A perfectly symmetrical small 4-legged table is standing in a large room with a continuous but uneven floor. Is it always possible to position the table in such a way that it doesn't wobble, i.e. all four legs are touching the floor?

No tricks. No . Serious question (with real-life applications too!) with a serious answer.

This might look like it'd fit better on Lifehacks.SE, but the answer has a nice mathematical proof/formula [depending on whether it's yes/no; I won't give the game away!] which is surprisingly simple and elegant.

  • 27
    $\begingroup$ What if the continuous uneven floor resembles a bed of nails where each nail is longer than a table leg and they are spaced such that there is no combination of four nails that the legs can touch at the same time, nor is there a space between nails such that all four legs could touch the ground at the bottom? $\endgroup$ Commented Aug 26, 2015 at 19:54
  • 6
    $\begingroup$ @IanMacDonald Come on, seriously! I said no tricks :-) $\endgroup$ Commented Aug 26, 2015 at 20:05
  • 6
    $\begingroup$ What if the floor of the entire room is a slightly uneven, extremely steep slope? $\endgroup$
    – Luke
    Commented Aug 26, 2015 at 20:06
  • 10
    $\begingroup$ Have we decided that "uneven floor" means "horizontal plane, but with flaws -- peaks and valleys -- which are considerably shorter than the table's height"? It seems a lot of people are being reprimanded for "lateral thinking", though the question doesn't comment on the floor's unevenness. $\endgroup$
    – Roland
    Commented Aug 27, 2015 at 15:18
  • 16
    $\begingroup$ Seems to me that the question is trying to have it both ways, as in mathematical model vs practice. On one hand it seems that the intended definition of a “table that doesn't wobble” is very mathematical (ends of legs are points and it suffices that the table can be rotated in a way such that all four touch the floor at some instant), but on the other hand people suggesting “unrealistic” floors get accused of trickery… If the point is to showcase a single pre-determined (albeit neat) solution, perhaps the question should define the desired model of table, floor, and wobble more accurately. $\endgroup$
    – Arkku
    Commented Aug 27, 2015 at 16:05

12 Answers 12


The answer is


Here's why:

Imagine that the table can pass through the floor. We're going to call one leg the "floating leg" - the other three are going to always be on the floor. Now, after we rotate the table a quarter turn, the floating leg is going to be above the floor if it was originally below, or vice versa. By the intermediate value theorem, it will be exactly on the floor at one point in that rotation.

More detailed proof:

Three legs of the table define a plane. Define the "offset" of a leg to be the distance above the floor if the other three legs are placed on the floor directly - negative if below, positive if above. In any arbitrary placement, the offset of any two adjacent legs will be positive and negative. (WLOG assume the one on the left is positive.) Offset is a continuous function because it's the distance from the floor. If you start with a positive offset on leg A and rotate the table 90° to the right, the offset on leg A (now in the original position of another leg) will be negative. This means that at some point, offset was 0, therefore the table was completely touching the floor.

  • 23
    $\begingroup$ What I really like about this solution is that it's genuinely useful. In a real-life situation, you actually can plonk 3 legs on the ground and rotate until it's stable. Normally there's far more intermediate stages of abstraction between a mathematical proof and a real-life method! $\endgroup$ Commented Aug 26, 2015 at 17:19
  • 13
    $\begingroup$ My only concern is ridiculous slants where the cent of mass cannot be between the points of contact. $\endgroup$ Commented Aug 26, 2015 at 17:23
  • 13
    $\begingroup$ @Deusovi "Yes" may very well be the answer, but I am not convinced that your explanation using the intermediate value theorem is sufficient. I can see how it would work if the table has only 1 leg, but with more than 1 leg, it is possible that at no point during the rotation that all legs are exactly on the floor. Please elaborate a little further if you believe your answer is correct. $\endgroup$
    – Aura
    Commented Aug 26, 2015 at 18:13
  • 11
    $\begingroup$ Watch the numberphile youtube on that same question. $\endgroup$
    – user15996
    Commented Aug 26, 2015 at 18:32
  • 8
    $\begingroup$ @Taemyr Why can we make that assumption? The question only says "continuous", which could be all sorts of crazy things. And how do you even define "roughly planar"? Without a good definition the answer isn't much better than "we can always balance the table because I only allow floors where it's always possible to balance the table". $\endgroup$ Commented Aug 27, 2015 at 9:23


