * 1.25, # 25% off the one year low
price >= high_52w * 0.85, # within 15% of the high
change_3m - spy_3m > 0, # ahead of the S&P 500
dollar_volume >= 10_000_000, # $10m a day
])
Given the same figures these return the same answer every time. The same threshold applies to a fund as to a company, including the liquidity bar, which has kept funds out of issues.
Each issue also reports the stage of a chart in Stan Weinstein’s sense: Stage 1 a base, Stage 2 an advance, Stage 3 a top, Stage 4 a decline. It is worked out from how far price sits above each moving average. A stage says where a chart is, not what to do about it.
What is judgement, and what is not
Step Reading the graphs, choosing sectors, funds and candidate companiesjudgement, by eyeThe ten testsmechanicalGrouping by business and keeping the strongest of eachmechanicalEvery figure reportedtaken from the data, not chosen
Another reader looking at the same graphs could reasonably reach a different list. That is a property of the method, not a defect in it, and it is why nothing here is presented as a rule that produces a right answer.
What we deliberately do not use
All of these are available in the data and none of them appears in an issue:
Analyst ratings, price targets and target upside
Every forward estimate, including next quarter revenue and earnings growth. A PEG ratio is suppressed for the same reason when a company has no trailing P/E, because the published one is built on forecasts
Any score, whether ours or anyone else’s. No conviction measure, no ranking, and no third-party quality scores such as Piotroski or Altman Z, all of which are in the data. The shortlist is the output of tests, not a rating, and a nine point quality score printed beside that claim invites a reader to treat it as one. The companies are not ordered by preference
The next scheduled results date is printed, because a date is a fact rather than a forecast.
Each company card also shows its P/E against its own three, five and ten year averages. Three averages cannot give a range, because averaging destroys the spread a range would have to be recovered from: an average of twenty could come from a series that ran ten to thirty or one that never left nineteen to twenty one. What they do show is whether a company is dearer or cheaper than it has been, and in which direction that has been moving.
Read a very large average as a warning rather than a valuation. A single period of near zero earnings sends the ratio to hundreds and drags the average with it, so a high average usually marks a bad year rather than an expensive share. Where a company has no P/E today, because it is losing money, the history is left blank rather than shown against nothing.
Sources
StockCharts.com provides the Relative Rotation Graphs. They are read there on a paid subscription, captured by hand, and reproduced under their reprint permission policy.
StockAnalysis.com provides every figure we print. Two exports are downloaded by hand each week, one covering US stocks and one covering funds.
Nothing is read out of a chart into our text, tables or cards. The graphs decide what we look at, not what we print about it.
Known limits
Stated plainly, because a method described without them is a sales pitch.
The thresholds are chosen, not proven. Twenty five per cent above the one year low, fifteen per cent below the high, ten million dollars a day: each is a judgement about what counts as an advance rather than a bounce. None has been validated against a future.
Relative strength describes what has already happened. It is not a forecast and does not predict returns.
Reading a graph by eye is fallible, most of all near the diagonals where a tail’s direction is genuinely hard to call.
One data provider, unverified. Figures are not checked against companies’ own filings, and an error upstream carries straight into an issue.
A chart is a moment. Everything in an issue can be undone by a single earnings report, which is why the next results date is printed beside each company.
Relative Rotation Graph and RRG are registered trademarks of RRG Research, used with permission. The graphs are published by StockCharts.com and are read there on a paid subscription. Neither is affiliated with this publication and neither has reviewed or endorsed anything in it.
The full Disclaimer applies to every issue.