The wisdom of crowds

[Squawk]

http://www.youtube.com/watch?v=3Sgd1Ibnz2s&feature=sub

My guess, 5

 

[Master_Ghost_Knight]

:facepalm:
I think someone failed dramatically at physics.

 

[nasher168]

:facepalm:
I think someone failed dramatically at physics.

Aye. The result will most likely be a totally random number.
Wisdom of crowds doesn't work if the crowd is guessing randomly. This is not at all like people guessing how many balls in a jar they can see. It is more like a crowd guessing how many balls in a jar they can't see and then filling the jar to see if they were right.

 

[5810Singer]

I was dubious about ShaneDK's experiment at first, but my doubt turned to firm disapproval at the point when he said he wouldn't be spinning the roulette wheel. :facepalm:

Why use a roulette wheel if you're not going to use it properly?

The rotation of the wheel has to count as at least 50% of the randomisation mechanism of the wheel, and IMO the removal of the wheel's rotation totally invalidates the whole experiment.

If ShaneDK was worried about motion blur he should have come up with a different experiment, rather than comprising the randomisation factors.

 

[AndromedasWake]

I really don't see how this tests the wisdom of crowds. When people 'guess' the number of jelly beans in a jar, they do so my constructing a rudimentary model of the jar, and guess the packing fraction of beans/volume of beans. Whether they like it or not, their brains construct models and this is why the wisdom of crowds is successful in that sort of scenario.

So why is he telling us not to try and plot the ball's trajectory? That's exactly what you'd do to form an intuitive guess. Why show us the ball or wheel at all if you aren't wishing to invite people to use visual information to construct a rudimentary model?

Secondly, the system is not that hard to model mathematically. Not spinning the wheel makes it orders of magnitude simpler, because (as was shown in the practise demonstration) if the ball contributes all the motion, it will not end up far from where it strikes the centre of the wheel. Knowing the initial velocity and position of the ball, and approximating the curvature of the table, most graduate physicists could construct the equations of dynamics of the system to predict roughly where the ball would end up much more accurately than blind guesses.

I would guess the ball would end up somewhere between 33 and 35 black, but only because I can see how its about to strike the numbers and make a crude estimate of its trajectory. If I had to guess at any other point (without my dynamical model) I, and by extension a crowd, are probably more likely to choose numbers which are more visually striking. It seems to me that someone has been watching too much Derren Brown.

 

[Master_Ghost_Knight]

It is not only because he never defined the averaging criteria (which he didn't) but because he cheated extremly bad on the experiment, even if he managed to actually do that sucessfully (which isn't hard to do with such a cheating) it would still be true that for the vast majority of aplication intuition gives an interly contradictory answer to what it really happens and thus aplying it would be just wrong.

 

[Aught3]

I hope in the follow up video he will explain the point of this experiment. Other than that I found the video pretty laughable.

 

[Squawk]

Err, ok either I missed the point or everybody else here did.

The point of this experiment was to see how well people just basic mechanics. He showed a video of the ball coming to a stop. He then repeated the throw and paused the motion just before the ball hits the wheel, and invites us to guess where it will stop. This becomes, then, a test of the perception of mechanics, of visual perception of movement.

There is of course a random element to the movement of the ball on the roulette wheel which is part of the experiment.


Lets say, for example, that the ball in the first experiment travelled for a distance of 25 numbers around the wheel before it came to rest. The "logical" place to guess for it to land on the second throw would be 25 numbers from the point where the ball again strikes the wheel. The actual location the ball finishes will form a pretty good probabilility distribution, and this experiment is to see how well the guesses of people observing it follow that probability distribution. Chances are that the crowd won't get it spot on due to random flucations, but I bet they get within two slots either side of it.

To anyone who said that it's not valid because he didn't spin the wheel, I guarantee you did not understand the experiment as portrayed.

 

[Squawk]

Meh, too annoyed, got to offer additional posts

I think someone failed dramatically at physics.

Explain pls.

Aye. The result will most likely be a totally random number.
Wisdom of crowds doesn't work if the crowd is guessing randomly.

This experiment was specifically set up so that you don't make a random guess. What exactly did you think the point of the first video in which you saw the ball come to rest was? Why do you think the roulette wheel was stationary rather than spinning?

We were shown a video so you could see an example of the balls behaviour after first striking the wheel. The wheel was stationary so that it was possible to note exactly where it struck the wheel in the first example, note where it stopped and subsequently get a reasonable idea of how far the ball can be expected to travel after stricking the wheel.

The second video was paused just before the ball hit the wheel precisely so you could use the information from the first spin to make an educated guess, not a random guess, as to the final outcome. Had the video been stopped before the ball left the rim the experiment would have been worthless. As it is it's valid.

This is not at all like people guessing how many balls in a jar they can see. It is more like a crowd guessing how many balls in a jar they can't see and then filling the jar to see if they were right.

