Lengthbashing

mira74

August 2022

Thanks to Luke Robitaille for reading this and giving lots of helpful suggestions. Also evan.sty <3.

Contents
1 Philosophy 2

2 Disclaimer-ish Things 4
2.1 Directed Angles and Lengths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Citing Stuff . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 Maybe Don’t Bash Too Hard Or Something . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.4 For Some Definition of Bashing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Lengths and the (Normal) Ratio Lemma 5

4 Quadrilateral Ratios, Ceva and Menelaus 8

5 Ratios on a Circle 12

6 Linearity of Power of a Point 13

7 Two Length Conditions 17

8 Appendix 1: General Advice and Random Things not in the Handout 20

9 Appendix 2: My IMO 2021/3 Thought Process 22

10 Appendix 3: Actually Writing Out Proofs 25

1
mira74 1 Philosophy

1 Philosophy
In some sense, geometry problems are often about trying to “understand” points. It’s been a while since I
wasn’t in the mindset of lengthbashing problems, but if my memory serves me correctly, the go-to ways of
thinking about points are:

• What lines and circles does it lie on?

• What angles does it form with other points?

• What are its cross ratios with other points?

People generally know lots of ways of working with the above types of information. The main goal of this
handout is to help you deal with the following types of information:

• What is its distance to other points?

• What is the ratio of its distances to other points?

These techniques are often powerful because, if the construction of something “isn’t that bad,” we can often
easily deal with it. You can view some sections of this handout as picking a particular way of thinking about
points, then finding theorems that make this way of thinking about points viable.

Ok so here’s an example of what I mean.

Example 1.1 (Iran Lemma)

Let ABC be a triangle. Let its incircle with center I touch CB, CA at D, E, and let the midpoints of
BC and BA be M and P . Prove that AI, DE, and M P concur.

E
P
I

B D M C

These lines aren’t related in any super obvious ways. However, if you look at it from the perspective of A,
none of them are actually that bad:

• For AI - walk at angle A/2 to AB and AC.

• For M P - start c/2 along AB, and walk at an angle of A away from AB.
• For DE - start s − a along AC, and walk at an angle of 90 − C/2 away from AC.

2
mira74 1 Philosophy

Hence, we should probably be able to deal with the distance from A to the concurrence point. This intuition
translates to the following proof:

Proof. Let M P meet AI at T , and let DE meet AI at T 0 . By the Law of Sines,

sin(∠AP T ) c sin(A)
AT = AP · = ·
sin(∠AT P ) 2 sin(A/2)
sin(∠AET 0 ) cos(C/2)
AT 0 = AE · = (s − a) ·
sin(∠AT 0 E) sin(B/2)

It now suffices to check that these are equal. At this point, we can convert everything in terms of half angles.
We have
 
cos(A/2) cos(B/2) sin(A/2 + B/2)
s − a = r cot(A/2) =⇒ c = (s − a) + (s − b) = r + =r· .
sin(A/2) sin(B/2) sin(A/2) sin(B/2)
cos(A/2) cos(C/2)
Plugging these back in for AT and AT 0 gives that they’re both r · sin(A/2) sin(B/2) , as desired.

A key takeaway from this proof is that throughout the proof, we try to “delete” more and more things. With
the length computations, we reduce the problem to something only referencing the triangle. Then, we try
converting all of our lengths to just r and trig functions of half angles.

The final computation may seem a bit tricky, but it’s worth noting that:

• There’s actually almost nothing going on. We already know that the two expressions have to be equal,
and none of the terms are unusual.
• Surprisingly, you won’t have to do things like that often.

• For a better finish, you can use the fact that AET 0 ∼ AIB. The intuition here is “if trig ratios are
surprisingly nice, you can often try to make them show up naturally.” Being able to see this more easily
probably just comes from practice.

3
mira74 2 Disclaimer-ish Things

2 Disclaimer-ish Things
2.1 Directed Angles and Lengths
This handout is going to use a lot of trig. The issue is that, as far as I know, there’s no canonical fix to the
following fact:

If we want to use directed angles, we need to consider angles mod 180◦ . However, sin(x) 6=
sin(x + 180◦ ).

I have a way of fixing this, see this handout, but its pretty annoying. I’m not going to direct angles and
lengths and trig in this handout since:

• I want this handout to be comprehensible without reading the directed trig and lengths handout.
• Honestly, it’s pretty likely best to try WLOG-ing away (or just ignoring if it’s obvious enough)
configuration issues for actual writeups.

But I still think the fact that we can get stuff to work with directed angles is pretty cool, and also works
surprisingly well. For example, we end up getting injectivity statements like the following:

• BP/CP is a bijection between line BC and R ∪ ∞ (directing lengths just on BC suffices for this).
• For the definition of sin in the handout, if sin α = sin β, then α ≡ β (mod 180◦ ).

2.2 Citing Stuff

Lots of claims on this handout aren’t citable. I think the law of sines, Ceva, Menelaus, (normal) ratio lemma,
and the statement of linearity of power of apoint are all safely citable. For other things, you’re probably best
off writing out the statement of the fact you’re trying to cite and its proof.

2.3 Maybe Don’t Bash Too Hard Or Something

Synthetic is good. Bashing is also good. The issue is that bash solutions have pretty unbounded ability to
get time-consuming, and synthetic solutions don’t. A few corollaries of this are:

• You should eventually get an intuitive grasp on how hard things are to compute. If something’s too
hard, it may be a good idea to look for better solutions (including better bashes).

• Trying bash on every problem before getting pretty good at angle chasing is dangerous.

2.4 For Some Definition of Bashing

Most of the “bashing” in this handout is very very clean. You won’t even have to add numbers that often.
Calling it bashing is kind of questionable but I’m too used to it to stop.

4
mira74 3 Lengths and the (Normal) Ratio Lemma

3 Lengths and the (Normal) Ratio Lemma

The first basic tool we use is the Law of Sines:

Theorem 3.1 (Law of Sines)

Let ABC be a triangle. Then, we have
a b c
= = .
sin A sin B sin C

Without actually going through any computations, we can think of this as:

• If we can deal with the angles, we can deal with the ratios of the side lengths.
• If we can deal with the ratio of two of the angles, we can deal with the ratio of the corresponding sides.