The reason is actually simpler and more intuitive than the other answers:

It's obvious that the table can always stand on three legs, because the ends of two legs define a line, and the third leg can be brought into contact by rotating around that line. Then, rotating the table around a vertical axis, the fourth leg must touch the floor before any of the other legs becomes unable to touch, because the fourth leg touching is the reason one of the other legs wouldn't be able to touch.

  • 3
    $\begingroup$ +1 because you're right, unlike several of the other answers. Although @MarchHo is right that this is equivalent to the top answer and describes the same method for stabilising the table, it's also more intuitive (if less rigorous) since it doesn't mention the intermediate value theorem (which is admittedly 'obvious'). $\endgroup$ Commented Aug 27, 2015 at 7:48
  • 8
    $\begingroup$ This gets my +1 for being the clearest explanation, despite not being a new theory. $\endgroup$
    – AndyT
    Commented Aug 27, 2015 at 9:50
  • 5
    $\begingroup$ Much clearer than the accepted answer. $\endgroup$
    – njzk2
    Commented Aug 27, 2015 at 14:11
  • 7
    $\begingroup$ except how do you demonstrate that the fourth foot must touch the floor? can't it stay woobling forever? (for example if the table is standing on an elevation of the floor where you can only set 3 feet at a time?) $\endgroup$
    – njzk2
    Commented Aug 27, 2015 at 14:14
  • 3
    $\begingroup$ @arivero: The other answer says to rotate the table while keeping the 3 legs on the floor but uses different wording: We're going to call one leg the "floating leg" - the other three are going to always be on the floor. The other answer is more specific, it states that the solution WILL BE FOUND within a quarter turn. $\endgroup$
    – slebetman
    Commented Aug 28, 2015 at 8:49

The answer is:


Reason by provable test case:

We have a table that is a solid, 3 foot tall, wide, and long, cube of wood, with 4 very short stubby legs on one side (we'll say 2 inches). If standing on a continuous but uneven floor, Say, a simple hexagonal pattern of half-spherical domes (5 inch high domes) (stylized like bubble paper, of course) the domes would lift the table off the floor, preventing it from balancing. At no point could you guarantee stable footing.

...Here's some images to help illustrate the problem.

A grid of bubbles

Here's a grid of bubbles covering the floor. we can surmise that a table could put its legs between these bubbles and touch the floor, but only on the condition that the floor gives space to do so. if not, then you could "balance" the table only in name, on top of the bubbles, but we all know it wouldn't last.

A hex grid of bubbles

When we switch to a hex-based formation, we see more problems. Firstly, depending on the dimensions of the table, it might not be possible to fit the legs inbetween the bubbles anymore. (In real life, tables don't have massless, frictionful legs because that would be pretty frickin dangerous. This question has what I like to call a spherical cow error [ https://en.wikipedia.org/wiki/Spherical_cow ] - Although it might be useful in real life scenarios, in order to say that it ALWAYS will be helpful in real life scenarios you have to keep adding criteria to the problem to eliminate "edge cases" which are valid concerns for other people, you just don't realize that they count just as much as you do)

A hex grid of bubbles from above

From above, the situation is a lot more clear. Square tables are going to have a hell of a time trying to balance on this kind of floor, and most often, they're just going to remain in a "good enough" position where they still slide a bit and you just have to remember not to do anything on that table that requires stability.

Final verdict:

The table can't be guaranteed to balance on the floor in any situation where the table legs have mass (and therefore volume).

