How pace24 Works
A Streak for a Sport Where Rest Is Part of the Plan
Duolingo says streaks bring people back. Training has rest days, illness and plans that change. Here are the rules we gave our streak and the score beside it, and what we have not tested.
The pace24 team5 min read
Key takeaways
- Duolingo reports that learners at a 7-day streak were 2.4 times as likely to come back the next day. That is the company's own figure, and it shows association, not cause.
- A streak that counts consecutive days of exercise punishes rest, the thing a plan most needs you to do.
- pace24 counts completed planned endurance days. Planned rest and changes you approved leave the streak alone.
- Extra training earns nothing, and the same ride imported twice cannot cover two sessions.
- The 0 to 100 momentum score is sessions done over sessions planned in the last 28 days. The screen shows the counts so you can check it.
- We have not yet tested whether any of this helps real athletes.
About this post: the Duolingo figures are results the company published on its own blog and have not been audited. Everything about pace24 describes how our app counts, from its code and plan record.
Streaks work, and we are not going to pretend otherwise. Duolingo says learners who reach a 7-day streak are 2.4 times as likely to come back the next day as learners without one. It also reports that when it separated the streak from the daily goal, so that one lesson was enough to keep it alive, day-14 retention rose 3.3 percent and the share of daily learners on a streak rose 10.5 percent in 20 days. All relative changes, all the company's own.
Read the first number carefully. People who keep a streak are the people who were going to come back anyway. It tells you the streak is a good signal. It does not tell you it is a good cause.
Then try it on a training week
Picture a normal week. Monday rest. Tuesday intervals. Wednesday easy. Thursday rest. Friday threshold. Saturday long ride. Sunday easy.
A streak that counts consecutive days of exercise has two options with that week. It breaks on Monday and Thursday, or it quietly rewards you for riding on days the plan told you to rest. Both are wrong. A rest day is a planned session, and doing it properly is the work.
Illness, a work trip, a session your coach moved: the same problem. A chain that snaps on the days an athlete is right to stay off the bike teaches the wrong lesson.
What our rules do
pace24's streak counts completed days of your planned endurance sessions, not app opens and not exercise volume.
| What happens | Effect on the streak |
|---|---|
| A planned rest day | Holds, adds nothing |
| You do the planned ride, run or swim | Adds one day |
| You approve a skip, a move or a drop through the coach | Holds |
| A strength or core session | Neutral either way |
| Yesterday's ride has not synced yet | Stays pending for a day |
| A planned endurance day with nothing recorded by the end of the next day | Resets |
| An extra ride on a day with none planned | Earns nothing |
| The same ride imported twice | Counts once |
A new training plan starts at zero, and the screen shows your best run in the current plan. On a day with two planned sessions, both need a recording before the day counts.
Two of those rows are choices. A missed planned day does reset the streak, because a streak that never resets is not saying anything. And extra training earns nothing on purpose: the number is meant to say "you did what the plan asked", and rewarding more volume would make it say "do more".
The score beside it
Next to the streak sits a number from 0 to 100. It is the sessions you completed out of the sessions the plan asked for in the last 28 days, set against the 28 days before. Eleven of twelve shows as 92. The window stops two days ago, so late syncs can land.
The screen prints the counts and the formula. Under four planned sessions it says plainly that this is a developing baseline. Rest days, dropped sessions and sessions you moved away add to neither side.
It uses no fitness score, no sleep, no heart rate variability and no sense of how hard a session was. No AI is involved: the same plan and the same recordings give the same number every time. A score of 100 means you did what was planned. It does not say the plan was good, or that you are fit, rested or healthy, and the screen does not say so either.
What we do not know
We do not know that this keeps anyone training. We have not run a pilot, and we claim no effect on consistency. Duolingo's results come from a very different activity, with a very different cost of showing up. The case for the design is narrower: a streak that treats rest as part of the plan is at least not working against the plan.
Want a streak that agrees with your rest days? Get early access.
Sources
- Duolingo (company blog), "Improving the streak": blog.duolingo.com/improving-the-streak. The 3.3 and 10.5 percent figures are relative changes from the company's own experiment; the 2.4 times figure is its own observational data.
- Our own plan records for the streak and momentum features and the screen's copy, 2026-11.
- No real athlete data, and no data of Strava origin, was used for this post.