• If we can deal with the ratio of two of the sides, we can deal with the ratio of the corresponding angles.

In our proof of the Iran Lemma, before actually computing AT and AT 0 , we know that they have to be easy
to deal with. We can deal with all the angles of AP T and AET 0 , and we can deal with AP and AE.
We’ll now find ways to deal with length and sine ratios turn, which out to be very useful.

Theorem 3.2 (Ratio Lemma)

Let 4ABC be a triangle, and let D be a point on BC. Then, we have
BD BA sin ∠BAD
= · .
CD CA sin ∠CAD

B D C

The proof is left as an exercise. It’s also written out in the third appendix. Intuitively, what this means is:

• If we can deal with BA/CA and BD/CD, we can find the “direction” of AD in ∠BAC.

• If we can deal with BA/CA and the “direction” of AD in ∠BAC, then we can deal with BD/CD.

We’re thinking about the trig ratio as the direction since it only depends on angles at A.
Here’s an example problem:

5
mira74 3 Lengths and the (Normal) Ratio Lemma

Example 3.3
Let ABC be a triangle with incenter I. Let its incircle meet BC, CA, AB at D, E, F , and let M be the
midpoint of BC. Then, AM , DI, and EF concur.

E
T

B D M C

Without actually computing anything, here’s how we’re going to deal with everything:

• Since ABC is well-behaved, and BM : CM is just 1, we can deal with the direction of AM in ∠BAC.
• Since we can deal with the direction of AM , and AEF is well-behaved, we can find the ratio that AM
splits EF into.
• Since all the angles between D, E, F, I are nice, we can deal with all the length ratios and directions, so
we can find the ratio that DI splits EF into.

The above should intuitively be enough to solve the problem, perhaps up to some conversion at the end.
Here’s the solution once we actually go through the computation:

Proof. Let AM and DI meet EF at T and T 0 . Our first step is:

sin(BAM ) BM AC AC
= · = .
sin(CAM ) CM AB AB
Our second step is:
ET sin(EAM ) EA AB
= · = .
FT sin(F AM ) F A CB
Our last step is:
ET 0 ED sin(EDI) cos(C/2) sin(C/2) sin(C)
= · = · = .
FT0 F D sin(F DI) cos(B/2) sin(B/2) sin(B)
ET ET 0
Hence, FT = FT0 , so they are the same point, as desired.

For points B, C, the function mapping T to BT /CT for points T on segment BC is injective, making the
last line of the proof above valid. If I was properly directing lengths (or even just directing lengths on line
BC), it would in fact be a bijection. This means we can actually just think about points T on line BC by
thinking about the signed value of BT /CT , which is nice.

6
mira74 3 Lengths and the (Normal) Ratio Lemma

Problem 3.4. Prove the Ratio Lemma.

Problem 3.5 (Steiner’s Theorem). Let ABC be a triangle. Let P, Q be points on segment BC such that
BQ BA2
∠BAP = ∠CAQ. Prove that BP CP · CQ = CA2

Problem 3.6 (IMO 2018). Let Γ be the circumcircle of acute triangle ABC. Points D and E are on segments
AB and AC respectively such that AD = AE. The perpendicular bisectors of BD and CE intersect minor
arcs AB and AC of Γ at points F and G respectively. Prove that lines DE and F G are either parallel or
they are the same line.

Problem 3.7 (USAMTS Year 32). Let ABC be a triangle with AB < AC. T is the point on BC such that
AT is tangent to the circumcircle of 4ABC. Additionally, H and O are the orthocenter and circumcenter of
4ABC, respectively. Suppose that CH passes through the midpoint of AT . Prove that AO bisects CH.
Problem 3.8 (XVII Olimpı́ada Matemática Rioplatense (2008)). In triangle ABC, where AB < AC, let
X, Y , Z denote the points where the incircle is tangent to BC, CA, AB, respectively. On the circumcircle
of ABC, let U denote the midpoint of the arc BC that contains the point A. The line U X meets the
circumcircle again at the point K. Let T denote the point of intersection of AK and Y Z. Prove that XT is
perpendicular to Y Z.
Problem 3.9 (USAMO 2013). In triangle ABC, points P , Q, R lie on sides BC, CA, AB respectively.
Let ωA , ωB , ωC denote the circumcircles of triangles AQR, BRP , CP Q, respectively. Given the fact that
segment AP intersects ωA , ωB , ωC again at X, Y , Z, respectively, prove that Y X/XZ = BP/P C.
Problem 3.10 (IMO Shortlist 2019). Let ABC be an acute-angled triangle and let D, E, and F be the feet
of altitudes from A, B, and C to sides BC, CA, and AB, respectively. Denote by ωB and ωC the incircles of
triangles BDF and CDE, and let these circles be tangent to segments DF and DE at M and N , respectively.
Let line M N meet circles ωB and ωC again at P 6= M and Q 6= N , respectively. Prove that M P = N Q.

7
mira74 4 Quadrilateral Ratios, Ceva and Menelaus

4 Quadrilateral Ratios, Ceva and Menelaus

We now extend our method of computing the “direction” of AD in ∠BAC to cases where D isn’t on line BC.

Theorem 4.1 (Quadrilateral Ratio Formula)

Let ABC be a triangle, and let D be a point. Then, we have

sin(BAD) sin(ABD) BD sin(ABD) sin(BCD)

= · = · .
sin(CAD) sin(ACD) CD sin(ACD) sin(CBD)

B C

Proof is left as an exercise. It’s also written out in the third appendix. Although this is really similar to
Ceva’s theorem (the latter equality is equivalent), it makes sense to think about it separately, since we’re
looking “from A”. Intuitively, this theorem states:

If rays BD and CD aren’t too bad, and we can either deal with BD/CD or the angles between
BC and DB, DC, we can find the direction of AD in ∠BAC.