  • 2
    $\begingroup$ Ah, c'mon - I said no tricks or lateral thinking! :-) $\endgroup$ Commented Aug 26, 2015 at 21:36
  • 20
    $\begingroup$ Stubby legs doesn't count as tricks! Don't discriminate against short tables! $\endgroup$
    – Kingrames
    Commented Aug 26, 2015 at 22:00
  • 10
    $\begingroup$ Why stubby legs? You just need a floor that's more uneven than the leg length. $\endgroup$ Commented Aug 27, 2015 at 3:20
  • 3
    $\begingroup$ @PeterOlson The question was meant to be about a real-life problem. Mathematically you can come up with pathological counterexamples to loads of things, but in real life anything that works practically for all cases you're likely to come across is good enough. $\endgroup$ Commented Aug 27, 2015 at 7:58
  • 13
    $\begingroup$ Well fundamentally, this being a "puzzle", a trick answer could be defined as "any correct answer other than the one I had in mind". But for obvious reasons that's not entirely satisfying to people who approach it as a practical problem, nor to people who approach it with a mathematical model different from the one you hoped for :-) $\endgroup$ Commented Aug 27, 2015 at 14:11

The answer is


The answers that say 'yes' seem to stop after showing how you can make all 4 legs reach the floor. However, not only is it sometimes impossible to make all 4 legs touch the floor (think floors with large troughs, mounds, or waves), even when they do, that condition alone does not mean it can be said the table is "balanced", "stable", or "doesn't wobble". Indeed, those answers accept out of hand that 3 legs of the table can touch the floor nicely, but no guarantee for such a condition is afforded anywhere in the problem statement.

Even a modestly turbulent surface that you or I wouldn't have trouble walking on won't hold a table steady unless most of the surface of each foot contacts the floor. This would be impossible if the floor was relentlessly bumpy, like a gravel road or a rocky coastline. If you want a more sterile example, consider something like z = sin x + cos y. There are other, gravel-less ways to get this too. One is where the floor is large tiles that are flat but have a variety of slants. More real-life places come to mind. Speaking of real life, the table will not experience infinite friction with the ground (or with objects placed on it), which means a limit needs to be respected on how far off level the table can be before it simply falls over (or objects slide off).

In conclusion,

there are many factors that determine how and whether or not a table wobbles on a surface. There are probably more than I realize, since I am not a physicist. As for this question, I do not think a large room with a continuous but uneven floor comes close to providing context sufficient to rule out the tendency to wobble.

  • 1
    $\begingroup$ As I said in a comment on the question, the legs are technically being modelled as lines, or at least the bottom of each leg is modelled as a point. $\endgroup$ Commented Aug 27, 2015 at 9:35
  • 8
    $\begingroup$ @randal'thor I don't understand your standard for being about a "real-life" problem. You reject Kingrame's answer and Ian McDonald's comment on this basis even though they present surfaces that are perfectly feasible to produce in real life, while you reject this answer for pointing out real-life limitations because your question is about a mathematical model. What are you really asking about, real life or a mathematical model? $\endgroup$ Commented Aug 27, 2015 at 9:40
  • $\begingroup$ You shouldn't think of table legs as lines if you want this problem to transfer to the real world. $\endgroup$
    – Geoff
    Commented Aug 27, 2015 at 9:42
  • $\begingroup$ The table legs could still be cone-shaped though, i.e. the bottom of each one is a single point. That's enough to invalidate this answer, I think. In reality the leg bases are usually smaller than the bumps in the ground. $\endgroup$ Commented Aug 27, 2015 at 9:51
  • 4
    $\begingroup$ @randal'thor At least 99% of the tables I've seen do not have cone-shaped legs and could easily made wobbly by being placed on top of gravel, and at least 99% of the floors I've seen don't look like a bed of nails. Why do you only seem to consider edge cases that point towards the answer being "yes"? $\endgroup$ Commented Aug 27, 2015 at 9:56

The answer is "YES"

Place the table on the floor, and label the legs, clockwise from above (A, B, C, D) and their points of contact on the floor (1, 2, 3, and 4) respectively. Note that despite the different labels, the four legs, by symmetry, must be identical in length.

