Triangle Center Calculator: How to Find the Centroid, Circumcenter, Incenter, and Orthocenter
About the Author
Marko Šinko
Co-Founder & Lead Developer, AI Math Calculator
A triangle center calculator has to start by answering a question most people never think to ask: which center? A triangle doesn’t have one middle. The Encyclopedia of Triangle Centers, the reference catalogue kept by Clark Kimberling at the University of Evansville, lists more than fifty thousand of them. Four do essentially all the real work — centroid, circumcenter, incenter, orthocenter — and on anything but an equilateral triangle they sit in four different places. On a plain 3-4-5 right triangle they spread 2.5 units from end to end, which is exactly half the hypotenuse. Pick the wrong one and you aren’t off by a rounding error. You’re off by a distance comparable to the triangle itself.
Four Centers, One 3-4-5 Triangle
Put the right angle at the origin: A(0, 0), B(4, 0), C(0, 3). Legs of 3 and 4, hypotenuse of 5, area of 6. Now compute all four centers on the same shape and watch them scatter.
| Center | Where three lines meet | Coordinates | Landed |
|---|---|---|---|
| Centroid G | Medians (vertex to opposite midpoint) | (1.333, 1) | Inside |
| Circumcenter O | Perpendicular bisectors of the sides | (2, 1.5) | On the hypotenuse |
| Incenter I | Angle bisectors | (1, 1) | Inside |
| Orthocenter H | Altitudes (vertex perpendicular to opposite side) | (0, 0) | On vertex A |
Three of those four are worth checking by hand, because each is a small proof in itself. The circumcenter sat at (2, 1.5), the midpoint of the hypotenuse — that is Thales’ theorem showing up as a coordinate. The orthocenter collapsed onto the right angle, because two of the three altitudes are the legs. And the incenter came out at (1, 1), one unit from each leg, which matches the inradius r = Area ÷ s = 6 ÷ 6 = 1 exactly. Only the centroid needed real arithmetic.
The Coordinate Formulas Behind a Triangle Center Calculator
Name the sides for the vertex they face — a = |BC|, b = |CA|, c = |AB| — and three of the four centers reduce to short expressions. Use the distance formula calculator to get those side lengths first if you are working on paper.
G = ( (x₁+x₂+x₃)/3 , (y₁+y₂+y₃)/3 )
I = ( (a·x₁ + b·x₂ + c·x₃)/(a+b+c) , (a·y₁ + b·y₂ + c·y₃)/(a+b+c) )
O: solve |O−A|² = |O−B|² = |O−C|² (a 2×2 linear system)
H = 3G − 2O
That last line is the shortcut worth stealing. Textbooks derive the orthocenter by intersecting two altitudes, which means writing two perpendicular slopes and solving a system — roughly fifteen lines of algebra with plenty of sign traps. But once you have G and O, the orthocenter is a single subtraction, because all three lie on the Euler line with a fixed ratio between them. On the 3-4-5 triangle: H = 3(1.333, 1) − 2(2, 1.5) = (4, 3) − (4, 3) = (0, 0). One line, no slopes, no negative reciprocals to get backwards.
The circumcenter is the only one that genuinely needs a system, and there is a determinant form that avoids the algebra. With d = 2·[x₁(y₂−y₃) + x₂(y₃−y₁) + x₃(y₁−y₂)], the circumcenter’s x-coordinate is [(x₁²+y₁²)(y₂−y₃) + (x₂²+y₂²)(y₃−y₁) + (x₃²+y₃²)(y₁−y₂)] ÷ d, with the mirrored expression for y. That d in the denominator is twice the cross product, which is also why it doubles as the degeneracy test — more on that at the end.
Why Averaging the Vertices Only Works for the Centroid
Here is the single most common error in coordinate geometry homework: averaging the three vertices and calling the result “the center.” That average is the centroid, and only the centroid. On our 3-4-5 triangle it gives (1.333, 1), while the incenter is at (1, 1) — off by a third of a unit, which is a third of the entire inradius.
The incenter is an average too, just a weighted one. Each vertex is weighted by the length of the side opposite it, so the longest side pulls the incenter toward the corner facing it. In the 3-4-5 case the weights are 5, 3 and 4 against vertices (0,0), (4,0) and (0,3), which is why the answer drifts toward the right angle rather than sitting where the plain average would put it. Weight by something else and you get a different center entirely — that is essentially how the fifty-thousand-entry catalogue is built.
Two more facts about the centroid are worth knowing precisely, because both get stated loosely and both matter in physics problems:
- It is the balance point of a solid triangular plate, and also of three equal masses pinned at the vertices. Both give the same point.
- It is not the balance point of a triangular wire frame. A frame made of three uniform rods balances at the Spieker center, the incenter of the medial triangle. Cut a triangle out of card and it balances at G; bend three equal-density rods into the same shape and it does not.
- It minimises the sum of squared distances to the three vertices — not the sum of distances. The point that minimises the plain sum is the Fermat point, which is a different location on any triangle with no angle above 120°.
The centroid also cuts each median in a 2:1 ratio, sitting twice as far from the vertex as from the opposite midpoint. That gives you a free check: compute the midpoint of a side with a midpoint formula calculator, then confirm G lies two-thirds of the way from the vertex to it.
Which Centers Leave the Triangle
Two of the four can wander outside the shape they belong to, and students routinely assume a negative or out-of-range answer means they made a mistake. Often it means the triangle is obtuse. The behaviour is completely predictable:
| Center | Acute | Right | Obtuse |
|---|---|---|---|
| Centroid | Inside | Inside | Inside |
| Incenter | Inside | Inside | Inside |
| Circumcenter | Inside | Midpoint of the hypotenuse | Outside |
| Orthocenter | Inside | At the right-angle vertex | Outside |
Notice that the circumcenter and orthocenter always change status together. That is not a coincidence — they swap places through the centroid, so if one crosses a side, the other has to as well. The right-triangle column is the crossing moment caught mid-flip: O has arrived exactly on the hypotenuse and H exactly on a vertex. Try the “Obtuse” preset above and both dots jump clear of the outline while the green and amber ones stay put.