Having this theorem greatly expands the set of ratios we can deal with. Here’s an example:

Example 4.2 (IMO Shortlist 2006)

Circles w1 and w2 with centres O1 and O2 are externally tangent at point D and internally tangent to a
circle w at points E and F respectively. Line t is the common tangent of w1 and w2 at D. Let AB be
the diameter of w perpendicular to t, so that A, E, O1 are on the same side of t. Prove that lines AO1 ,
BO2 , EF and t are concurrent.

8
mira74 4 Quadrilateral Ratios, Ceva and Menelaus

By homothety at E, F , we have that ED and F D pass through B and A, respectively.

F
T
E

O1 D O2

A B

Now, without explicitly doing any computations, we check if we can deal with ET /F T , where T is the desired
concurrence point. A pretty natural way of constructing the diagram is starting with A, B, E, F and then
constructing everything else, so we’ll try finding things in terms of those points. We have:

• ∠AEO1 , ∠ADO1 , AD, AE, and EO1 : DO1 are all reasonable. Hence, we can compute the direction
of AO1 in ∠EAF , so we can compute the ratio AO1 splits EF into.
• The above holds for BO2 .
• All the angles between D, E, F and the common tangent are nice, so we can compute the ratio the
common tangent splits EF into.

It looks like we have enough info to prove the result. Filling in the details gives this proof:

ET AE·BE
Proof. We claim that they concur at the point T on EF such that FT = AF ·BF . We have

sin(EAO1 ) sin(AEO1 ) DO1 sin(BAE) BE AE sin(EAO1 ) AE · BE

= · = = =⇒ = .
sin(F AO1 ) sin(ADO1 ) EO1 sin(F AB) BF AF sin(F AO1 ) AF · BF

Similarly, BO2 passes through that point. Finally, if T meets EF at T 0 , we have

ET 0 ED sin(EDT 0 ) EA sin(BAE) EA BE ET
0
= · 0
= · = · = ,
FT F D sin(F DT ) F B sin(ABF ) F B AF FT

as desired.

Finally, we’ll see how to convert common goals of geo problems to ratios:

9
mira74 4 Quadrilateral Ratios, Ceva and Menelaus

Theorem 4.3 (Ceva’s Theorem)

Let ABC be a triangle, and let D, E, F be points. Then, AD, BE, and CF concur iff

sin(BAD) sin(CBE) sin(ACF )

· · = 1.
sin(CAD) sin(ABE) sin(BCF )

Theorem 4.4 (Menelaus Theorem)

Let ABC be a triangle, and let D, E, F be points on BC, CA, and AB. Then, DEF are collinear iff
BD CE AF
· · = 1.
CD AE BF

Thinking with the tools we’ve developed, Ceva’s theorem states:

If we want to prove AD, BE, CF concur, it suffices to find the directions of AD in ∠BAC and
similar for the others.

Menelaus can be thought of similarly, though it’s restricted to the case where D, E, F are on the opposite
sides.

Example 4.5 (Jacobi’s Theorem)

Let ABC be a triangle. Points X, Y , Z are chosen such that ∠BAZ = ∠CAY = α, for some α, and
there exist β and γ with similar properties. Show that AX, BY , and CZ concur.

A
Y
Z

B C

X
10
mira74 4 Quadrilateral Ratios, Ceva and Menelaus

The key observation here is that since X is defined by given angles off of AB, AC, and BC, the direction
of AX should be manageable. Similarly, the directions of BY and CZ should be manageable. By Ceva’s
theorem, we know that this is enough to check concurrence.
Now, we can just go through the computations.

Proof. By our Quadrilateral Ratio Formula (in its second form),

sin(BAX) sin(BCX) sin(ABX) sin γ sin(B + β)

= · = · .
sin(CAX) sin(CXB) sin(ACX) sin β sin(C + γ)

The cyclic product of this is 1, so by Ceva’s theorem, we are done.

Problem 4.6. Prove the Quadrilateral Ratio Formula
Problem 4.7. Let ABC be a triangle. Let the feet of the altitudes from B and C to AC and AB be E and
F , respectively. Let D be the intersection of the tangents to the circumcircle of ABC at B and C. Prove
that AD bisects EF .

Problem 4.8 (Simson Lines). Let ABCD be a cyclic quadrilateral. Prove that the feet of the altitudes from
D to the sides of ABC are collinear.
Problem 4.9 (Steinbart’s Theorem). Let ABC be a triangle with incircle ω. Let ω meet BC, CA, and AB
at D, E, and F . Suppose X, Y , and Z are chosen on ω such that DX, EY , and CZ concur. Show that AX,
BY , and CZ concur.

Problem 4.10 (Fall SDPC 2019-2020). Let 4ABC be an acute, scalene triangle with orthocenter H, and
let AH meet the circumcircle of 4ABC at a point D 6= A. Points E and F are chosen on AC and AB such
that DE ⊥ AC and DF ⊥ AB. Show that BE, CF , and the line through H parallel to EF concur.
Problem 4.11. Let ABC be a triangle. Suppose the Euler line of ABC meets BC at P . In terms of the
angles of ABC, compute BP
CP .

Problem 4.12 (APMO 2021). Let ABCD be a cyclic convex quadrilateral and Γ be its circumcircle. Let
E be the intersection of the diagonals of AC and BD. Let L be the center of the circle tangent to sides
AB, BC, and CD, and let M be the midpoint of the arc BC of Γ not containing A and D. Prove that the
excenter of triangle BCE opposite E lies on the line LM .

Problem 4.13 (ELMO Shortlist 2019). Let ABC be an acute triangle with orthocenter H and circumcircle
Γ. Let BH intersect AC at E, and let CH intersect AB at F . Let AH intersect Γ again at P =
6 A. Let P E
intersect Γ again at Q 6= P . Prove that BQ bisects segment EF .
Problem 4.14 (TSTST 2020). Let A, B, C, D be four points such that no three are collinear and D is
not the orthocenter of ABC. Let P , Q, R be the orthocenters of 4BCD, 4CAD, 4ABD, respectively.
Suppose that the lines AP , BQ, CR are pairwise distinct and are concurrent. Show that the four points A,
B, C, D lie on a circle.
Problem 4.15 (Balkan MO 2019). Cevians AP and AQ of a triangle ABC are symmetric with respect to
its bisector. Let X, Y be the projections of B to AP and AQ respectively, and N , M be the projections of
C to AP and AQ respectively. Prove that XM and N Y meet on BC.