If all four legs are in contact, then the condition is satisfied.

If not, then without loss of generality we can set Leg A as not in contact with Floor Point #1.

Then rotate the table by one-quarter turn clockwise, while allowing it to be supported by the floor throughout the rotation.

Now Leg A is at Point #2. and in contact with Point #2. After all, Leg B was in contact before the rotation, and Leg A is identical to Leg B.

On the other hand, Leg D, formerly in contact with Point 4, has rotated around to Point 1, where it is out of contact, since it is replacing the identical Leg A, which was out of contact.

So Leg A made contact, while Leg D lost contact. If Leg A made contact first, then at that point all four legs are in contact. If Leg D left first, then the table is resting on two legs, and it tips over...

Modesty and honesty require that I credit Martin Gardner http://www.scientificamerican.com/article/strange-but-true-turning/

And yes, I did read and recall the original 1973 article...

  • 1
    $\begingroup$ I think this is somewhat better than Deusovi's answer, but it still assumes the existence of some continuous motion from the original position to the rotated position that keeps at least three legs on the ground at all times. It's not clear that you wouldn't run into a situation where leg D loses contact, the table tips over, and it lands on leg A. $\endgroup$ Commented Aug 27, 2015 at 3:00
  • 1
    $\begingroup$ I actually use this technique fairly regularly since I was first shown it as an undergrad in the faculty of Math at Waterloo in the late 70s. Much better than sticking paper under one leg and really does work whenever the issue is the floor rather than the table. $\endgroup$ Commented Aug 27, 2015 at 11:07
  • 1
    $\begingroup$ @slebetman: How do you know? Even for quite ordinary floors, how do you know that such a thing won't happen, aside from having never seen it happen? $\endgroup$ Commented Aug 28, 2015 at 14:13
  • 1
    $\begingroup$ @slebetman: Deusovi's answer doesn't explain that either, and I've made similar comments on that answer. $\endgroup$ Commented Aug 29, 2015 at 3:52
  • 1
    $\begingroup$ @slebetman: No, the burden of proof is on the person claiming to have a solution. Simply asserting that there are no counterexamples is not a solution; it's as bad a non-answer as simply asserting that the table can always be balanced. $\endgroup$ Commented Sep 1, 2015 at 1:01

I can't tell if anyone has posted this reference, but here is an arXiv link entitled ON THE STABILITY OF FOUR-LEGGED TABLES with the following abstract:

We prove that a perfect square table with four legs, placed on continuous irregular ground with a local slope of at most 14.4 degrees and later 35 degrees, can be put into equilibrium on the ground by a “rotation” of less than 90 degrees. We also discuss the case of non-square tables and make the conjecture that equilibrium can be found if the four feet lie on a circle.

So your answer is it has been proven for certain constraints on how irregular the ground it.

  • 2
    $\begingroup$ Interesting constraints. Deserve upvotes, XD incredible someone studied that with a research XD. $\endgroup$ Commented Apr 8, 2016 at 8:57

The correct proof

First we assume that the floor has elevation that is a continuous function of the position bounded between zero and the $p$ where $p$ is some parameter that depends on the shape of the table (see below for details), and that the table legs are lines that end in the points of a square $ABCD$. The problem is then assumed to be to position the table so that all four legs touch (and do not go through) the floor.

In the following method note that every step is physically possible in the sense that the table never intersects the floor.

First note that there is a rotational orientation of the table (based on $p$), such that for any point on the floor we can hold the table in that orientation such that one leg touches the floor at that point. This is possible because we can hold the table sufficiently tilted and above the floor, so that the end of the lowest leg is directly above the desired point, and now lowering the table results in that leg touching the floor first by the intermediate value theorem (IVT) and the tilt. We can assume that $A$ is the end of that leg touching the floor at the chosen point. Now tilt the table around the line through $A$ that is perpendicular to $AB$ and horizontal, such that $B$ goes towards the floor, until the leg ending at $B$ touches the floor by IVT. Tilt the table around the line through $AB$ until a third leg touches the floor by IVT. Note that if the third leg is not there we can tilt some more so that the last leg touches the floor by IVT.

