The daily game
What is the daily trend guessing game?
Trend Guess is a free trend guessing game with a new daily
challenge. Each UTC day it gives you five charts built from
real-world data. The chart shows
the years we know and hides the next stretch. Your job is to draw
the missing line before the real data appears.
A round takes about three minutes. There is no account and no
leaderboard. The point is to read the shape of the data:
direction, pace, turning points, and the scale of the chart. After
you submit, you see the real curve, your score, the year you
missed most, and the source behind the series.
It is for anyone who likes a short daily puzzle and wants to get
better at reading charts. You do not need a statistics background.
If you can compare a line with the axes, you can play. Teachers and
data-curious teams can use a round as a warm-up, but the game is
not a course and does not claim to teach forecasting.
Free Play uses the same question pool but does not change the daily
five. Use it to warm up, try a category you missed, or keep playing
after the daily result is saved.
How scoring works
Each question is scored from 0 to 1,000. The closer your line is
to the hidden values, the higher the score. The formula is
1000 × e−3 × normalized error.
Normalization compares your miss with the range of the visible
data, so a five-point miss on a 0–100 chart is treated
differently from a five-point miss on a 0–1,000 chart.
A perfect line scores 1,000; a line that is far off approaches
0. Five questions make a total out of 5,000. You must set every
hidden year before submitting. The line starts at the last known
value, and you can drag it or use the arrow keys to adjust it.
Where the trend data comes from
Questions are built from public datasets, not generated numbers.
The current pool includes energy and electricity series from
Our World in Data,
U.S. baby-name shares from the
Social Security Administration,
phrase frequencies from
Google Books Ngram Viewer,
and historical NBA team strength from
FiveThirtyEight's Elo dataset.
Every question keeps its unit, year range, and source. After you
submit, the reveal links to the source and explains what the
series measures. Some archives end before today; those questions
use the years that are actually available. This is a game about
reading the past, not a live forecast or a market prediction.
The current pool spans three categories:
- Culture — book-language frequencies and NBA team strength.
- Demographics — U.S. baby-name shares.
- Tech — energy and electricity mixes.
Categories are labels for orientation, not a promise that every
chart fits one theme neatly.
How to read a trend line
Start with the last known point. Ask whether the line is rising,
falling, or flat, and how steep the recent slope is. Then check
the chart's scale: a move that looks large can be small in the
data's units.
- Match the direction first. A line that continues the recent slope is often closer than a dramatic reversal.
- Use the visible range. If the known values stay between 20 and 30, a guess of 200 is almost certainly outside the chart's world.
- Expect noise, not a perfect curve. Real series have one-off years, revisions, and changes in definition.
- Treat a turning point as a claim. Only predict a sharp change when the visible data gives you a reason.
A new challenge every day at 00:00 UTC
The daily set changes at 00:00 UTC. Everyone who opens the same
UTC date gets the same five questions in the same order. Your
progress is stored in this browser, so a refresh returns you to
the same question or result. If you want another round after the
daily five, Free Play draws another question from the pool.
Results can be shared as a link. The link contains the date,
challenge number, and five scores only; it does not contain
answers, sources, or explanations. Whoever opens it sees the
scorecard and a button back to the current daily challenge.
How the daily five are chosen
The daily trend guessing game uses a curated question pool rather
than scraping live. Each day's set aims for a mix of categories
and readable charts. The import validator checks the data shape
and requires enough points for a complete hidden run; it does not
filter out every flat line, so some charts move very little. When
the pool is smaller than the schedule, the game reuses the
questions that have gone longest without appearing.
Each question records its unit, hidden run, and explanation. Some
series are noisy or contain structural breaks; the reveal
explains what the series measures rather than implying a single
cause. New questions can be added in batches, and future dates can
be re-scheduled without changing past results.
What the score does not measure
Trend Guess is a game of estimation, not a test of forecasting
ability. A high score means your line was close to one historical
series on one day; it does not mean you can predict markets,
elections, or the next data release. A low score can come from
one unusual year or a chart with a sharp change.
There is no leaderboard, no account, and no prize. The score is
feedback for the next round, and the reveal is there to show you
how the series actually moved.