There is a practical payoff hiding in the right-triangle row. Because O is the midpoint of the hypotenuse, the circumradius of any right triangle is exactly half the hypotenuse. A right triangle calculator gives you c; halve it and you have R without touching the abc ÷ 4K formula.
The Euler Line and Its Stubborn 2:1 Ratio
Leonhard Euler proved in 1765 that the circumcenter, centroid and orthocenter of every triangle fall on one straight line, with the centroid always sitting twice as far from H as from O. Not approximately. Exactly, on every triangle that exists. Check it on our 3-4-5 numbers: O(2, 1.5), G(1.333, 1), H(0, 0).
O → G = (−0.667, −0.5)
G → H = (−1.333, −1) = 2 × (−0.667, −0.5)
|OG| = 0.8333 |GH| = 1.6667 ratio = 2.0000
|OH| = 2.5, which happens to equal R here
The midpoint of O and H has a name of its own: the nine-point center. The circle drawn around it with radius R ÷ 2 passes through nine specific points — the three side midpoints, the three feet of the altitudes, and the three midpoints between H and each vertex. On the 3-4-5 triangle that centre is (1, 0.75) with radius 1.25, and you can verify it by hand: the midpoint of the hypotenuse is (2, 1.5), and its distance from (1, 0.75) is √(1 + 0.5625) = 1.25. Every one of the nine lands on the same circle. The Euler line carries the nine-point center too, at the exact midpoint between the two outliers.
The incenter is the notable absentee. It joins the line only when the triangle is isosceles, in which case the axis of symmetry swallows every center at once. And on an equilateral triangle the Euler line disappears entirely, because all four centers collapse onto a single point and no unique line passes through one point. If your calculator reports an undefined Euler line, check whether you typed an equilateral triangle before assuming it broke.
Equidistant From the Corners, or From the Sides?
Outside a geometry class, this is the only question that matters, and it has a clean answer. The circumcenter is equidistant from the three corners. The incenter is equidistant from the three sides. Those are different jobs, and picking by feel is how siting decisions go wrong.
| You need… | Use | Because it minimises |
|---|---|---|
| One point reaching three fixed locations | Circumcenter | The worst-case distance to a corner |
| The largest circle that fits inside a plot | Incenter | Nothing — it maximises clearance from every edge |
| A pivot, balance point or centre of mass | Centroid | The sum of squared distances to the corners |
| Heights, altitudes, or a proof about them | Orthocenter | Nothing physical — it is a construction point |
Run the numbers on a real siting problem and the gap is easy to price. Three villages sit at (0, 0), (12, 0) and (4, 9), distances in kilometres, and you need one transmitter that reaches all three. The circumcenter lands at (6, 2.722) with every village exactly 6.589 km away. Put the mast at the centroid instead — (5.333, 3), the intuitive “middle” — and the nearest village is 6.12 km off but the farthest is 7.31 km. You now need 11% more range, and since coverage area scales with the square of the radius, roughly 23% more radiated power to serve the same three villages. Same triangle, same three villages, one wrong center.
The incenter answers the opposite question. Given a triangular plot, the largest circular structure that fits is centred on the incenter with radius r = Area ÷ s. Compute the area with Heron’s formula when you only have the three side lengths, or with the full triangle calculator when you have a mix of sides and angles.
Flat Triangles and Runaway Circumradii
Every triangle center formula divides by something that vanishes when the three points fall on a line. For the circumcenter it is d = 2·[x₁(y₂−y₃) + x₂(y₃−y₁) + x₃(y₁−y₂)], which is twice the cross product and therefore four times the area. Hit zero and the circumcenter and orthocenter genuinely do not exist — parallel lines never meet.
The dangerous case is not exactly collinear, it is almost collinear, because nothing errors out. Take A(0, 0), B(10, 0), C(5, 0.001). The area is 0.005, and R = abc ÷ 4K = 250 ÷ 0.02 = 12,500. The circumcenter sits at roughly (5, −12500), twelve and a half thousand units below a triangle that is ten units wide, and it is technically correct. Shift that third y-coordinate to 0.002 and R halves to 6,250. A change in the fifth decimal place of an input moved the answer by six thousand units, which is why survey and CAD pipelines flag sliver triangles before computing anything about them.
Three more failure modes worth recognising:
- Duplicate vertices. Two identical points make one side length zero, the perimeter collapses toward the remaining sides, and the incenter formula divides by a weight sum that no longer describes a shape.
- Mislabelled sides. The incenter weights must pair a with vertex A, and a is the side opposite A, not the one starting at it. Swap the convention and the answer stays plausible-looking while being wrong — the most expensive kind of error. The law of cosines uses the same convention, so if the two agree you probably have it right.
- Assuming vertex order matters. It doesn’t. Listing the vertices clockwise instead of counterclockwise flips the sign of the cross product but leaves all four centers untouched, since each formula either uses |d| implicitly or divides two sign-flipped quantities. That is the opposite of the shoelace formula, where orientation flips the sign of the result.
None of this is exotic mathematics. It is one triangle, four ways of splitting it in half, and a stubborn line that refuses to let three of the answers stray apart. Enter your coordinates above, toggle the medians and altitudes on, and watch the four dots move as you drag a vertex around — the pattern becomes obvious long before the algebra does.
