No. 18 soapbox
Same Day, Opposite Year
One percent a day is ×38 or ÷39 by New Year's Eve, if the year goes exactly the way the arithmetic assumes.
There is a famous thought experiment that gets sold as a life plan. Take one ordinary day and live it 365 times. In one version you end each day one percent better than you began it. In the other, one percent worse. Nothing else about the two years differs, and by December they are unrecognizable to each other.
Compounding does all the work. Multiply 1.01 by itself 365 times and you get 37.8, so the improving year ends nearly thirty-eight times ahead. Multiply 0.99 instead and you get 0.026, a fortieth of where you started. Those two numbers are pure arithmetic. Nothing was measured.
| Day | 1% better each day | 1% worse each day |
|---|---|---|
| 0 | ×1.00 | 1.000 |
| 91 | ×2.47 | 0.401 |
| 182 | ×6.12 | 0.161 |
| 273 | ×15.13 | 0.064 |
| 365 | ×37.78 | 0.026 |
I run a site whose whole pitch is one small thing done every day, so I am exactly who this graph was drawn for. It flatters my product, which is why I wanted to poke at it, the way I poked at my token bill.
What the arithmetic assumes
- That you are a number. Money compounds because a dollar earned is itself a dollar that earns. A skill has no closing price. The graph needs you to be a number before it can multiply you, and you are not one.
- That every day counts. The model has you improving on the day of the flu, the funeral, and the month the day job eats every evening. It also scores rest as decay: a day that merely holds steady counts as a day you got worse.
- That setbacks are polite. Real ones are rarely a tidy one percent, and they do not check the streak before arriving. The rough quarter lands whether or not you had momentum.
A year that could actually happen
So here is a model I believe more, built from dice. Most days are flat: you showed up, did the thing, and nothing measurably moved. About one day in four earns a real one percent. Knocks arrive on their own schedule, three or four percent backwards. And a handful of days a year, something clicks and pays five at once.
Every number in this section is modeled, not measured. The dice are mine and the sizes are guesses, so distrust them: the buttons under the chart change the luck, and the chart redraws the year that follows.
For the builders: the dice this chart rolls
Each day rolls once against fixed odds. A decent year: a 25% chance of a 1% gain, 5% of a 3% knock, 2% of a 5% big day, flat otherwise. The rolls come from a small seeded generator (mulberry32) with a fixed seed per scenario, so the prerendered page and the hydrated one draw the identical year. The spread quoted under the chart was measured over 2,000 rerolls of the same odds.
This drawn year improves on 103 days, takes 20 knocks, catches 6 big days, and lands at ×2.0. Rerolling the same dice lands between ×1.6 and ×2.7 four years in five.
| Day | A decent year |
|---|---|
| 0 | ×1.00 |
| 91 | ×1.11 |
| 182 | ×1.63 |
| 273 | ×1.76 |
| 365 | ×2.03 |
The decent year lands at ×2.0, improving on 103 days, taking twenty knocks, and catching six big ones. The rough year is not the same dice rolled worse. I loaded them: fewer good days, more knocks, and each knock a little heavier. It improves on 72 days, takes 29 knocks, catches four big ones, and still ends a quarter behind its start.
The charmed year, luckier than anyone should plan on, improves on 112 days, takes fourteen knocks anyway, catches ten big ones, and lands at ×3.2. That is a typical charmed year, not the best one. Reroll the same generous odds two thousand times and half finish above ×3.2, with the luckiest at ×7.0. Even that best year is under a fifth of what the arithmetic promised.
What the smaller number buys
Twice as good in a year is still an outrageous return: twice the words found before the buzzer, or twice as fast at the thing your job actually is. Nobody who has measured themselves at anything would sneeze at double. The poster was right about the direction and wrong about the slope.
My dice cannot explain why the flat days matter, because they roll all 365 days whether or not anyone shows up. Quitting is the risk the model leaves out, and I suspect it is the largest one. A big day can only land on someone still rolling. What the dice do show is that a knock is the same three percent whether it falls in January or November, so nothing already banked is at extra risk.
This is the quiet argument of a daily-games site, so discount my enthusiasm as marketing if you like. But I already run the roster on it. A new game stays only if somebody I know is still playing it five days a week, four weeks after launch, and that verdict counts the showing up, never the size of any one day's score.