Coins on a grid

A coin is randomly thrown on a square grid. As the size of the grid approaches infinity, what is the probability that the coin does not touch the gridlines?

Assume that the thickness of the grid lines is negligible.

2 3 \frac23 1 4 \frac14 1 2 \frac12 1 3 \frac13

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

8 solutions

David Vreken
Sep 29, 2018

Since the coin has a radius of 1 2 cm \frac{1}{2} \text{ cm} , if the center of the coin is within 1 2 cm \frac{1}{2} \text{ cm} from a grid line, then it will cross it, otherwise it will not cross it. The red area below shows where the center of the coin can land and cross a grid line, and the green area below shows where the center of the coin can land and not cross a grid line.

Therefore, the probability that the coin does not touch the grid lines is the area of the green square to each grid square, which is 1 1 2 2 = 1 4 \frac{1 \cdot 1}{2 \cdot 2} = \boxed{\frac{1}{4}} .

How does considering the probability that the coin doesn't touch the grid lines w.r.t EACH GRID SQUARE extend to the ENTIRE INFINITE GRID SYSTEM ?

Aniruddha Bagchi - 2 years, 8 months ago

Log in to reply

The same pattern is repeated in the entire grid system, so the ratio of the areas in the entire grid system is the same as the ratio of the areas in each grid square.

David Vreken - 2 years, 8 months ago

Log in to reply

The infinity of the grid is really neither here nor there (except that by NOT having a boundary, we don't need to concern ourselves with the possibility - or probability - of the coin landing outside of that boundary) . The probability for a single square is equal to probability for 10 or 1000 or 100,000 or ANY number of squares.

John Conway - 2 years, 8 months ago

Log in to reply

@John Conway "The infinity of the grid is really neither here nor there." Actually, the infinity of the grid is important, because there is no translation invariant probability measure on the infinite plane . As I mention in my solution, and in a report which I don't know if you can see, because it is still marked "pending", there are ways to make sense of the problem that get around this. However, you should not dismiss the very valid concerns of anyone who does not see how to make sense of the problem because of the infinite nature of the grid.

Varsha Dani - 2 years, 8 months ago

@Aniruddha Bagchi You ask a reasonable question. Please see the comment I posted after my solution.

Varsha Dani - 2 years, 8 months ago

Why is the center placement uniformly distributed in the plane? The problem did not state this...

C . - 2 years, 8 months ago

Log in to reply

How else would you distribute the center placement? The coin is thrown randomly, and the grid is infinite.

In real life it makes sense that the coin will not end up a kilometer (or even more) away, but if I were to throw a coin so that it lands randomly with respect to the grid, I probably would throw it at least a couple of meters away. With 2 cm per cell, that means I'm probably throwing it somewhere between 100 and 200 cells away, and the position within the cell that contains its center will likely be very random indeed, as we're talking about tens of thousands of cells it could have landed in.

So in such a practical case, the answer I would expect will be very close to the mathematical answer for a uniform distribution over this infinite plane, i.e. 25% to not cross any of the grid lines.

Roland van Vliembergen - 2 years, 8 months ago

Log in to reply

On the contrary, this is a valid complaint. You say that you "expect will be very close to the mathematical answer for a uniform distribution over this infinite plane". But in fact there is no such thing as a uniform probability distribution over an infinite plane. That there are ways to interpret this problem so that it makes sense does not alter that fact.

Varsha Dani - 2 years, 8 months ago

I might have a confession to make. Bertrand's paradox about "choosing a random chord" has biased me, and now i don't trust any multi-dimensional probability that isn't clearly specified. ^^

C . - 2 years, 8 months ago

Consider a 2cm coin, which intersects the line 100 percent. By shortening its radius to 1cm, it will be about a quarter of its original size, so its probability is about a quarter. Is that ok?

zhihao luo - 2 years, 8 months ago

Log in to reply

Although you wrote 'radius' I assume you mean 'diameter'.

Your argument is flawed.

You are correct that a 2cm coin always intersects a grid line. However 1/4 is the probability that the 1cm coin misses the grid, not that it hits the grid.