11
mira74 5 Ratios on a Circle

5 Ratios on a Circle
The exact set of content I’d want to include here (which is probably too big for one section anyways) is
included in my ratio lemma handout. The handout is not too well titled. What it actually does is, given
points on a line ` or circle ω through points B, C:

1. Shows how to get between BP/CP , BQ/CQ, and BR/CR for collinear points P, Q, R with P, Q on ω
and R on `
2. Shows how to, given only the ratios above, check if P QRS is cyclic, where we have that BCP Q and
BCRS are both either collinear or concyclic.
3. Given a triangle ABC and points E, F, P on AC, AB, (ABC) such that AEF P is cyclic, shows how to
compute BP/CP

The tools we developed in the previous 2 sections are really useful for this; think about, for example, problem
3.8. I’ll also add the following remark:

Theorem 5.1
Let ABC be a triangle, and let P be any point on its circumcircle. Then, you can go between AP : BP ,
BP : CP , and CP : AP by Ptolemy’s theorem.

This is useful since it allows us to go between ratios to different points on a circle.

For those of you who’ve already done the handout, here are some more problems.
Problem 5.2 (Sharygin 2022). Let ABCD be a cyclic quadrilateral, E = AC ∩ BD, F = AD ∩ BC. The
bisectors of angles AF B and AEB meet CD at points X, Y . Prove that A, B, X, Y are concyclic.
Problem 5.3 (Jerabek’s Theorem). Let ABC be a triangle, and let P, Q be points in its interior. Suppose
AP and AQ meet the circumcircle of ABC again at P1 and Q1 , and let P1 Q1 meet BC at A0 . Define B 0 and
C 0 similarly. Prove that A0 , B 0 , and C 0 are collinear.
Problem 5.4 (AoPS). Let 4ABC have orthocenter H, circumcenter O, and let O0 be the reflection of O
through BC. Let E and F be the intersections of BO0 and CO0 with AC and AB, respectively. Prove that
the circle with diameter AH, the circumcircle of ABC, and the circumcircle of AEF are coaxial.

Problem 5.5 (USAMTS Year 30). Acute scalene triangle 4ABC has circumcenter O and orthocenter H.
Points X and Y , distinct from B and C, lie on the circumcircle of 4ABC such that ∠BXH = ∠CY H = 90◦ .
Show that if lines XY , AH, and BC are concurrent, then OH is parallel to BC.
Problem 5.6. Let ABC be a triangle with incenter I, let the external angle bisector of ∠A meet the
circumcircle at M , and let A0 be the antipode of A in the circumcircle of ABC. Prove that A, the second
intersection of A0 I and the circumcircle, the midpoint of BC, and the intersection of M I and BC are
concyclic.
Problem 5.7 (Sharygin 2020). Let H be the orthocenter of a nonisosceles triangle ABC. The bisector of
angle BHC meets AB and AC at points P and Q respectively. The perpendiculars to AB and AC from P
and Q meet at K. Prove that KH bisects the segment BC.

12
mira74 6 Linearity of Power of a Point

6 Linearity of Power of a Point

Let Pow(P, ω) denote the power from P to ω. Linearity of pop refers to the following observation:

Theorem 6.1
For any two circles ω1 , ω2 , the function f (P ) = Pow(P, ω1 ) − Pow(P, ω2 ) is a linear function of P (i.e.
linear in the coordinates of P ).

Proof. If ω has equation x2 + y 2 + ax + by + c = 0, the power from P = (x0 , y0 ) to ω is x20 + y02 + ax0 + by0 + c.
Subtracting two such expressions gives a linear function in x0 and y0 , as desired.

One can think of this as “Pow(P, ω) is a linear function plus some term that’s universal for all circles.”

One common use of linearity of pop is computing powers at random points. In particular, we have:

If Z = xP + yQ + zR, then f (Z) = xf (P ) + yf (Q) + zf (R). Hence, if we know the differences of

powers at 3 points, we can compute that difference everywhere.

Since there are good and relatively known handouts regarding this (see Kagebaka’s, for example), I’ll focus
on different uses of linearity of pop. It’s a really versatile tool, and probably the hardest technique on this
handout to get used to. Loosely, “it generally helps you deal with power-of-a-point information.”
The following special case is very useful, and works well with previous sections of this handout:

We can go between power differences and length ratios on a line! If P, Q, T are on a line, then
PT f (P ) − f (T )
= .
QT f (Q) − f (T )

Here’s an example of a problem in which this is useful:

Example 6.2 (Brazil Olympic Revenge 2017 (sorta))

Let ABC be an acute triangle and I its incenter. Let E and F be the intersections of BI and CI with
sides AC and AB. Let M and N be the midpoints of minor arcs AB and AC and let P be a point on
BC such that AP is tangent to AEF . Prove that M , N , and P are collinear.

N
A

M E
F

P B C

13
mira74 6 Linearity of Power of a Point

Relabel P to be on BC such that AP is tangent to (AEF ). Line M N is pretty well behaved, so the hard
part of this problem is to get some grasp on P . The key idea is the following characterization:

P is the point on BC such that P A2 is equal to the power from P to (AEF ).

The motivation here is that A and (AEF ) are both reasonable to access from B and C. Now, defining f (X)
to be XA2 − Pow(X, (ABC)), we have
BP f (B) AB · AF
= = ,
CP f (C) AC · AE
which is just something in terms of the side lengths of ABC. We can also compute the ratio M N splits BC
into by, say, ratio lemma. This should be enough to solve the problem.

Proof. Defining f (X) to be XA2 − Pow(X, (AEF )), we have

BP f (B) AB · AF c a+c
= = = · .
CP f (C) AC · AE b a+b
BM BN
It suffices to check that this is equal to CM · CN . By Ptolemy, we have
BM c
c · CM = BM · b + AM · a = (a + b)BM =⇒ = ,
CM a+b
BM BM BN BP
and similar for CN , giving CM · CN = CP , as desired.