Note that in the above subroutine it is easy to choose the initial orientation of the table such that rotating the table by any angle around the vertical will give yet another initial orientation that works. So we get a function with its inputs being the desired point and the initial rotation around the vertical and its output being a table 'position' with $ABC$ on the floor (and $D$ possibly below). It is not hard to prove that this function is continuous, given some weak conditions on the floor (see below). This is the key.

Perform this subroutine once to get $ABC$ on the floor. To be precise we simply get any three legs on the floor, and then rotate the table such that those legs are the ones that actually have endpoints being $ABC$. Then choose any path $P$ from $A$ to $B$ on the floor. Drag the table on the floor such that $A$ follows $P$. We claim that we can do so with $B,C$ remaining on the floor. This follows from the subroutine's continuity since we can keep the rotation around vertical fixed and just use the subroutine on the points along the path. In fact this is what can actually happen physically if you constrain the table's movements. Now drag the table on the floor such that $A$ remains where it is while $B$ goes to where $C$ originally was. Again we claim that we can keep $B,C$ on the floor in the process of dragging. This is because the subroutine applied to all possible rotations around the vertical gives all possible table 'positions' where $A$ is in that place and $B,C$ are on the floor. In particular, a quarter turn around the vertical gives a table 'position' where $A,B$ are at the original locations of $B,C$, and in that case $D$ is below the floor. By IVT, there is a smallest rotation less than that quarter turn such that the subroutine will give a table 'position' with $D$ exactly on the floor. This means that we can drag the table to move $B$ 'around' $A$ keeping $B,C$ on the floor because of the subroutine's continuity, and before a quarter 'turn' we would have found the solution.

Table parameter

Clearly if the table top is much wider than the distance between legs, the method can fail to work simply because the side of the table top hits the floor. In the worst case the legs are so short and the table top has some downward protrusions that reach the plane through the square! In that case certain floors would make the problem impossible. However, as long as there is a plane that completely separates the leg ends from the table top, there will be some suitable parameter $p > 0$ for the above method to work.

Similarly if the legs point outwards, the middle part of the legs may hit the floor if we use the method. Again, for reasonable tables there is some suitable parameter $p$ that works. (No I'm not going to rigorously prove this.)

Floor conditions

Also, to ensure that the subroutine works, any sphere with centre on the floor and radius $AB$ must intersect the floor at points that are at unique angles from the centre. I'm not sure whether this condition can be removed without making the solution non-constructive, in the sense that we can easily prove existence of a solution but there may not be a physical method (like the one above) to obtain it systematically from any random starting position.

In any case, this condition is fulfilled by various simpler conditions such as having Lipschitz constant at most $1$.

  • 1
    $\begingroup$ Wow. This deserves more upvotes, even if it would be wrong. +1 $\endgroup$
    – user21233
    Commented Sep 15, 2016 at 15:37
  • 1
    $\begingroup$ @RudolfL.Jelínek: I'm a mathematician and I can certainly assure you that my answer is the only one that uses IVT correctly. My answer had been downvoted by some people who did not like mathematical correctness, which is also why I quit participating on Puzzling SE. $\endgroup$
    – user21820
    Commented Sep 16, 2016 at 1:28
  • $\begingroup$ Yes, indeed, this is beautiful. $(+1)$ :D $\endgroup$
    – Mr Pie
    Commented Oct 4, 2018 at 21:32


The answer is no, there exist at least one counter example of floor that prevent the table to be placed, don't get me wrong, it is perfectly possible that the chance to that floor to exists is so low that don't really matters, but I'll show that exists a whole family of floors preventing table placement.


So given I gave a counter-example, no, it is not always possible to place the table


Assume the floor seen from aboxe is just a square/rectangle, one of the corners of the square is almost (I give a small delta so I can claim the floor to be continuos) centered on a "hole" and the vertical floor section is 1/x.enter image description here

More Info:

You can get a table touching on 4 legs if it is a non-squared shape (in example something like a piece of cake), since the table leg's are placed around corners of a rectangle/square, the 4th leg can't however be placed correctly. I builded this answer around Kevin answer, he sais it is possible, however I used his construction to find out that actually is it not possible (you were so near to the solution XD).

