A sikku kolam can be drawn without lifting the hand because of a theorem Leonhard Euler proved in 1736: a connected drawing can be traced in one stroke exactly when every junction has an even number of line-ends, or at most two do not. Every crossing in a kolam has four ends, so the rule is met automatically, and the only question left is whether the line closes into one loop or several. Logos can be checked the same way, but the single-stroke rule is a useful design constraint, not a proven secret of memorable logos.
- Euler, 1736: a figure is one stroke if it is connected and has 0 or 2 points with an odd number of line-ends.
- 2k odd points = k strokes: the letter H has 6 odd points, so it needs 3 strokes; O and 8 need 1.
- 4 ends per crossing make every kolam crossing "even", so a connected kolam is always one closed line.
- 2^40, about 1.1 trillion: crossing patterns on a 5 × 5 grid of square kolam tiles, before symmetry.
- 1 in 85: people who drew the simple Apple logo perfectly from memory in a 2015 study. Simplicity alone does not make a logo memorable.
This is a tier-4 "collision": two things that should have nothing in common, tested for a real link. The claim is in the title. A sikku kolam, the knot pattern drawn in rice flour at Tamil doorsteps, is one line you never lift your hand from. Is a good logo the same? The maths says something precise about the first half and something more modest about the second.
What makes a kolam one continuous line?
Three rules, and one consequence. In a sikku (knot) kolam, a grid of dots called pulli is laid down first, then a line is drawn that loops around every dot, crosses itself only at points, and leaves no loose ends.
Venkatraman Gopalan of Penn State states the rules plainly in his 2023 paper:
- every dot is individually encircled;
- line crossings are discrete points;
- there are no loose ends.
A large kolam drawn in one sweep was traditionally a point of pride. ThePrint's 2026 report on a kolam-drawing algorithm noted that people usually manage patterns up to 19 dots, and 31 in exceptional cases.

Königsberg, in what was then Prussia, had seven bridges linking two river banks and two islands in the Pregel. Its people wondered whether a walk could cross every bridge exactly once. In 1735 Leonhard Euler, then in St Petersburg, turned the question into points and lines: each land mass a point, each bridge a line. A walk entering and leaving a point uses two bridges each time, so every point except the start and the end needs an even number of bridges. In Königsberg one land mass had five bridges and the other three had three each. Four odd points, so no such walk existed. The paper, Solutio problematis ad geometriam situs pertinentis, is usually counted as the first theorem of graph theory.
Wikipedia: Seven Bridges of Königsberg
When can a drawing be made in one stroke?
When it is connected and has zero or two points with an odd number of line-ends. With zero, the stroke closes on itself; with two, it starts at one odd point and ends at the other. Here is the proof, in four steps.
The worked derivation
Turn any line drawing into a graph. A vertex is any end, corner-junction or crossing; an edge is a piece of line between two vertices; the degree of a vertex is the number of line-ends that meet there.
- Handshake lemma. Every edge has two ends, so summing degrees counts each edge twice:
\[ \sum_{v} \deg(v) = 2E \]The right side is even, so the number of odd-degree vertices is always even: 0, 2, 4 and so on.
- Passing through costs two. A pen that enters a vertex and leaves it uses two line-ends there. Only the first and last vertex of the stroke can use an odd number. So one stroke needs at most two odd vertices.
- Sufficiency (Hierholzer, 1873). If the drawing is connected and all degrees are even, you can always trace a closed stroke that uses every edge once. Carl Hierholzer's posthumous 1873 paper gave the first complete proof and a step-by-step construction.
- The general count. With \( 2k \) odd vertices, add \( k \) imaginary edges pairing them up. Now every degree is even, so one closed stroke exists. Delete the \( k \) imaginary edges and the stroke falls into \( k \) pieces:
\[ \text{minimum strokes} = \max\!\left(1,\ \frac{\#\text{odd vertices}}{2}\right) \]for one connected drawing.
Whether a drawing can be done in one stroke is not a question of talent. Count the odd points; the answer is already decided.
Use this to check what you just read. Count the odd points in any mark, a letter, or your own signature.
How many strokes does a drawing need?
Why is every kolam crossing "even"?
Because a crossing is where two passes of the line meet, and two passes bring four line-ends. Degree four is even, so a kolam drawn only of loops and crossings has no odd vertices at all.
That means Euler's condition is satisfied by construction. The only way a kolam can fail to be one line is by splitting into several separate closed loops. This is exactly what kolam artists manage by eye, and what mathematicians have now written down.