Varsha Dani - 2 years, 8 months ago

Log in to reply

The easiest way to show that the argument is flawed is using a different scale: suppose the diameter was shortened to 0.2 cm (10% of the original). The size is now a hundredth of the coin's initial size, but the probabilities are 64% for miss and 36% for hit. ;-)

C . - 2 years, 8 months ago

oh,I see.You are right.

zhihao luo - 2 years, 8 months ago

Sorry, but how do you arrive at 1.1/2.2 (which is a half, not a quarter)? Or is it 1/2 x 1/2 ? Although I think it is most likely (1 x 1)(/2 x 2) . Anyway, I divided one square up into 16 small squares, the green area occupying 4 of these small squares. Then I said the probability of the coin landing in the green space = Area of Green/Area of Red = 4/16 = 1/4. Am I correct in saying so ? Or am I over-simplifying it ? :-)

John Conway - 2 years, 8 months ago

Log in to reply

Just chill bro

Arnav Gupta - 2 years, 8 months ago

Yez, I mean 1/2 x 1/2 = 1/4. Dividing one square into 16 squares is essentially the same approach.

David Vreken - 2 years, 8 months ago

John, the fraction was G r e e n S i d e × G r e e n S i d e R e d S i d e × R e d S i d e = 1 × 1 2 × 2 \frac{GreenSide\times GreenSide}{RedSide\times RedSide}=\frac{1\times1}{2\times2} ... Meaning it was indeed (1 x 1)/(2 x 2). The fact is that x y x\cdot y is another notation for x × y x\times y , and in fact i have even seen multiplication written as x . y x . y online, probably because a period is easier to type than an interpunct and less ambiguous than using an "x".

C . - 2 years, 8 months ago

Can we say coin can land in 4 ways in 1 box, in 2 box, in 3 box, in 4 box. By geometry we can see all 4 are possible. So, only in case of 1 box it won't touch or Cross any grid line and in rest cases it would. So probability is 1/4

maths jeet - 2 years, 8 months ago

Log in to reply

You don't use the size of coin , and it is relevant

Jesus Nievas - 2 years, 8 months ago

Well... maybe, but it depends on what exactly you mean by "in 1 box, in 2 box, in 3 box, in 4 box"... I don't see any numbered boxes anywhere... ;-??

C . - 2 years, 8 months ago

By 1box I mean any box in grid, by 2 box I mean coin lands simultaneously in any 2 adjacent boxes, any 3 boxes forming L shape is what I mean by 3 boxes, by 4 boxes I mean 4 boxes forming a bigger square of side 4cm.so when the coin lands on more than 1 box it has to touch the grid line so only case of not touching left is of 1 box so probability should be 1/4

maths jeet - 2 years, 8 months ago

Log in to reply

Oh, that's what you meant, 1 box vs 2 boxes vs ... Sorry, but it doesn't work, because in general the 4 cases are not equiprobable.

For the current case, there's 25%, 50%, 5.4% and 19.6% ( 25 π 4 \cfrac{25\cdot\pi}{4} ).

And if you consider the case of a coin of radius equal to 10% the size of the grid, the values become 64%, 32%, 0.86% and 3.14%. :-B

C . - 2 years, 8 months ago

I think that this solution is correct if a sphere is launched. If a coin is tossed, it can spin, and it is more likely not to touch the net.

Jesus Nievas - 2 years, 8 months ago

Log in to reply

I agree. The answer 1/4 is true only if you assume the plane of the falling coin is parallel to the grid in the moment it goes through it, and that any possible horizontal speed of the coin is equal to zero.

Carlos Reche - 2 years, 8 months ago

This solution is right. However, it can be somewhat simplified.

Instead of tracing center of the coin, we can take a bounding rectangle and trace top left (or any other) corner.

Inside a single cell, top left coin rectangle corner can be set in any point in a sub-figure of area 3 and intersect grid lines. If we place the coin corner anywhere in top left quarter of the cell, it will not intersect any grid lines.

The answer is the same, the ratio of the relevant areas is still 1/4.

Arkadi Kagan - 2 years, 8 months ago
Chez Townsend
Oct 2, 2018

Instead of considering areas, consider the position of the center of the coin, first in the x-direction and then in the y-direction. Since the coin is randomly thrown, these are independent of each other. We can simplify to consider just one square.

X-direction : The center of the coin has to land between x = 0.5cm and x = 1.5cm to ensure the coin does not touch a vertical line. This is ( 1.5 0.5 ) 2 \frac{(1.5-0.5)}{2} , i.e. 1 2 \frac{1}{2} .

Y-direction : Analogous to the x-direction, so the probability is the same, i.e. 1 2 \frac{1}{2} .

Overall : Multiple these together to get the overall probability that the coin does not touch a vertical or horizontal grid line: 1 2 \frac{1}{2} x 1 2 \frac{1}{2} = 1 4 \frac{1}{4} .

This is the way I did it - makes good algebraic sense - Steve Tomlinson

Stephen Tomlinson - 2 years, 8 months ago

Since the coin is randomly thrown, these are independent of each other.

