Reference
Stock Screens and What They Select
A screen is one indicator with a threshold attached, applied to every instrument at once. What it selects follows from that arithmetic, which is why most-traded lists fill with cheap stocks and oversold lists fill with deteriorating businesses.
Every screen is a filter with a threshold, and the results are entirely determined by what the underlying measure is denominated in. That sounds like a technicality and it is the whole story: a list ranked by share volume is mostly a list of low-priced stocks, a list ranked by an unbounded oscillator is mostly a list of volatile ones, and neither has selected on anything a reader intended.
The pages beneath this one are guides rather than live tables, for a stated reason. A screen needs a market data feed to mean anything, and a list of names dated eighteen months ago reads as current to anyone arriving from a search result. So what is here is what each filter selects and how to run it on data you already have.
What each filter actually selects
Reading a screen means reading its denominator. Four common filters, and what the arithmetic puts at the top of each list:
| Screen | Ranks by | What ends up at the top |
|---|---|---|
| Most traded | Share volume | Low-priced issues, by arithmetic. Rank by dollar volume instead and the list changes almost entirely. |
| Most volatile | Range or return dispersion, as a percentage | Small companies and recent listings, scale-free, so at least the bias is real rather than arithmetic. |
| Overbought | An oscillator above a threshold | Instruments in strong trends. The reading is produced by persistence, which is the opposite of what "overbought" suggests. |
| Oversold | The same oscillator below a threshold | Deteriorating businesses, disproportionately, because a sound company rarely stays oversold for long. |
None of these is an argument against screening. They are an argument for knowing which question a filter answered, because in every case the list is exactly what the arithmetic implies and nothing more.
Building one yourself
Every screen described here can be run end-of-day from free daily price and volume files, and the arithmetic is a few lines over a table of closes. The work that decides whether the output is worth reading is not the calculation.
Use adjusted prices. An unadjusted split or a large dividend leaves a step in the series, and every measure spanning it (a rate of change, an average, an oscillator) is meaningless until the window clears. A screen run over unadjusted data will faithfully return a list of recent corporate actions.
Set a liquidity floor first. Filter out anything below a minimum dollar volume before ranking on anything else. Without it, the top of every percentage-based list is occupied by instruments where a handful of trades set the price, and the reading describes those trades rather than a market.
Exclude what is not an operating company. Exchange lists contain closed-end funds, preferred shares, exchange-traded products, warrants and multiple classes of the same business. They pass price filters perfectly well and they are not what anyone screening for stocks intends to find.
Combine two unrelated conditions rather than tightening one. A stricter momentum threshold returns the same kind of name with a more extreme reading. A momentum condition plus a liquidity floor plus a volume comparison against the instrument’s own history returns a genuinely shorter list, because the conditions are not measuring the same thing.
Volume and activity screens
Lists ranked by how much trading happened. The ranking is dominated by share price unless the screen corrects for it, which most do not.
- Most tradedHighest share volume, and why the same universe ranked by money gives a different list.
- Most traded of the monthA monthly total belongs to its two or three busiest sessions, so the list ranks news rather than liquidity.
- Most volatileLargest range relative to price, and the two definitions that disagree.
- Most volatile todayThe same filter over one session, which is a different question.
Momentum and extreme screens
Lists ranked by an oscillator reading. Each inherits every limitation of the indicator underneath it, including the ones the screen cannot show you.
- OverboughtWhat an RSI above 70 does and does not say about a stock.
- OversoldThe mirror screen, and why its results skew towards falling businesses.
- Biggest losersRanked by percentage change, which is at least scale-free.
Building and reading a screen
The construction around the filters, and the one market-wide measure in this section. Both pages compute their own example rather than describing one.
- Stock selection and screen designThe universe, the gates and the order to apply them in, with the survivor count at each stage, which no published screen states.
- The TICK indexA live census of which way the last trade went across every listed issue, and why its published thresholds belong to a smaller market.
The measures behind the screens
Every filter above is one of these indicators with a threshold attached. Reading the indicator page first is the difference between using a screen and being used by it.
- Relative Strength IndexWhat "overbought" is actually measuring, and why the level is an observation rather than an instruction.
- VWAPVolume weighting done properly, inside one instrument.
- Short volume researchReal data on how much of any volume figure is mechanical rather than directional.
Frequently asked questions
Why are most-traded lists always full of low-priced stocks?
Because share volume counts shares, and a cheap stock produces far more of them for the same money. Ten million dollars buys five million shares of a $2 stock and forty thousand shares of a $250 one, a factor of a hundred and twenty five, from arithmetic alone. Any list ranked by share volume is therefore substantially a list of low-priced issues, and the fix is to rank by dollar volume: share volume multiplied by price.
What does "most volatile" mean on a screen?
It depends on the screen, and the two common definitions disagree. One ranks by the average true range as a percentage of price, which measures how far the instrument travels in a typical session. The other ranks by the standard deviation of daily returns, which measures how much those returns scatter. A stock that gaps overnight and then sits still scores high on the second and low on the first. Neither is wrong; a list that does not say which it used is not usable.
Is an overbought screen a list of stocks to sell?
No, and read that way it is close to an inverted signal. An oscillator reaching an extreme most often means a strong trend is under way — that is what produces the reading — and in a sustained advance RSI can sit above 70 for weeks. A list of overbought names is a list of instruments that have risen persistently, which is a description. Whether any of them is worth selling requires everything the screen did not measure.
Why do oversold screens fill with falling companies?
Because the filter selects on price behaviour and nothing else, and the most reliable way for a stock to be persistently oversold is for the business to be deteriorating. The screen cannot distinguish an ordinary pullback in a sound company from the early part of a permanent decline, since both look identical in the closing prices. This is the single most expensive misuse of a mechanical screen, and the correction is not a better threshold but information from outside the price series.
Should a screen use dollar volume or share volume?
Dollar volume, for anything comparing instruments. It is share volume multiplied by price, so it measures the money committed rather than the number of certificates that moved. Share volume remains the right input when comparing an instrument against its own history, which is the only comparison this site treats as safe. The general rule: shares for the same instrument over time, dollars across instruments.
How many results should a screen return?
Few enough to look at each one. A filter returning two hundred names has not narrowed anything; it has moved the problem. In practice the useful settings are the ones that produce a list short enough to read in an afternoon, which usually means combining two unrelated conditions rather than tightening one, a liquidity floor plus a momentum condition, say, rather than a stricter momentum threshold.
What does a screen never tell you?
Why the condition exists. A stock appears on an oversold list whether it fell because of a disappointing quarter, a sector rotation, an index deletion, or nothing identifiable at all, and those four cases have nothing in common. Screens are a way of shortening a list of candidates for actual work, and treating the output as a conclusion is what gives mechanical filters their poor reputation.
Do these pages run live screens?
Not yet. Running a genuine screen requires a market data feed, and publishing a stale list of names would be worse than publishing none, a table dated eighteen months ago reads as current to anyone arriving from a search result. What is at these paths for now is what each filter selects, how it is built, and how to run it yourself with data you already have. The intention is a live version later, at the same paths.
Can I build one of these without a data subscription?
Yes, for anything end-of-day, which is what all the screens described here are. Free daily price and volume files cover the American market adequately, and every calculation on this site (RSI, average true range, dollar volume, a rate of change) is a few lines of code over a table of closes. The work that actually matters is not the arithmetic but the housekeeping: adjusted prices, a liquidity floor, and excluding the issues that are not operating companies.