If the choice of f still feels too unmotivated, it may be worth noting that the choice of function “difference
between power to a circle and power to a point on the circle” is, in general, very useful. We have the following:

Corollary 6.3
Let A be a point on a circle ω. Let P be a (non-fixed) point, and let AP meet ω at a point Q other
than A. Then, AP · AQ is a linear function of P .

Proof. Rewrite AP · AQ as AP 2 − P Q · P A = AP 2 − Pow(P, ω)2 . Conclude by linearity of pop.

We’ll now see another problem that uses a similar application of linearity of power.

14
mira74 6 Linearity of Power of a Point

Example 6.4 (IMOC Shortlist 2021)

Let I be the incenter of the acute triangle 4ABC, and BI, CI intersect the altitude of 4ABC through
A at U , V , respectively. The circle with AI as a diameter intersects (ABC) again at T , and (T U V )
intersects the segment BC and (ABC) at P , Q, respectively. Let R be another intersection of P Q
and (ABC). Show that AR k BC.

A R
T

V
I
U

S B P C

By Reim’s, AR||P Q is equivalent to AT ∩ BC also lying on (T U V ). Let S = AT ∩ BC. By LOS, we should

be able to deal with AU · AV , so by power of a point, it suffices to compute AT · AS. Now, note that:

• By spiral similarity, we should be able to deal with BT : CT

• By Ratio Lemma, we should be able to deal with BS : CS
• By our previous corollary of linearity of power of a point, since B, C have reasonable powers to (AI),
we should be able to compute AT · AS.

Actually going through with these computations turns out to be slightly more annoying than usual, but still
not bad. Here’s the resulting solution:

Proof. Let E, F be the B and C intouch points. By spiral similarity,

BT BF s−b
= = .
CT CE s−c
Now, by ratio lemma,
BS BT BA c(s − b)
= · = .
CS CT CA b(s − c)
Let f (Z) denote the function AZ 0 · AZ, where Z 0 is the second intersection of AZ and (AI). By the corollary,
this is linear in Z. Now,
c(s − b) BS f (S) − f (B) f (S) − (s − a)c
= = = =⇒ f (S)s(b − c) = (s − a)bc(b − c).
b(s − c) CS f (S) − f (C) f (S) − (s − a)b
15
mira74 6 Linearity of Power of a Point

Now, we have
bc(s − a)
f (S) = .
s
We need to check that this is equal to AU · AV , and do this via the law of sines:

sin(B/2)
AU = AB · = AB cot B/2 =⇒ AU · AV = bc cot B/2 cot C/2.
sin(90◦ + B/2)
IF r
To deal with cot B/2, note that it’s BF = s−b , giving

bc bc [ABC]2 bc(s − a)
AU · AV = r2 = · 2
= ,
(s − b)(s − c) (s − b)(s − c) s s

where the last equality is by Heron’s, as desired.

Problem 6.5. Let BP QC be on a line in that order, and pick R on the line such that RB · RC = RP · RQ.
BQ
Prove, in a way different from the ratio lemma handout, that BR BP
CR = CP · CQ .

Problem 6.6 (USAMO 2013). In triangle ABC, points P , Q, R lie on sides BC, CA, AB respectively.
Let ωA , ωB , ωC denote the circumcircles of triangles AQR, BRP , CP Q, respectively. Given the fact that
segment AP intersects ωA , ωB , ωC again at X, Y , Z, respectively, prove that Y X/XZ = BP/P C.
Problem 6.7 (AoPS). Let 4ABC be a triangle with circumcenter O. The perpendicular bisectors of the
segments OA, OB and OC intersect the lines BC, CA and AB at D, E and F , respectively. Prove that
D, E, F are collinear.
Problem 6.8. Let ABC be a triangle, and let ω be the circle of points T with BT : CT = λ. Prove that
2 2
−AB 2
the power from A to ω is AC λ·λ2 −1 .
Problem 6.9 (IMO 2017, Reworded). Let ABCD be a trapezoid with AB||CD, and let M be the midpoint
of AC. Show that AD is tangent to the circumcircle of ABM if and only if BC is tangent to the circumcircle
of CDM .

Problem 6.10 (GGG1). Let ABC be an acute triangle, and let D, E, F be the feet of the altitudes from
←→
A, B, C, respectively. Let EF meet the circumcircle of ABC at points S1 and S2 . Let P be the intersection
of BE and DF , and let Q be the intersection of DE and CF .
Prove that the circumcircles of triangles DS1 S2 and DP Q are tangent to each other.
Problem 6.11 (USA TST 2019). Let ABC be a triangle and let M and N denote the midpoints of AB and
AC, respectively. Let X be a point such that AX is tangent to the circumcircle of triangle ABC. Denote by
ωB the circle through M and B tangent to M X, and by ωC the circle through N and C tangent to N X.
Show that ωB and ωC intersect on line BC.
Problem 6.12 (India TST 2019). Let ABC be an acute triangle with circumcircle Ω and altitudes AD,
BE, CF meeting at H. Let ω be the circumcircle of 4DEF . Point S 6= A lies on Ω such that DS = DA.
Line AD meets EF at Q, and meets ω at L 6= D. Point M is chosen such that DM is a diameter of ω. Point
P lies on EF with DP ⊥ EF . Prove that lines SH, M Q, P L are concurrent.

16
mira74 7 Two Length Conditions

7 Two Length Conditions

You can view a lot of this handout as proving theorems that we “should” have. For example, in the “Ratios
on a Circle” section, we think about points in terms of the ratio they form with distances to B and C. Since
this uniquely determines points on a fixed circle through B, C or line BC, we “should” be able to check when
things are collinear or concyclic, and find ways to do this in that section. This section answers the question:

Say I fixed a point O, and gave you points P1 , P2 , P3 by telling you the distance from O and the
angle they leave O at. Can you tell me if they’re collinear? If they’re concyclic?

Here are the two length conditions:

Theorem 7.1 (Concyclicity Length Condition)

Let O be a fixed point and `1 , `2 , `3 be fixed lines through it. Let θ1 , θ2 , and θ3 be the angles between
`2 and `3 , `3 and `1 , and `1 and `2 . Points P1 , P2 , P3 are chosen on each of these lines. Then, OP1 P2 P3
is a cyclic quadrilateral if and only if

sin(θ1 )OP1 + sin(θ2 )OP2 + sin(θ3 )OP3 = 0

with signs chosen appropriately.