The authors build the heart-shaped Hridaya kolam from \( m \) dots per arm and \( n \) arms using the sequence \( a_k = (k\,n) \bmod m \). When \( \gcd(m, n) = 1 \), the arm count generates every residue, the dot graph has equal in- and out-degree at every dot, and the pattern closes as one Eulerian circuit, so it can be drawn in one stroke. They extend earlier odd-only constructions to even \( m \) (examples with \( m = 4 \) and \( 8 \) and \( n \) from 3 to 13). Their October 2025 follow-up shows that when \( \gcd(m, n) \ne 1 \) the pattern breaks into several disjoint loops. The limit: the single-stroke claim is shown with examples and figures, not a formal proof, and connectivity is assumed rather than checked.
Read the paper (free)How many kolams are there?
Far more than anyone could draw. Gopalan's square-tile model treats a kolam on an \( n \times n \) grid of dots as a set of tiles whose shared edges either carry a crossing or not, which gives a clean count.
- Shared edges. An \( n \times n \) arrangement of square tiles has \( 2n(n-1) \) internal shared edges.
- Two states each. Each shared edge is either a crossing or not, so without symmetry
\[ N(n) = 2^{\,2n(n-1)} \]
- Plug in. \( n = 3 \): \( 2^{12} = 4{,}096 \). \( n = 5 \): \( 2^{40} \approx 1.1\times10^{12} \). \( n = 6 \): \( 2^{60} \approx 1.15\times10^{18} \).
- The Euler echo. Gopalan also proves that the number of tiles with an odd number of loose ends (his "Drop" and "Fan" tiles) must be even for a kolam to close. That is the handshake lemma again, in rice flour.
Note that this count includes kolams that close into several loops. The single-line ones are a subset, and the bigger the grid, the more skill it takes to find one.
See the data as a table
| log10 of 2^(2n(n-1)) | |
|---|---|
| 2 x 2 | 1.2 |
| 3 x 3 | 3.6 |
| 4 x 4 | 7.2 |
| 5 x 5 | 12.0 |
| 6 x 6 | 18.1 |
Source: Formula from Gopalan, Symmetry Classification and Enumeration of Square-Tile Sikku Kolams (arXiv 2023) · A value of 12 means about a trillion. Symmetric kolams are a small fraction of these.
Gopalan, a materials scientist at Penn State, builds sikku kolams from six square tiles with 0 to 4 loose ends each, so that two of the three kolam rules hold automatically and only "no loose ends" remains. He classifies kolams by point-group symmetry, derives counting formulas for each symmetry type, and proves a necessary condition: the number of odd-ended tiles must be even. He notes pulli kolam is practised across five southern states with a combined population above 260 million. The limit: only six tile types are used, so many traditional kolams fall outside the model.
Read the paper (free)Does the single-line rule make a better logo?
It makes a more constructible logo, not automatically a more memorable one. The memory research is humbling: people who see a simple logo daily still cannot draw it.
Across four experiments, participants briefly saw either the real Apple logo or one with the bite flipped, then had to pick the correct logo from a set of look-alikes. Seeing the altered version made people less accurate at picking the real one, though the effect faded after a short delay. The authors recap the 2015 study by Blake and colleagues in which only 1 of 85 people drew the logo perfectly from memory, average accuracy was about 40%, and fewer than half picked the real logo from eight options. The limit: one very familiar logo, tested on students, so it says little about new marks.
Read the paper (free)What the single-stroke constraint does give a logo designer is concrete:
- Coherence. One path means no stray parts; the mark reads as one object, which is the Gestalt idea of good continuation.
- Reproducibility. A one-line mark can be embroidered, laser-cut, drawn in neon or traced in a stroke animation without breaking.
- A drawing procedure. Like a kolam, it can be taught as a sequence of moves. Whether that improves recall is my hypothesis, not a tested result.
- A cost. It rules out letterforms like H and T, crossbars and separate dots, which is why few wordmarks are single strokes.
| Shape | Odd points | Minimum strokes | One-line logo friendly? |
|---|---|---|---|
| Circle, O, 8, infinity sign | 0 | 1 (closed) | Yes, closes on itself like a kolam |
| L, S, Z, a spiral | 2 | 1 (open) | Yes, starts and ends at the two odd points |
| T, E, A, X | 4 | 2 | Only by redrawing the joins |
| H, the Königsberg bridges | 6 (H), 4 (bridges) | 3 (H), 2 (bridges) | No |
| Any sikku kolam | 0 | 1 per separate loop | Yes, if the loops join into one |
How sensitive is the conclusion?
