Ask the old Trade Timing page whether a stock was trending and it would hold a small election among its moving averages, count the ballots, and very often announce "Neutral, 50 % confidence" — the statistical equivalent of a shrug. The Current Trend card now gives you one number instead. Same scale for every stock, one honest caveat attached.

The short version

• One number per stock: how far the recent trend sits from flat, measured in standard deviations.
• The same scale everywhere: a +1.0 on a sleepy utility means the same as a +1.0 on a jumpy chip stock, so you can compare them side by side.
• It describes, it doesn't predict: it tells you what a stock has been doing. We tested whether it also forecasts what comes next. On single stocks it doesn't, and the page says so.

How to read it

Open Trade Timing, pick a ticker, run the analysis and scroll down to Advanced Pattern Analysis. The first card under Trend Detection is the new one:

Under the number sits a line of context — "higher than 30 % of its readings over the last 3 years" — because a +0.5 means one thing for a stock that spent three years at +2 and another for one that usually naps near zero. The old moving-average vote is still there, one line further down, for anyone attached to it. It is on its way out.

On the day of writing, Alphabet reads +0.55 (a modest uptrend), Nvidia +0.89 (uptrend) and Netflix −1.40 (a downtrend, lower than all but 4 % of its readings over the past three years). These numbers change every day. That is rather the point.

What's under the bonnet

Each day's return is first divided by how volatile the stock has recently been, so a 2 % day on a calm stock counts for more than a 2 % day on a wild one. Those adjusted returns are then blended into a running average in which older days slowly fade. How slowly: a day's influence halves after about 78 trading sessions, roughly three and a half months.

That speed isn't our invention. It comes from research on 70 futures markets over 33 years (Grebenkov & Serror, 2014; Valeyre, 2026), which found that one well-tuned average of this kind does about as well as the elaborate indicator baskets people like to stack on top of each other — and that the result barely changes anywhere between about 50 and 150 days. Nobody had to curve-fit the dial.

Geek aside: the formula we started from promises a reading in standard deviations, but with its scaling factor the numbers come out about 30 % too small — a standard deviation of 0.71, not 1. We fixed the factor, so here "+1.5" really does mean 1.5 standard deviations. It is the kind of bug that survives for years because nobody plots the histogram.

The gauge needs about 1.6 years of prices to warm up, so keep the date range on the default five years. Shorter ranges get a polite "needs more history" instead of a made-up number.

A thermometer, not a crystal ball

Here is the part most product announcements leave out. Before shipping, we ran the trend number through our own Signal Quality page, which scores a signal against what actually happened next — and against a baseline of pure random noise. On the stocks this platform covers, the trend number did not beat the noise. Stocks that had been trending kept trending about as often as a coin would suggest. (Trend-following works on the futures in the research, going long and short across dozens of markets. A long-only stock portfolio is a different animal.)

So the card does the honest thing: it describes the trend and says plainly that it does not forecast the next move. A thermometer is still useful. It just shouldn't be asked about next week's weather.

You can check our homework: on Signal Quality, pick Trend EMA (112-day) as the source and any stock, and read the verdict next to the noise baseline.

The same research also describes a portfolio-sizing rule built on this signal. We parked it: a sizing rule that leans on a signal that doesn't predict would mostly generate trades, and trades are not a strategy. If the evidence changes, so will we.

Why we bothered

Because "Neutral, 50 % confidence" was technically true and practically useless. Because one number with one setting is much harder to fool yourself with than a committee of indicators that each get a vote — fewer knobs, fewer ways to tune a backtest into a fairy tale. And because a number on a page should mean the same thing on every stock. Standard deviations do. "Two out of three moving averages agree" doesn't.

Where to find it

• Trade Timing → run any ticker → Advanced Pattern Analysis → Current Trend.
• Signal Quality → source Trend EMA (112-day), to see how it scores against noise.
• Or press Ctrl+K and type "trade timing".

Informational only, not investment advice. A trend reading describes the past; the market reserves the right to disagree.