To me, that statement is not immediately obvious. For example, if the axis were angled at 6 0 60 ^ \circ , are these still independent of each other? Why, or why not?

Calvin Lin Staff - 2 years, 8 months ago

Log in to reply

Hi Calvin,

Yes they will. If you redefine the measure on the plane via a different lattice it will still be the product measure. Naturally the actual probability will depend on the size of the parallelogram. In particular, if the distance between the parallel lines in your new lattice is still 2, then the probability will still be 1/4. If the distances are different, that will affect the probabilities of the two independent events, and therefore the probability of the event under consideration.

Of course, all of this has to be seen in the right interpretation. Here, as in the original problem, the product "measure" we are talking about is the Lebesgue measure. In order to get a probability measure from this, we need to mod out by the lattice. Without that, the problem is meaningless, because (as I have pointed out repeatedly in my solution and in one of the other threads) there is no uniform probability measure on the infinite plane, and no non-uniform measure has been specified.

Varsha Dani - 2 years, 8 months ago

Log in to reply

With regards to your first paragraph, it depends on what "size" you are measuring. You seem to be hinting at it by saying "the distance between the parallel lines". If we're simply looking at 2 × 2 2 \times 2 parallelograms (by side lengths), then the statement isn't true.

With regards to your second paragraph, yes that is a common misconception. The way to "fix" it is to take the limiting probability on an ever larger grid, which is what people often mean.

Calvin Lin Staff - 2 years, 8 months ago

Log in to reply

@Calvin Lin " If we're simply looking at 2x2 parallelograms (by side lengths), then the statement isn't true."

Which statement do you mean? The statement that the probability is 1/4 is not true. However

  1. The events that the x-coordinate lies between 0.5 and 1.5 and the y coordinate is between 0.5 and 1.5 are independent. They simply do not correspond to the event that the coin misses the grid lines.

  2. The events that (the y-coordinate lies in an interval such that the coin misses the horizontal lines) and that (the projection of the center of the circle in the direction of the slanting lines lies in an interval that causes the coin to miss the slanting lines ) are also independent

For (2) the point is that the coordinates of the center are independent. But if X and Y are independent random variables, and Z= aX+bY (mod the lattice) where b is not 0, then X and Z are also independent.


"The way to "fix" it is to take the limiting probability on an ever larger grid, which is what people often mean."

Sure, as long as you take the limit over "nice" regions, such as rectangles. Perhaps it is suffices for them to be convex. But in general, it is possible to write down a sequence of bounded regions such that taking the limit on the restriction of the grid to those regions will not yield 1/4. That's why I prefer the quotient approach, where you consider the problem mod the lattice. However, I do realize that if I wasn't already quibbling before, I am definitely quibbling now. :)

Varsha Dani - 2 years, 8 months ago

Log in to reply

@Varsha Dani I'm clarifying that we're talking about 2 different scenarios, and (I believe) we are in agreement about what happens in each of these scenarios.

I am referring to a 2 × 2 2 \times 2 parallelogram (by side length) with a 1 1 ^\circ vertex angle. In this case, then the area of the parallelogram is 2 × 2 × sin 1 < π 4 2 \times 2 \times \sin 1^ \circ < \frac{\pi}{4} and clearly we can't fit a circle in it.

You are referring to "perpendicular distance between the parallel lines is still 2 cm", which means we have a 2 × 2 sin 1 c i r c 2 \times \frac{2}{\sin 1^ circ } parallelogram (by side length).


The approach is to use similar figures (whether it's circles, squares, triangles, or pretty much any measurable region). The reason for doing so is that there are times when the measure isn't evenly distributed, and so the "mod the lattice" approach would not work.

E.g When they say that the optimal packing of circles in the plane is 90.69 % \approx 90.69\% , it doesn't just apply to lattice packings (which is the "easy" case), but also to irregular packings (which is much harder).

Calvin Lin Staff - 2 years, 8 months ago

Log in to reply

@Calvin Lin Hi again,

To your first point:

I think you are misunderstanding what I am saying. I am not talking only about latices where the distance between the parallel lines is 2. I absolutely agree that you cannot fit a circle in the parallelogram if the angle is 1 degree.

Your original question (unless I am mistaken) was whether the events would be independent if the grid lines were slanted. What I am trying to point out is that the coordinates of the center of the coin are still independent, so even with your grid lines at a 1 degree angle, you CAN still compute the probability (which is zero) by writing down two independent events whose conjunction is the desired event. (Naturally one of them will have probability zero.)