This analagy doesn't work for the reasons given above.

was dubious about ShaneDK's experiment at first, but my doubt turned to firm disapproval at the point when he said he wouldn't be spinning the roulette wheel. :facepalm:

Why use a roulette wheel if you're not going to use it properly?

Because it would have made the experiment completely random from the point of view of the person guessing, therefore invalidating it. You were dubious, then disapproving? I was completely the other way, because I recognised that if he had spun the wheel it would have been random and as such about as useful as Derren Brown's wisdom of crowds. As it was, by holding the wheel stationary he gives a valid experiment.

The rotation of the wheel has to count as at least 50% of the randomisation mechanism of the wheel, and IMO the removal of the wheel's rotation totally invalidates the whole experiment.

You didn't understand the point of the experiment as precisely the opposite is true.

If ShaneDK was worried about motion blur he should have come up with a different experiment, rather than comprising the randomisation factors.

He isn't compromising random factors, he is eliminating them.

I really don't see how this tests the wisdom of crowds. When people 'guess' the number of jelly beans in a jar, they do so my constructing a rudimentary model of the jar, and guess the packing fraction of beans/volume of beans. Whether they like it or not, their brains construct models and this is why the wisdom of crowds is successful in that sort of scenario.

So why is he telling us not to try and plot the ball's trajectory? That's exactly what you'd do to form an intuitive guess. Why show us the ball or wheel at all if you aren't wishing to invite people to use visual information to construct a rudimentary model?

Andromedaswake, I'm pretty sure you get it, but misunderstood his modeling thing. Shane is, with this video, not asking you to pick a random number, but is instead interested in our inate ability to work out basic mechanics. It should be excellent, most people can catch a ball which, when you think of the calculations of trajectory involved, is quite staggering. The question is, how well do people judge the trajectory of a ball on a roulette wheel when the roulette wheel is stationary and the ball is about to impact the wheel, given that you have already seen it.

You allude to this in the second part of your post, but then dismiss it with the final sentence which is why I said I don't think you thought it through. I suspect that you were possibly thrown by the comment at 3:04 where he says don't model it. I took that to mean that I should base my guess on intuition based solely on that seen by the first spin, pretty much the same as you did. I tried to figure out where it first struck the wheel and inferred from that how much further around the wheel it would go.

It is not only because he never defined the averaging criteria (which he didn't) but because he cheated extremly bad on the experiment

You're going to have to explain to me how you cheat on an experiment which you have defined yourself. He explained what he was doing and how he was doing it.

Even if he managed to actually do that sucessfully (which isn't hard to do with such a cheating) it would still be true that for the vast majority of aplication intuition gives an interly contradictory answer to what it really happens and thus aplying it would be just wrong.

I have no idea what you are getting at here, I think you have misunderstood the goal of this experiment badly. It would seem you think this experiment is, as AW noted, something akin to Derren Brown and the lottery.

 

[5810Singer]

Err, ok either I missed the point or everybody else here did.

The point of this experiment was to see how well people just basic mechanics. He showed a video of the ball coming to a stop. He then repeated the throw and paused the motion just before the ball hits the wheel, and invites us to guess where it will stop. This becomes, then, a test of the perception of mechanics, of visual perception of movement.

There is of course a random element to the movement of the ball on the roulette wheel which is part of the experiment.


Lets say, for example, that the ball in the first experiment travelled for a distance of 25 numbers around the wheel before it came to rest. The "logical" place to guess for it to land on the second throw would be 25 numbers from the point where the ball again strikes the wheel. The actual location the ball finishes will form a pretty good probabilility distribution, and this experiment is to see how well the guesses of people observing it follow that probability distribution. Chances are that the crowd won't get it spot on due to random flucations, but I bet they get within two slots either side of it.

To anyone who said that it's not valid because he didn't spin the wheel, I guarantee you did not understand the experiment as portrayed.

Actually I did think of exactly this explanation, and then rejected it as I thought it would produce an extremely narrow range of potential answers, so narrow as to make the analogy to the jar of beans practically meaningless.

After coming up with my own guess, taking Andromeda's into account, and then looking at a few guesses on ShaneDK's channel, IMO the current answers give a potential range that is so narrow as to make the test itself meaningless.

When you judge basic mechanics you have so many more visual clues / references than with the jar of beans, that predicting the landing position of the ball is child's play. The only really prominent variable left in the vid sequence is the final bounce/s of the ball, which probably won't change the outcome by more than about 3 places without the wheel spinning, and the impetus of the ball will lend a tendency for the final square to be slightly further on from the landing point, and we can factor that bias into our guesswork.

I don't think any of the guesses are going to be more than three squares out, so what exactly will we learn from proving that a group of very close guesses will, when aggregated, produce a very close aggregate guess?



EDIT: I should have made this plainer in my first post.

I take your point about the need to turn this into a matter of judgement rather than a purely random spin, but my problem with the experiment is that it is now too predictable, and therefore the potential range of answers will be so narrow as to produce a null/inconclusive result.

 

[Squawk]

Given that my guess and AW's guesses are completely on opposite sides of the boards this seems a little dubious.