The family of floors

All non-linear polar constructed floors, monotonic 2d functions windows that are contained within a quadrant, don't allow the table to be placed. (the family can be actually bigger)


You can always find 2 points on the surface to get a line, however you place the line, if you turn the table to get the third leg touchin the 4th leg will not be touching.

What's new

Other answers give either a wrong answer (Yes), or mention cases that are either not clear where the solution lies (need constraints, and readers have to figure them out) or have no intuitional proof. I think the one I give here is very easy to reason about.


The construction still holds, but needs some adaptation, as noted in the comments, I can still place legs of the table on 2 elevation rings, however If I skew the floor along 1 axis, and I cut out simmetric planes, I'm no longer able to place the table: enter image description here

Conjecture (true, but need proof)

Given 2 ellypses of different diameter (same center and not circles), except for pairs centered on a simmetry axis, there are no pair of parallel lines whose intersections with the 2 ellypses are the corners of a rectangle.

In the end:

We need also to "stretch" the floor in one direction and cut out simmetry axis in additiction to the family of floors I constructed above.

  • $\begingroup$ "The only way to get something with 4 legs to touch 4 points on the floor is by having it to be 'cake piece shaped'." I don't believe this. Why should it be true? $\endgroup$
    – f''
    Commented Apr 8, 2016 at 18:02
  • $\begingroup$ Well, it was a small inaccuracy, " a sufficient condition to make legs fits is to have a table chake-piece-shaped". I'll edit the answer, the rest is still perfectly valid. $\endgroup$ Commented Apr 11, 2016 at 9:05
  • $\begingroup$ Well @f'' it is true at least on the "1/x" floor. $\endgroup$ Commented Apr 11, 2016 at 10:45
  • $\begingroup$ A cake-shaped table can be placed on the floor, but that doesn't prove that a square table can't. In fact, there is a way to place a square table on this surface. $\endgroup$
    – f''
    Commented Apr 11, 2016 at 14:14
  • $\begingroup$ True, however a slight scale along one of the 2 axis (top view) would prevent the table to be placeable. $\endgroup$ Commented Apr 11, 2016 at 14:30

I'm not entirely sure if this counts as lateral thinking, but couldn't you just chop off a leg? Looking at the answers the problem seems to be one leg connecting with the floor preventing the opposite leg doing so.

  • 1
    $\begingroup$ But then it wouldn't be a 4-legged table anymore, which goes against the very question itself. $\endgroup$
    – McMagister
    Commented Aug 28, 2015 at 1:28
  • $\begingroup$ @McMagister is right: the table has to remain 4-legged and perfectly symmetrical, as stated in the OP. This does count as lateral thinking, I'm afraid. $\endgroup$ Commented Aug 28, 2015 at 8:00
  • 1
    $\begingroup$ @McMagister BTW, you can chop just a part of it ;) $\endgroup$ Commented Aug 28, 2015 at 9:06
  • $\begingroup$ What with my other question, I don't seem to be able to find a middle ground between overly lateral and overly literal... $\endgroup$
    – YoshiBoy13
    Commented Aug 28, 2015 at 10:00
  • $\begingroup$ You know what, the table does have to remain $4$-legged, but for uniqueness, Imma give you a $(+1)$ :D $\endgroup$
    – Mr Pie
    Commented Oct 4, 2018 at 21:36

I don't think you can solve this question infallibly unless you approach it using a theoretical approach specifically in multivariable calculus. If you reword your question into a mathematically approachable model, then the question goes:

Let $m > 0$ be the greatest slope that is acceptable for a "realistic floor." Let $l$ and $w$ be the length and width of the chair.