The maths is exact; the design claim is not. Here is what changes it.
- Counting retraces. If retracing a line is allowed (pen down, going back over ink), anything connected is one stroke. Kolam practice and Euler both forbid it.
- Lifting for dots. Separate dots, like the one over an i, each add a piece and so at least one stroke. A kolam's dots are not part of the line, so they do not count.
- Loops vs one line. A kolam with several closed loops still obeys every rule. Single-line kolams are a stricter, prouder subset, as the gcd condition shows.
- Memory. The Apple logo data say a simple mark can still be badly remembered. Single-stroke is a structural property, not a memory guarantee.

Every sikku kolam is one stroke or a set of closed strokes because each crossing has four line-ends, the exact condition Euler proved in 1736; a single loop needs only connectivity, which for Hridaya kolams means gcd(dots, arms) = 1. A logo meets the same rule only if it has 0 or 2 odd points, so O, S and 8 pass and H needs 3 strokes. The rule guarantees coherence and reproducibility, not memorability: only 1 in 85 people could draw the simple Apple logo correctly. Change "a good logo" to "a well-constructed mark", and the claim holds.
For the practical side, see the logo design process, vector vs raster for logos and why logos look dated. The story of a single mark designed for a whole country is in the rupee symbol, and floor and wall art's longer history is in Madhubani and Warli. More from the studio at Kala.
What's new
- A neural network draws one-stroke sikku kolams with 501 dots, in a study in Heritage Science ThePrint (PTI), Mar 2026
- Hridaya kolams with non-coprime dots and arms split into several closed loops arXiv, Oct 2025
- Single-stroke Hridaya kolams extended to even numbers of dots arXiv, Jun 2025
- Pulli kolam proposed as a craft medium for showing data arXiv (CHI 2026 workshop), Mar 2026
What if next?
- Truck art and app icons solve the same problem: both have under a second to be recognised among a hundred others.
- Your eye and a camera both lie in the same places: a hole in your vision about 5.5 degrees wide, filled in so well you never see it.
- A sari and a sports bike are both exercises in the same curve: the other half of line-drawing maths, where curvature matters instead of junctions.
Questions people ask
What is the difference between a pulli kolam and a sikku kolam?
A pulli kolam is any kolam built on a grid of dots. A sikku (knot) kolam is the kind where a line loops around each dot and crosses itself without loose ends, often as one continuous line.
What is an Eulerian path?
A route through a network that uses every connection exactly once. It exists in a connected network when zero or two points have an odd number of connections; with zero, the route can end where it started.
Can every kolam be drawn in one line?
No. Every crossing has an even number of line-ends, so a kolam never has loose ends, but it can split into several separate loops. For Hridaya kolams, one loop needs the number of dots per arm and the number of arms to share no common factor.
Is there mathematics in kolam?
Yes. Kolams have been studied as picture languages, Eulerian graphs, knot diagrams and symmetry groups since the 1970s, from Gift Siromoney's work in Chennai to the 2023 to 2025 arXiv papers cited here.
Sources
- Wikipedia: Seven Bridges of Königsberg (source for Euler's 1735 reasoning and the 5, 3, 3, 3 bridge counts)
- Wikipedia: Eulerian path (source for the theorem, Euler's 1736 paper, Hierholzer's 1873 proof and the handshake lemma)
- Gopalan: Symmetry Classification and Enumeration of Square-Tile Sikku Kolams, arXiv (2023) (source for the three rules, tile model, parity condition and 260 million figure)
- Chakraborty and Manna: Extending Hridaya Kolam to Even-Ordered Dot Patterns, arXiv (2025) (source for the gcd condition and single-stroke construction)
- Manna and Chakraborty: Extending Hridaya Kolam to Multiple Loops, arXiv (2025) (source for multiple loops when gcd is not 1)
- Whatley, Schwartz, Block and Castel: Memory, metamemory, and false memory for features of the Apple logo, Applied Cognitive Psychology (2023) (source for the 1 in 85 and 40% figures from Blake et al. 2015)
- Wikipedia: Kolam (source for regions, materials and the sikku pattern)
- ThePrint: Now AI can draw pulli kolams beyond human imagination (2026) (source for the 19 and 31 dot limits and the 501-point kolam)
- Ramesh and Rajabiyazdi: Pulli Kolam, a traditional South Indian craft practice for representing data, arXiv (2026)
