Free tool · no signup
1% better calculator (the honest version)
Improving 1% a day for 365 days gives 1.01^365 = 37.78, the number Atomic Habits made famous. It assumes you never miss a day, nothing plateaus, and nothing you gain ever fades. Set a realistic consistency rate below and this calculator plots that ideal curve against one that accounts for missed days, a ceiling on how far a single habit can take you, and the decay that happens when you stop. The realistic result is smaller, and it is still worth doing.
Runs in your browser · Nothing uploaded · No account, no email
1% better · naive vs realistic
in your browser
1% is the famous figure.
292 of 365 days actually done.
You give back half of a day’s gain. Skills and fitness fade slowly, so this is closer to true than zero.
Gains flatten as you approach 5× your starting point.
- Naive
- Realistic
Log scale: equal height means equal ratio, so both curves show their real growth rate. Switch to linear to see the famous hockey stick.
Smaller than 37.78×, still a lot
Doing it 80% of days for 365 days puts you at 3.98×, which is 11% of the 37.78× the naive version promises. You passed 2× on day 115. Keep going past 365 days and it settles near 4.50×.
Naive, 1% better
37.78×
1.01^365, the number with no missed days and no ceiling.
Naive, 1% worse
0.0255×
0.99^365, the mirror image everyone quotes next to it.
Days actually done
292
73 missed out of 365.
Day you hit 2×
115
On the realistic curve.
Both curves are computed in this browser from the settings above. Nothing is uploaded, and the realistic model is written out in full further down the page so you can reproduce it.
Where 37.78 comes from
The arithmetic is exact. The claim around it is not.
The calculation became famous through James Clear’s Atomic Habits, which puts it plainly: “if you can get 1 percent better each day for one year, you’ll end up thirty-seven times better by the time you’re done.” Clear is illustrating a principle, and the principle is sound. Small repeated inputs beat sporadic large ones, because each one lands on a base that the previous ones already raised. That is a real property of compounding and it is worth internalising.
The trouble starts when the illustration gets treated as a measurement. Ask what 37.78× is 37.78 times of and the sentence falls apart. There is no unit. You cannot be 37.78 times better at sleeping, or at reading, or at being a decent colleague. Percentages compound only when they apply to a quantity, and the quantities you can actually measure in a person behave nothing like a savings account: they have floors, ceilings, and a strong tendency to regress when you stop.
Three assumptions are doing all the work in 1.01365, and all three are hidden. You never miss a day. Nothing you gain ever fades. And the 1% keeps being 1% forever, so your three-hundredth day of practice moves you as much as your third. Change only the first of those, from 100% to a realistic 80%, and add a small cost for the days you skip and a ceiling on one habit, and 37.78× becomes roughly 4×. That is the calculator above.
How to read your result
Four knobs the other calculators hardcode.
- 01
Set a daily gain
The famous version uses 1%. It is a placeholder, not a measurement, because nobody can tell you what 1% of "better at running" is.
- 02
Set a consistency rate
The share of days it actually happens. The naive curve assumes 100%, which is the single biggest reason it lands at 37.78×.
- 03
Say what a missed day costs
Nothing, or a slow decay where you give back half a day’s gain. Fitness and skill fade when you stop, so zero is generous.
- 04
Put a ceiling on one habit
Gains from a single habit flatten as you get good at it. The ceiling is how far you think this one habit can carry you.
The method
Here is the formula, so you can check it.
The realistic model
Let g be the daily gain, c the share of days you do it, and d the fraction of a day’s gain you give back on a day you don’t. Starting from V₀ = 1:
headroom(V) = (ceiling − V) / (ceiling − 1), clamped to 0…1
V(t+1) = V(t) × (1 + c·g·headroom(V(t)) − (1 − c)·d·g)
The naive curve is the same thing with c = 1, d = 0 and no ceiling, which collapses to (1 + g)t. Set the controls that way and the two lines sit exactly on top of each other, at 37.78× on day 365.
Each day is a coin flip you win with probability c. Because the daily factors are independent and multiply, the expected multiplier is the product of the expected daily factors, which is exactly the line above. So with the ceiling switched off, this is not an approximation of “what usually happens”: it is the average outcome, computed directly. With the ceiling on, the daily factor depends on where you already are, so the curve is the mean path rather than the exact expectation. It is also deterministic on purpose, because a simulation would draw you a different chart every time you loaded the page.