Theorem 7.2 (Collinearity Length Condition)

Let O be a fixed point and `1 , `2 , `3 be fixed lines through it. Let θ1 , θ2 , and θ3 be the angles between
`2 and `3 , `3 and `1 , and `1 and `2 . Points P1 , P2 , P3 are chosen on each of these lines. Then, P1 , P2 ,
and P3 are collinear if and only if

sin(θ1 ) sin(θ2 ) sin(θ3 )

+ + =0
OP1 OP2 OP3
with signs chosen appropriately.

Being able to state the theorem without the “with signs chosen appropriately” thing requires actually learning
how to do trig with directed angles.

There are plenty of proofs of these conditions. I’ll just outline of 1.5 proofs for each:

• Concyclicity: apply Ptolemy’s.

• Collinearity: compute [P1 P2 P3 ].

• The conditions are equivalent: Invert at P .

The proofs are also fully written out in the third appendix.

The philosophy behind applying these conditions is simple - if we can understand the directions and lengths
well enough from the perspective of a certain point, you can check collinearity and concyclicity at that point.
I’ll give an example.

17
mira74 7 Two Length Conditions

Example 7.3 (ELMO 2021)

In 4ABC, points P and Q lie on sides AB and AC, respectively, such that the circumcircle of 4AP Q
is tangent to BC at D. Let E lie on side BC such that BD = EC. Line DP intersects the circumcircle
of 4CDQ again at X, and line DQ intersects the circumcircle of 4BDP again at Y . Prove that D, E,
X, and Y are concyclic.

Observe that (BDP ) and (CDQ) are tangent. Since homothety at D maps these circles to each other, we
have DP BY ∼ DXCQ. Now, we have a pretty good grasp on E, X, Y from the perspective of D, and we
can finish using the concyclicity criterion.

P Q

B D E C

Proof. Assume AB < AC, so DX is contained in ∠Y DE. We want to prove

DE sin(XDY ) + DY sin(XDE) − DX sin(Y DE) = 0.

For the angles, we have ∠XDY = 180 − ∠A, ∠XDE = ∠BDP = ∠BAD, and similar for the last remaining
one. For the lengths,
DC DC BD DC
DX = DP = · · AD = · AD,
DB DB BA BA
where we use BDP ∼ BAD for the second to last step. Hence, we want to show
DB DC
(DC − DB) sin(A) + · AD sin(BAD) − · AD sin(CAD) = 0.
CA BA
To get rid of the annoying sines, note that by ratio lemma,
sin(BAC) DA BC
= · .
sin(DAC) BA DC

Hence, dividing through by sin(A), we want to check

DB 2 DC 2
(DC − DB) + − =0
BC BC
which is clear, as desired.

18
mira74 7 Two Length Conditions

Before going into practice problems, I’ll mention a couple ideas used to make computation with these
conditions easier. There are a couple interesting connections between the concyclicity length condition and
other theorems:

• Corollary 6.3 applied to three collinear points P is pretty similar. For instance, in the above solution,
we have ∠XDE = ∠BAD, ∠EDY = 180◦ − ∠DAC. Therefore, we can translate and rotate D, E, X, Y
so D goes to A, line DX goes to line AB, line DE goes to line AD, and line DY goes to line AC.
Now, by Corollary 6.3, the concyclicity is equivalent to there being a line passing through the points
(0, DX · AB), (BD, DE · AD), and (BC, −DY · AC).
• Say Pi , Qi are chosen on `i for i = 2, 3, and we have that OP1 P2 P3 and OP1 Q2 Q3 are cyclic quadrilaterals.
Then, if we subtract the concyclicity length conditions, we get
P2 Q2 sin(P1 OP2 )
= .
P3 Q3 sin(P1 OP3 )
This can also be proved by using spiral similarity at P1 .

Also, for either length condition, if we have a triangle with angles θ1 , θ2 , and θ3 , we can replace the sines in
the length condition with the side lengths of that triangle, which is often useful.
Problem 7.4 (AoPS). Let ABC be a triangle with orthocenter H. Let X, Y, Z be points on AH, BH, and
CH. Show that [BXC] + [CY A] + [AZB] = [ABC] if and only if HXY Z is cyclic.
Problem 7.5 (APMO 2017). Let ABC be a triangle with AB < AC. Let D be the intersection point of
the internal bisector of angle BAC and the circumcircle of ABC. Let Z be the intersection point of the
perpendicular bisector of AC with the external bisector of angle ∠BAC. Prove that the midpoint of the
segment AB lies on the circumcircle of triangle ADZ.
Problem 7.6 (Sharygin 2017). Let ABC be a triangle with incenter I. Show that the orthocenter of AIB,
the orthocenter of AIC, and the A-intouch point are collinear.
Problem 7.7 (IMO Shortlist 2018). Let ABC be a triangle with AB = AC, and let M be the midpoint of
BC. Let P be a point such that P B < P C and P A is parallel to BC. Let X and Y be points on the lines P B
and P C, respectively, so that B lies on the segment P X, C lies on the segment P Y , and ∠P XM = ∠P Y M .
Prove that the quadrilateral AP XY is cyclic.
Problem 7.8 (IMO Shortlist 2015). Let ABC be an acute triangle and let M be the midpoint of AC. A
circle ω passing through B and M meets the sides AB and BC at points P and Q respectively. Let T be the
point such that BP T Q is a parallelogram. Suppose that T lies on the circumcircle of ABC. Determine all
BT
possible values of BM .
Problem 7.9 (Fake USAJMO 2020). Let 4ABC be a triangle. Points D, E, and F are placed on sides
BC, CA, and AB respectively such that EF k BC. The line DE meets the circumcircle of 4ADC again at
X 6= D. Similarly, the line DF meets the circumcircle of 4ADB again at Y = 6 D. If D1 is the reflection of
D across the midpoint of BC, prove that the four points D, D1 , X, and Y lie on a circle.
Problem 7.10 (IMO Shortlist 2018). Let ABC be a triangle with circumcircle Ω and incentre I. A line `
intersects the lines AI, BI, and CI at points D, E, and F , respectively, distinct from the points A, B, C,
and I. The perpendicular bisectors x, y, and z of the segments AD, BE, and CF , respectively determine a
triangle Θ. Show that the circumcircle of the triangle Θ is tangent to Ω.
Problem 7.11 (ELMO Shortlist 2010). Let ABC be a triangle with circumcircle Ω. X and Y are points on
Ω such that XY meets AB and AC at D and E, respectively. Show that the midpoints of XY , BE, CD,
and DE are concyclic.
Problem 7.12 (USA TSTST 2019). Let ABC be an acute triangle with orthocenter H and circumcircle Γ.
A line through H intersects segments AB and AC at E and F , respectively. Let K be the circumcenter of
4AEF , and suppose line AK intersects Γ again at a point D. Prove that line HK and the line through D
perpendicular to BC meet on Γ.