Define $S$ to be the set of all continuous surfaces such that for each surface $S_i$, $|{dz \over {dx}}|,|{dz \over {dy}}| < m$ at every point in $S_i$.

Define the set $P_{A,B}$ be all the possible contact points (x,y,z) between a continuous surface $A$ and a plane $B$, both of which are in $\mathbb R^3$ (the 3rd dimension).

Does $S$ have the property that there always exists a plane $B_i$ such that $P_{S_i, B_i}$ contains 4 contact points that form a rectangle of dimensions $l, w$ when connected?

I simply don't have the knowledge to derive and justify the result, and I don't think that kind of infallible solution would be easy to get here; math.stackexchange would be more appropriate.

I have only posed more questions so far, so to make up for this (kind of), I will hypothesize that the answer to the question is no; if I assume that $S$ does have said property for dimensions $l,w$, then the same property must be applicable for really large dimensions and really small dimensions (e.g. where $l,w \in (0, \infty)$ ), and I simply find that unfeasible.

  • 2
    $\begingroup$ This is an interesting study in how to make a simple problem more complicated and harder to understand, but it doesn't really answer the question I asked... $\endgroup$ Commented Aug 27, 2015 at 18:12
  • 3
    $\begingroup$ Sorry, translations into math cannot always be simple without losing precision. The 4 points of contact must lie on a flat plane, as described by your legs-with-the-same-height description. I define $P_{A,B}$ as the family that those 4 points must lie in. If you want to prove that this applies to every realistic surface you can imagine, then you want to define a set of those realistic surfaces for notation purposes. That explains $S$. Is there anything else that I'm missing? Or can you please explain why your interpretation conflicts with my mathematical translation $\endgroup$
    – user263104
    Commented Aug 27, 2015 at 18:37
  • 1
    $\begingroup$ Oh nevermind, I understand what you mean. I intended to make a comment to offer this as help, but having only 1 point limits my options. $\endgroup$
    – user263104
    Commented Aug 27, 2015 at 18:47
  • $\begingroup$ That's fair enough. I would flag a mod to convert this to a comment, but it's way too long for a comment! I actually know more than enough maths to formulate this in some incomprehensible way, but I liked the simplicity and practicality of the original problem. $\endgroup$ Commented Aug 27, 2015 at 18:50

You may be asking the wrong question. As stated, no it isn't always possible to position an 'ideal table' on an uneven floor and guarantee that all four legs have firm contact with the ground. The real question is it is possible to always distort a floor such that an ideal table will wobble.

Start with an 'ideal floor' (perfectly smooth and level) made from soft wood, place your ideal table anywhere. No wobble. Now take a hammer and knock out a small uneven plug from the floor under one of the legs. Now the table wobbles. Move table and if doesn't wobble, repeat hammer knock out. Repeat until you have a floor upon which there is no position where the table doesn't wobble.

Given a randomly sized (table width, depth) and a randomly uneven floor, you might be able to place the table somewhere it doesn't wobble but you might not. Therefore it isn't always possible.


I decided to update my answer since people seem to have misunderstood.

I'm not suggesting that the question ask was wrong. I simple thought that reframing the question might help people see the answer.

I'm not suggest that the floor is continuously moving. The floor is "uneven" there are lots and lots and lots of ways that a floor can be uneven. I simply started with an even floor and made it uneven to make a point.

How do I know that my "repeat until you have ..." loop is guaranteed to terminate? Given the problem statement we know there are a finite number of positions available for the table on the floor (there must be some distance, no matter how small that defines a "new" position for the table). Lets call the number of possible configurations n. Take the set of all uneven floors, in that there has to be at least one floor where there is only one position where the table wobbles (if there were no positions then the floor would be even). There must also be a floor where there are only two positions where the table wobbles. One where there are only three positions, etc. Therefore, there must be at least one floor that has only one position where the table table wobbles. Start with that floor. Take hammer and 'unlevel' the floor under one of the legs, call that point p. In order for the loop to never end the change of p must lead to a new position which is a rotation around point p where the table does not wobble, but there is still only one position on the floor where the table doesn't wobble. Hit p again, repeat. Since there are only n possible positions of the table in the room you cannot repeat the procedure more than n times. Since you are not changing any other points you will by eventually have a floor where there is no position where the table doesn't wobble. Therefore, there is at least one floor where you can not position the table in a wobble free position. Thus it is not ALWAY possible to position the table so that it doesn't wobble.