The ceiling is the honest term, and it is the one the naive model omits entirely. Habit formation research finds an asymptote rather than a straight line: Lally and colleagues fitted an asymptotic curve to each participant’s automaticity over 84 days, with a median of 66 days to reach 95% of that personal plateau and a range from 18 to 254 days. Their asymptote is about how automatic a behaviour feels, not how much you improve at it, so it is not a direct measurement of this ceiling. The performance side points the same way: fitting individual learning data, improvement approaches a limit rather than continuing at a fixed rate.
What this model still cannot see: it has no notion of which habit you picked, and a 1% daily gain in sleep duration and a 1% daily gain in deadlift are not comparable quantities. It treats missed days as evenly spread, when in real life they clump into bad weeks. It gives you one habit, when most people’s year contains several that interact. And the ceiling is a number you chose, not one measured from your data. If you want a curve with none of those assumptions in it, you have to log the real thing and plot that instead.
Sources
- [1]How are habits formed: Modelling habit formation in the real world
Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010) · European Journal of Social Psychology, 40(6), 998–1009
The reason this calculator has a ceiling at all. The authors fitted an asymptotic curve to each participant’s automaticity scores over 84 days: habits strengthen toward a personal plateau rather than without limit. Median time to reach 95% of that plateau was 66 days, range 18 to 254 days. A single missed opportunity lowered automaticity by an average of 0.29 points on the 0–42 subscale, which they call "a very small decrease". Their asymptote is about how automatic a habit feels, not about how much you improve at it.
- [2]The power law repealed: The case for an exponential law of practice
Heathcote, A., Brown, S., & Mewhort, D. J. K. (2000) · Psychonomic Bulletin & Review, 7(2), 185–207
Why the daily gain shrinks as you improve. Fitting individual, unaveraged learning data, performance approaches a limit exponentially rather than following a power law. Either way the curve flattens, which is the opposite of a fixed 1% a day forever.
- [3]Atomic Habits
Clear, J. (2018) · Avery / Penguin Random House
Where the number got famous: "if you can get 1 percent better each day for one year, you’ll end up thirty-seven times better by the time you’re done." Clear presents it as an illustration of compounding, not as a measurement.
This models a hypothetical. Your own numbers are not hypothetical.
Every curve on this page starts from an assumption you typed in. Habit Pocket records the real figure each day, on iPhone and the web, and charts what actually happened instead of what a model says should have.
5 habits free, forever. No credit card. Your data stays private.
FAQ
Questions about the 1% rule
1.01^365 = 37.7834, usually rounded to 37.78. That is where "get 1% better every day and you will be 37 times better in a year" comes from. The mirror figure is 0.99^365 = 0.0255, so the same arithmetic run downward leaves you at about 2.6% of where you started.
The arithmetic is exact but the assumptions behind it are not realistic for one habit over one year. It requires you to improve every single day without missing one, for gains never to fade, and for nothing to plateau. Change only the consistency assumption to 80%, add a small cost for missed days and a ceiling at 5×, and the same 1% daily gain lands near 4× instead of 37.78×.
Missing one day costs you very little, and the research on habit formation says the same thing. In Lally et al. (2010), automaticity fell by an average of 0.29 points on a 0–42 subscale after a single missed opportunity, which the authors call a very small decrease. The damage in this calculator comes from a sustained miss rate, not from any individual day: at 80% consistency you are doing it 292 days out of 365, and that is what pulls the curve down to about 4×.
The principle holds even though the number does not. Small repeated inputs beat sporadic large ones because they accumulate on a base that keeps growing, and a realistic model of that still returns roughly a 4× improvement in a year, which is a large change in anything you can actually measure. What does not hold is treating 37.78× as a forecast, because "1% better" has no unit and nothing about a person compounds without limit.
1.01^30 = 1.3478, so a perfect month of 1% daily gains is about a 35% improvement. At 80% consistency, with a small decay on missed days and a ceiling at 5×, the same month gives 1.22×, a 22% improvement. Thirty days is short enough that the two curves have barely separated, which is part of why the idea feels so convincing early on.
Last updated August 2026 · I wrote the model on this page myself and checked 1.01365 and 0.99365 against the published figures · By Bohdan Stefaniuk