19
mira74 8 Appendix 1: General Advice and Random Things not in the Handout

8 Appendix 1: General Advice and Random Things not in the

Handout
Here are some things I didn’t get to include, if you’re interested. Luke heavily helped this section.

1. The method of using linearity to compute power at random points was omitted, maybe read Kagebaka’s
handout.
2. As mentioned after the Iran Lemma, if you want an angle (or ratio) to be nice, you can try to force it
to synthetically show up.
3. I don’t use the Forgotten Coaxiality Lemma often enough to write about it, but it’s pretty strong.
4. If P, Q are points and ω is a circle, P is on Q’s polar if and only if P Q2 = Pow(P, ω) + Pow(Q, ω).
RMM 2013 #3, for example, can be solved with this.

5. It’s often useful to apply Menelaus or Cevas to weird triangles. You can often motivate these weird
applications by thinking about the information you have about points. This AoPS post is an example
of a weird Menelaus application that I remember.
6. If you have two linear functions of a point, you basically have a coordinate system. This is already the
idea behind barycentric coordinates, but more random coordinate systems can be useful. Examples:

• EGMO 2022 #1 with “component along AB and AC.” It suffices to check that the ordered pairs
(component along AB, component along AC) of the three desired points lie on a line.
• See the first problem of section 3.3 of Patrik Bak’s Master’s Thesis (GeoGen!). The point is, for
general points D, E on BC, F B = F C is equivalent to some linear relationship between D and E.
• There’s some old ISL problem but I can’t remember it. The interested reader should solve every
ISL problem from the 21st century since I’m pretty sure it was after 2000.
7. You can often think about whether or not points are “algebraically distinguishable.” A common example
is if a circle and a line (or two circles) intersect at two points P and Q in some configuration, one of
the following is happening:
• P and Q are somehow distinguishable points. For example, there could be some special line that
P lies on and Q doesn’t lie on. If you want to deal with these points you’ll generally need to find
these better characterizations, since you need the points to have different properties.
• P and Q are indistinguishable. In this case, there are a few things you may want to do:
– If the the problem is symmetric in P, Q, it’s often helpful to work with symmetric properties
of P, Q. Stuff like properties of line P Q, the midpoint of P Q, powers of points on line P Q, etc.
This is analogous to Vietas problems on computational contests but with more possibilities.
– If the problem isn’t symmetric in P, Q, and another pair of algebraically indistinguishable
points R, S comes up, there may be some way roots of this quadratic correspond to each other.
For a simple example:
Let P be a point and ω be a circle with center O. Lines `1 and `2 through P
are symmetric about P O, and they meet ω at Q, R and S, T , with P Q < P R and
P S < P T . Prove that P QOT is cyclic.
Q, R are, on their own, algebraically indistinguishable, and so are S, T , but we don’t have stuff
like “P QOS is cyclic”. This is because the P Q < P R and P S < P T conditions are actually
there to ensure reflection over P O maps Q to S and R to T .

20
mira74 8 Appendix 1: General Advice and Random Things not in the Handout

8. It’s useful to think about “as some point moves, how do other points move?” Even if there aren’t free
points, this can be helpful in figuring out what you care about. For example, in the second bullet point
of (3), we thought about “how do D, E move if F moves along the perpendicular bisector?”
If you know any moving points (“/degree counting”), intuition from that is probably useful here. Yayups’
moving points tutorial is pretty good for this.

9. I’ll end on this one cuz its the corniest. You shouldn’t let yourself be restricted lengthbashing along
methods in this handout! I think one of the important takeaways from this handout is that trying to
bash things that you feel like you should be able to deal with. The existence of this section implies that
there are a ton of more ad-hoc lengthbashing methods out there to use or something.

I considered adding an additional problems section, but this would distract from the fact that you can
probably go out and do a random geo problem, and be able to use lengthbash with decent probability.

21
mira74 9 Appendix 2: My IMO 2021/3 Thought Process

9 Appendix 2: My IMO 2021/3 Thought Process

I’m including this for a couple reasons:

• It’s a good representative of the techniques and intuition from this handout, but isn’t direct from it.
• You scan all of your scratch work on the IMO since coordination exists, so I have a relatively good idea
of what was going through my head during the contest.
• Some parts of this solution are pretty funny and I want to show off.
Problem 9.1 (IMO 2021 #3). Let D be in acute triangle ABC with AB > AC so that ∠DAB = ∠CAD.
Point E on segment AC satisfies ∠ADE = ∠BCD, point F on segment AB satisfies ∠F DA = ∠DBC, and
point X on the line AC satisfies CX = BX. Let O1 and O2 be the circumcenters of the triangles ADC and
EXD, respectively. Prove that the lines BC, EF, and O1 O2 are concurrent.

The first step of any geometry problem is drawing an accurate, to-scale diagram:

The first natural goal here is to figure out “what” point Y is, in terms of (triangle ABC and) D. Because of
AE AF
Menelaus, it suffices to compute CE and BF .