As to your second point, the point I was trying to make that it is NOT sufficient to use "pretty much any measurable region". Circle, squares, other convex regions are nice. Arbitrary measurable regions can be pretty nasty. And you don't even have to look very nasty regions in order to break the limit. For instance if you consider the j-th region in your sequence to be the intersection of the square of side length 200 j centered at the origin with the area between the hyperbolas x y = 2 j xy=2j and x y = 2 j xy= -2j , then in the limit these regions will cover the plane, but if you solve the coin problem on successive regions of this type and take the limit, you will not get 1/4 because the regions will include a large contribution from the thin strip around the axes.

Varsha Dani - 2 years, 8 months ago
Varsha Dani
Oct 1, 2018

Consider the grid as the set of lines { x = 2 n n Z } { y = 2 n n Z } \{x=2n | n \in \mathbb{Z}\} \cup \{y =2n | n \in \mathbb{Z}\} . Let ( a , b ) (a,b) be coordinates of the center of the coin. Then there are indepedent identical random variables. The coin does not intersect a vertical line if a [ 1 / 2 , 3 / 2 ] m o d 2 a \in [1/2, 3/2] \mod 2 and similarly it misses a horizontal line if b [ 1 / 2 , 3 / 2 ] m o d 2 b\in [1/2, 3/2] \mod 2 . Thus it misses the horizontal lines with probability 1/2 and the vertical lines also with probability 1/2. By independence, the probability of missing the grid is 1/4.

HOWEVER

I would like to point out that technically , the problem is not well-defined because there is no translation invariant probability measure on R 2 \mathbb{R}^2 , or for that matter on R \mathbb{R} . One way to get around this is to view the plane as a torus by identifying the grid lines. This is implicitly what I am doing in my solution by reducing the coordinates mod 2. Another way is to think of the coin as being fixed at the origin, and asking about a random placement of the grid lines. Since the space of placements of the grid lines is compact, it does have a uniform probability measure.

This last paragraph is not correct. Lebesgue (uniform) is obviously so. But there are others which may be constructed. https://math.stackexchange.com/questions/147089/translation-invariant-measures-on-mathbb-r

Joe Plotkin - 2 years, 8 months ago

Log in to reply

Lebesgue measure in not a probability measure. A probability measure has measure 1 for the whole space , while the Lebesgue measure of the plane is infinity. Also, according to the link you posted, up to a constant factors Lebesgue measure is the only translation invariant measure on the plane. This means all such measures are infinite on the whole space, which in fact proves my claim that there is no such probability measure.

Varsha Dani - 2 years, 8 months ago
Parth Sankhe
Sep 25, 2018

The total area the coin can cover with respect to one grid is 12 + π ( This includes area of the square + area the coin can cover while still touching the square from the outside.), while the area required will be such that the coin is completely inside the square, which is 3 + π/4. Hence, the probability is 1/4

Same thing I did.

John Smith - 2 years, 8 months ago

Not sure what you're talking about... Do you think you can maybe sketch a diagram of the two cases?

C . - 2 years, 8 months ago
Dhruva Sakhare
Oct 4, 2018

The area of one square of the grid is 2 2 = 4 c m 2 2*2=4 cm^2 and the area of the square which is surrounding the circle is 1 1 = 1 c m 2 1*1=1 cm^2 . the ratio of their areas is 1 : 4 1:4 .

Ben Krekorian
Oct 7, 2018
  • We know the area inside a square, in this grid, is 2 cm squared (= a 4 cm area.)

  • And we know the diameter and area of a single coin (circle) is 1 cm squared.

  • So according to though measurements, we can only fit 4 coins/circle in one square, without touching the lines.

-{ Ego, a single coin is 1/4 of the area inside a square. | In other words, there is 1/4 chance for the coin to stay inside the square and ego not to touch the grid lines. }-

Ervyn Manuyag
Oct 2, 2018

1/2*1/2=1/4

Krishna Deshmukh
Oct 1, 2018

First consider a square in a grid.The total area where the coin of radius 1/2 cm can move so that it does not actually touch the grid lines is pi(1)^2(which is the area of the inscribed circle in square). The actual area of coin is pi(1/2)^2. Probability=area of coin /area for free movement I.e inscribed circle =1/4 No matter there be any number of grids,the combined probability is same as considering only one square in this case.

Place a 1x1 square in the centre of the 2x2 square and turn it 45 deg. If the center of the coin lies inside the rotated square the circumference of the coin will not touch the edge. P = 1/4.

Ian Fowler - 2 years, 8 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...