Why a timer app has never changed anyone's semester, what code can do that paper cannot, and which tempting features you have to leave out to keep a ladder worth living with
Because the entire category measures input. A pomodoro app records that you sat
there for twenty-five minutes and has no opinion about whether anything came out
of it, and after nine months of honest logging the only thing I could tell you
about myself was a duration.
The verdict — did this sitting achieve the thing I said it would — is the only
field in the whole system that touches output, and no timer asks for it.
Paper is better than code for the sitting itself. It is faster, it never
crashes, and nothing about a five-minute tick needs a database.
Code buys exactly one thing, and it is worth the trouble: a curve. Two
hundred verdicts is a shape. It tells you that your Tuesdays are real and your
Thursdays are theatre; that the run you remember as a disaster was four losses
inside a month of gains. You cannot hold that in your head and you cannot see it
in a shoebox of paper.
So the rule for what to build: store the verdicts, derive everything else. If
the rating is a stored number you will eventually have a stale one; if it is
recomputed from an ordered list of wins and losses, then re-judging a sitting
from six weeks ago correctly redraws the whole history.
We will read one implementation together. It is faithful to a real ranked system
in the parts that matter — ten tiers, a hundred points per division, placements
that can only gain, a hidden rating whose lead or lag drives how much a single
result swings, drops gated behind a game taken at zero, and protection for a few
games after a promotion.
It has one house rule, which exists to answer the objection twenty minutes
isn’t enough to do anything. Length picks the mode. Under ten minutes is a
remake: somebody went offline, the game never counted, it touches nothing.
Ten to twenty-five is the casual mode, at half the swing. Twenty-five and up is
ranked, full swing. Suddenly a twenty-minute gap in your day is a thing you can
do something with.
Three things were deliberately left out, and this is the part to copy.
No decay. Real ladders take points off you for not playing. This one never
does. A mechanism whose job is to survive your bad weeks cannot have a rule that
punishes you for having one, on its own initiative, while you are not even
looking at it.
No win-streak bonus. Undocumented in the original and tier-dependent. It
would add noise to a curve whose whole job is to be legible.
Nothing pops up when you stop. No summary card, no score screen, no
in-session checklist panel. Those were built and then cut, because they made
studying feel like a performance of studying.
Write your refusals down. In the demo I will ask you for one, and “I didn’t have
time” is a different answer from “I decided against it”.
It is faster, it never crashes, and nothing about a five-minute tick needs a database.
So this is not a lecture about upgrading. It is about the one thing paper cannot do.
3
Why not just use a timer app
Because the entire category measures input.
A pomodoro app records that you sat there for twenty-five minutes and has no opinion about whether anything came out of it. Nine months of honest logging left me able to tell you one thing about myself: a duration.
The verdict — did this sitting achieve the thing I said it would — is the only field in the whole system that touches output, and no timer asks for it.
4
Code buys you exactly one thing: a curve
5
Two hundred verdicts is a shape
FigureAn example rating curve recomputed from two hundred logged sittings. The line climbs unevenly across the whole run. One stretch of thirty sittings is shaded and labelled the month you remember as a disaster; inside it exactly 4 losses are marked with dots, and the line rises through all of them. To the right, an inset of win rate by weekday shows Tuesday and Wednesday well above the halfway line and Thursday well below it. The caption reads: the rating is recomputed from the verdicts every time, so re-judging a sitting from six weeks ago redraws all of this.
6
So: store the verdicts, derive everything else
If the rating is a stored number, you will eventually have a stale one.
If it is recomputed from an ordered list of wins and losses, then re-judging a sitting from six weeks ago correctly redraws the whole history.
7
The worked example, and what it is faithful to
Ten tiers. A hundred points per division. Placements that can only gain.
A hidden rating whose lead or lag drives how much one result swings you.
Drops gated behind a game taken at zero, and protection for a few games after a promotion.
8
One house rule: length picks the mode
It exists to answer the objection twenty minutes isn't enough to do anything.
Under ten minutes — a remake. Somebody went offline, the game never
counted, it touches nothing.
Ten to twenty-five — casual, at half the swing.
Twenty-five and up — ranked, full swing.
Suddenly a twenty-minute gap in your day is a thing you can do something with.
9
The refusals matter more than the features
FigureThree feature panels drawn as faint outlines and struck through: a decay meter, a win-streak multiplier, and an end-of-session summary card. Each was built and then taken out again.
10
No decay
Real ladders take points off you for not playing.
This one never does.
A mechanism whose job is to survive your bad weeks cannot have a rule that punishes you for having one — on its own initiative, while you are not even looking at it.
11
No win-streak bonus
Undocumented in the original, and tier-dependent.
It would add noise to a curve whose whole job is to be legible.
12
Nothing pops up when you stop
No summary card. No score screen. No in-session checklist panel.
Those were built and then cut, because they made studying feel like a performance of studying.
13
"I didn't have time" is a different answer from "I decided against it".
Write your refusals down. In the demonstration I will ask you for one.
14
Assessment 1 demonstrations
Second half of the session. About six minutes each.
Paper or coded — both are fully marked, and a paper system in daily use marks better than a half-finished program.