The first observation I made was that DE is tangent to (CDT ), and similar for DF , since angle conditions
BD 2
are scary. Now, linearity of PoP on Pow(P, (DCT )) − P D2 gives CE
AE
, and we can compute BY


CY = CD . I
wrote this down in my scratch paper as “Y D is tangent to (BDC)”.
Remark: The linearity of PoP here is kinda silly, ratio lemma just works.
Now, since there are lots of circles through D, inversion at D looks nice. I tried that for a bit and gave up
once I realized X was impossible to deal with after inversion. I also introduced D0 = (DEX) ∩ (DAC) at
some point in this process

22
mira74 9 Appendix 2: My IMO 2021/3 Thought Process

After giving up, I observed that the original problem, the collinearity is equivalent to the circle centered at
Y passing through D also passing through D0 , which is just the D-Apollonius circle. Since we know where
(DEX) and (DAC) meet line AC, I decided to reduce to the following problem:

BD BP BQ Q AP
Given P, Q on AC with CD = CP = CQ , compute CP .

For ratio lemma reasons (or the forgotten coaxiality lemma), it suffices to check that what we get is equal to
AX AE
CX · CE . This also gets rid of pretty much all the weird points in the problem.

AP/CP and AQ/CQ should be roots of some quadratic. Stewart’s theorem is really the only theorem relating
BP to the location of P on AC, so that should be enough to literally be able to find the quadratic. Indeed,
we can make Stewart’s homogenous in AP, BP, CP as follows:

AP
where λ = CP and the last step is Vieta’s.
Remark: Exercise 6.8 gives a better way of doing this step, the intended solution is around 1 line.
AX
To compute CX , do linearity of CP 2 − BP 2 . We computed AE
CE earlier. After actually writing out the
computations, we want to check:

BD 2 2
AX AE AB 2 − AC 2 AD · DT AP AQ CD 2 AC − AB 2
· = · equals · = .
CX CE BC 2 CD2 CP CQ BC 2
Unfortunately, this isn’t trivial. But its relatively small so it can’t be that bad. After cross multiplying, both
sides are quadratics in D, so you can just check three cases and win. I ended up checking D = A, T and the
arc midpoint, all of which work! GG. 

Remark: The last step also follows from exercise 6.8 on D and the A-Apollonius circle.

23
mira74 10 Appendix 3: Actually Writing Out Proofs

10 Appendix 3: Actually Writing Out Proofs

Ok now I’ll finally write out the full proofs of the theorems I skipped over earlier.

Theorem 10.1 (Ratio Lemma)

Let 4ABC be a triangle, and let D be a point on BC. Then, we have
BD BA sin ∠BAD
= · .
CD CA sin ∠CAD

B D C

Proof. By the law of sines on ABD and ACD, we have

BA sin(BAD) CA sin(CAD)
= and = .
BD sin(ADB) CD sin(ADC)

Since ∠ADB + ∠ADC = 180◦ , we have

BD CD
sin(BAD) · = sin(ADB) = sin(ADC) = sin(CAD)
BA CA
which rearranges to
BD BA sin ∠BAD
= · ,
CD CA sin ∠CAD
as desired.

Theorem 10.2 (Quadrilateral Ratio Formula)

Let ABC be a triangle, and let D be a point. Then, we have

sin(BAD) sin(ABD) BD sin(ABD) sin(BCD)

= · = · .
sin(CAD) sin(ACD) CD sin(ACD) sin(CBD)

24
mira74 10 Appendix 3: Actually Writing Out Proofs

B C

Proof. By the law of sines on ABD and ACD, we have

sin(BAD) sin(CAD)
AB · = AD = AC · .
sin(ABD) sin(ACD)

This gives the first equality. The second equality follows from the law of sines of BCD.

Theorem 10.3 (Concyclicity Length Condition)

Let O be a fixed point and `1 , `2 , `3 be fixed lines through it. Let θ1 , θ2 , and θ3 be the angles between
`2 and `3 , `3 and `1 , and `1 and `2 . Points P1 , P2 , P3 are chosen on each of these lines. Then, OP1 P2 P3
is a cyclic quadrilateral if and only if

sin(θ1 )OP1 + sin(θ2 )OP2 + sin(θ3 )OP3 = 0

with signs chosen appropriately.

Proof. We show that, assuming P2 lies within ∠P3 OP1 , we have that OP1 P2 P3 is a cyclic quadrilateral if
and only if
sin(θ1 )OP1 + sin(θ3 )OP3 = sin(θ2 )OP2 .
Let P20 be the point where OP2 meets the circumcircle of OP1 P3 . Letting R be the circumradius of OP1 P3 ,
we have, by Ptolemy’s theorem,

P20 P3 P 0 P1 P1 P3
OP1 · + OP3 · 2 = OP2 · .
2R 2R 2R

25
mira74 10 Appendix 3: Actually Writing Out Proofs

By the extended law of sines, this is

sin(θ1 )OP1 + sin(θ3 )OP3 = sin(θ2 )OP20 .

Now, both the length condition from earlier and P2 being on (OP1 P3 ) are equivalent to OP20 = OP2 , as
desired.

Theorem 10.4 (Collinearity Length Condition)

Let O be a fixed point and `1 , `2 , `3 be fixed lines through it. Let θ1 , θ2 , and θ3 be the angles between
`2 and `3 , `3 and `1 , and `1 and `2 . Points P1 , P2 , P3 are chosen on each of these lines. Then, P1 , P2 ,
and P3 are collinear if and only if

sin(θ1 ) sin(θ2 ) sin(θ3 )

+ + =0
OP1 OP2 OP3
with signs chosen appropriately.

Proof. We show that, assuming P2 lies within ∠P3 OP1 , we have that P1 , P2 , and P3 are collinear if and only
if
sin(θ1 ) sin(θ3 ) sin(θ2 )
+ = .
OP1 OP3 OP2
Upon multiplying both sides by 12 OP1 · OP2 · OP3 , we have that it is equivalent to

1 1 1
OP2 · OP3 sin(θ1 ) + OP2 · OP1 sin(θ3 ) = OP1 OP3 sin(θ2 ).
2 2 2
By the 12 ab sin C area formula, this is equivalent to [OP2 P3 ] + [OP2 P1 ] = [OP1 P3 ], which is equivalent to
[P1 P2 P3 ] having area 0, as desired.

26