How to combine fundamental and technical filters in one stock screen

A real four-filter question

A stocks-llm visitor asked for companies with a P/E under 15, revenue growth above 10%, a dividend yield above 2% and a 14-day RSI below 40. Three fundamental conditions and one technical one. We ran the same question on readings from the October 2 close.

In that run, 143 companies passed the three fundamental cuts. The RSI condition narrowed them to 41.

The answer listed the first 15 matches, and 10 of them were companies the catalog labels as mortgage REITs: $NLY, $ORC, $ADAM, $AGNC, $IVR, $DX, $EFC, $ARR, $PMT and $FBRT. Their RSI readings ran from 13 to 22. Their yields were high even for an income list, 19.9% for $ORC and 25.5% for $PMT.

The other five rows were $LVS, $STC, $FIS, $SAFE and $FNF. $LVS made it with a P/E of 14.1, just inside the ceiling of 15.

The screen did what it was asked. So why did four filters that sound independent agree so readily on one kind of company?

Why three of the four filters pull the same way

Look at what each number is made of. P/E is the share price divided by earnings per share over the past year. Dividend yield is the annual dividend divided by the share price. RSI is a 0 to 100 gauge of how hard the price has been pushed up or down over the last 14 sessions.

Price sits in all three. When a share price falls and the company has not yet reported new earnings or changed its dividend, the P/E drops, the yield rises and the RSI sinks, all from the same move. A screen that asks for a low P/E, a high yield and a low RSI is asking three times whether the price has fallen.

Revenue growth is the only one of the four that the share price cannot move. The mortgage REITs passed it too: their reported revenue had grown, by 50.9% for $NLY and 127.5% for $ORC.

The yield leg suits mortgage REITs by design. The SEC's investor bulletin on publicly traded REITs notes that REITs have to distribute at least 90 percent of their taxable income each year, so high payouts are built into the structure. The same bulletin says mortgage REITs tend to be more leveraged than REITs that own property. When their prices fall together, as they had before October 2, the group can fill a list like this in one go.

RSI below 40 is a loose condition

J. Welles Wilder, who described RSI in his 1978 book, considered readings below 30 oversold. Forty is a looser line, and the ChartSchool guide to RSI explains why it catches so much. Citing Constance Brown's work on RSI ranges, it says RSI tends to fluctuate between 10 and 60 in a downtrend. A reading under 40 is ordinary for a stock that has been falling for a while.

The product counts show the difference. On the same October 2 readings, 698 covered companies had a P/E under 15. Asking for an RSI below 30 kept 97 of them. Asking for an RSI below 40 kept 306.

So the threshold matters more than it looks. At 30 the technical leg picks out sharp recent declines. At 40 it mostly confirms that a cheap-looking stock has been going down, which the low P/E was already hinting at.

Pair filters that check each other

A technical condition earns its place when it tests something the fundamental filters cannot see. One way to get that is to pair a valuation filter with a trend filter that points the other way.

Take dividend yield above 3% and a price above the 200-day moving average. On October 2, 560 covered companies had a yield above 3%. The 200-day condition removed 368 of them and left 192. Because yield rises as price falls, requiring a price above its long-term average screens out the yields that come mainly from a falling share price.

It does not test whether the dividend will be paid again. Near the top of that list sat the crude oil tanker operators $DHT, yielding 20.97%, and $INSW, yielding 18.17%. Their prices were well above their 200-day averages, and the screen has no view at all on the earnings behind those payouts.

Quality filters pair with trend filters in a similar way. Profitable companies with a return on equity above 15% numbered 709 on the same readings. Asking that they also trade near their 52-week high left 27, among them $JCI, $EXPD, $TSM and $NVDA. Here the technical leg asked something ROE cannot answer: whether the share price sits near the top of its range for the year.

Read the date on each leg

A combined screen mixes numbers that were measured at different times. In the four-filter run, RSI and price came from the October 2 close, the P/E was dated October 2, and the dividend yield was dated September 20.

Return on equity is older still. It comes from annual filings, so in the 52-week-high run most rows used fiscal 2025 figures and $TSM used fiscal 2024. The price condition is refreshed every night. The ROE beside it may describe a year that ended many months ago.

The screen is still accurate about the data it holds. A company can pass today on a fundamental figure that its next report will change.

Every filter is also a data filter

A company with a missing figure cannot pass a filter on that figure. In stocks-llm it is excluded rather than estimated, so each condition you add also removes companies whose data is incomplete.

Technical legs have their own version of this. In the P/E and oversold run, four companies were not evaluated because their price series had stopped updating, and one was skipped because its price history showed a split that had not been adjusted. They were left out, so they never appeared as failures. A short list can come from strict conditions or from thin data, so check how many companies were skipped.

A practical way to build one

Start with the fundamental question, the one you would ask even with no chart in front of you. Keep the thresholds to what you would defend out loud.

Then add one technical condition and ask what it measures that your fundamental filters do not. If it moves with price in the same direction as your valuation filter, it is probably repeating that filter.

Before reading names, read the counts: how many passed the fundamentals, and how many the technical leg removed. Then look at the sector mix of the result. If one industry fills the first page, find out which filter put it there.

Last, open the company pages and check the dates beside each figure that got the company onto the list.

How stocks-llm helps

The chat runs a combined screen as a single question, and its answer can tell you how many companies the technical condition removed. You can try the two pairings above: dividend stocks with a yield above 3% trading above their 200-day moving average and profitable companies with ROE above 15% near their 52-week high. Each row carries the figures it was selected on and their dates.

The indicator guide explains how each technical reading is calculated, and the screens directory lists the ready-made screens.

This article is for information only and is not financial or investment advice. A technical screen describes past price behavior and predicts nothing. Verify material figures against primary sources before making an investment decision.