  • 2
    $\begingroup$ ...The question he asked is perfectly acceptable. $\endgroup$
    – Deusovi
    Commented Aug 27, 2015 at 1:27
  • 6
    $\begingroup$ I agree about the equivalence: if it's possible to construct a floor on which the table cannot fit, then it follows it's not always possible to make the table fit on every floor. But you haven't shown that the procedure of knocking holes in the floor actually terminates with a floor on which the table doesn't fit. So you haven't shown a way to construct (or, in this case, destruct) the floor you're looking for. For example, how do we know that I won't eventually knock a hole in every location, and I'm right back where I started? $\endgroup$ Commented Aug 27, 2015 at 14:17
  • $\begingroup$ @Steve Jessop, by definition you don't stop if the table doesn't wobble. Look at it another way, if you position the table so it doesn't wobble I can always 'adjust' the floor so that it does. The original problem states only that the floor is uneven (not how it is uneven) and that the table 'always' can be positioned to not-wobble. $\endgroup$
    – Dweeberly
    Commented Aug 27, 2015 at 23:22
  • 1
    $\begingroup$ I think as per the definitions of the problem by the OP we can safely assume that a continuously moving floor doesn't qualify as a "floor" (not without violating the "no tricks" clause) $\endgroup$
    – slebetman
    Commented Aug 28, 2015 at 9:03
  • 1
    $\begingroup$ What makes you think that your "repeat until you have ..." loop is guaranteed to terminate? $\endgroup$ Commented Aug 28, 2015 at 12:20

The answer is certainly no, the best way to think about this is in terms of basic linear algebra, or if math is not your strong point just in terms of a 3d graph.

Your questions translates to "for any four points in 3d space, does there always exist a single 2d plane containing them all". For our problem: the four points are the bottoms of the four legs and the 2d plane is the floor.

We can certainly put any three points in the same plane by just drawing lines between them and defining that as the plane, but we will not always be able to add the fourth (in fact almost every other point will not be in this plane).

To visualise this consider the four points of the form (x,y,z) in Cartesian (normal) coordinates:

  • a = (0,0,0) i.e. this signifies x=0, y=0, z=0
  • b = (1,0,0)
  • c = (0,1,0)
  • d = (1,1,1)

Clearly, a,b and c all lie in the (x,y)-plane, we will call this the floor. However, d is by definition NOT in the (x,y)-plane as the x value is set to zero. It lies exclusively in the (y,z)-plane and will always be shifted away by one unit no matter how much you rotate the points.

Try this out by plotting each of these points in the following link:

point/graph plotter

Rotate the graph around and line 3 points up into the same plane, the third will always appear a unit of 1 to the side.

I hope this helps.


I changed the points to the "table-like" layout

  • 2
    $\begingroup$ Maths is my strong point, don't worry! You're right that 4 points don't define a plane, but you couldn't make a perfectly symmetrical 4-legged table with ends at those 4 points. $\endgroup$ Commented Aug 26, 2015 at 20:08
  • 13
    $\begingroup$ Err, the problem was that the table is perfect (so the for ends of the legs by definition are on a plane) while the floor is not (the floor isn't a plane, but a general continuous surface). $\endgroup$
    – celtschk
    Commented Aug 26, 2015 at 20:48
  • 1
    $\begingroup$ @celtschk is right. I don't know why this has got so many upvotes. $\endgroup$ Commented Aug 26, 2015 at 22:03
  • 1
    $\begingroup$ You show that there exist a non-valid position on a floor you don't showed all positions are invalid. $\endgroup$ Commented Apr 8, 2016 at 8:29

Not the answer you're looking for? Browse other questions tagged or ask your own question.