I confess AW's guess confuses me, I didn't see the guesses in the channel. I have now watched that video of the first spin at least 4 times. I think the ball first hits the wheel somewhere between 35 and 4. I conclude this based on both audio and visual perception. The ball seems to bounce off somewhat, before finally coming to rest between 2/3 and 3/4 of the way further around the wheel.

In the second vid, the one that gets paused, it seems to me that the ball strikes the wheel around the zero for the first time, pretty much exactly where the video is paused. It's a bit of a trick of the mind since the audio continues for half a second.

I then simply guessed a number that was just under 3/4 of the way around the board, and I expected most others to come to a similar, ish, conclusion. I expected most answers to be spread somewhere around my own guess.

What can we predict? As for the ball itself I'm not sure. I don't know how random the bouncing on the wheel really is, but you could spin the ball 10 times and figure that one out pretty quickly. We can make a prediction for the wisdom of crowds.

IF visual perception is good enough to infer where the ball first struck the wheel on the first video and if people can interpolate from that the distance travelled (without counting, which I didn't do) then the most popular square chosen in the second video will be precisely the same distance from the point at which that ball strikes as the distance travelled by the ball in the first video after it struck the wheel, with the squares to either side being of slightly lower probability, and so on and so forth as you move away.

 

[Master_Ghost_Knight]

The matter of fact is I did understood the point of the experiment squawk, aparently you are the one that didn't.
If you replay the clip you will hear that he clearly mentioned that the point of the experiment is to show that extremely elaborated chaotic systems can be predicted by the "wisdom" of the crowds, well this is a complete nonesense, because in chaotic systems if you do not know enough vairables the system will change so widly that it could end up basically almost anywhere (whit different shades of probability of course) rendering it for all effects random. If you have taken probability you would know that you can not cheat a random process because the process itself only apears random and their probability is dependent on the information that you have about the system, and so figuring out the outcome would mean that you have acess to information that you don'thave when you have acessed the probability in the first place.

Now the way he cheated is simple, the system is not chaotic. Because he has freezed the experiment near it's end (eliminating almost all of the unknown factors) and a small change in the factors in this experiment does not result in a big change in the outcome (in fact the change is almost linear) so just by watching the previous experiment and watching the current one you can make a gross estimate of the outcome (because the system will behave similarly).
You have been using this information all along to make your estimate and relying on the fact that the system is NOT chaotic, even you realised this in your comments, you just didn't noticed.

If he in the other hand had spined the wheel (even if slow) and he freezed the video a couple of seconds near the beggining I bet the guesses would be more random. And sense this is a sort of a wrap-arround system (for which he never defined the averaging criteria, so he can make several of them to adjust the result) the result would have been inevitably more like a random guess, and this is why casinos make money.

 

[Squawk]

The matter of fact is I did understood the point of the experiment squawk, aparently you are the one that didn't.

If you replay the clip you will hear that he clearly mentioned that the point of the experiment is to show that extremely elaborated chaotic systems can be predicted by the "wisdom" of the crowds

Actually he didn't. He stated the following.

I wonder, for example, how it applies to basic physics...
very basic newtonian stuff
Can chaotic newtonian systems, that would be a nightmare to work out mathematically, be predicted by the wisdom of crowds

What do we infer from this, and indeed from his experimental setup. Simple, we know that the wisdom of crowds works when given appropriate information. We also observe that a large number of variables have been eliminated. In particluar the wheel is not spinning and we observe the ball as it is about to strike the wheel. The quesiton will be whether enough randomness has been removed for predictions to yield a viable result. It would be possible, as I stated above, to predict which numbers would be predicted the most by observes IF their observation of both spins was accurate. Ie, if they could establish visually where the ball struck the wheel, and the final resting place, then they could extrapolate from this where the ball would likely land next time. If enough randomness has been removed from the system (and my intuition is that it probably has), the distribution of guesses produced by the wisdom of crowds should produce a sound probability distribution that reflects where indeed the ball will land. It would have produced a far more accurate one if two initial spins had been observed.

well this is a complete nonesense, because in chaotic systems if you do not know enough vairables the system will change so widly that it could end up basically almost anywhere (whit different shades of probability of course) rendering it for all effects random.

Which is precisely why the experiment seeks to eliminate a number of variables, the non spinning wheel and the video frozen as the ball hits the wheel rather than earlier.

If you have taken probability you would know that you can not cheat a random process because the process itself only apears random and their probability is dependent on the information that you have about the system, and so figuring out the outcome would mean that you have acess to information that you don'thave when you have acessed the probability in the first place.

The argument I am putting, and indeed the point of this experiment, could be considered a test of whether or not enough randomness has been removed from the system to enable prediction. You will note that this is entirely the point of the experiment, to see if the wisdom of crowds can apply to this particular circumstance. If it cannot we can infer that not enough information could be glened.

Now the way he cheated is simple, the system is not chaotic. Because he has freezed the experiment near it's end (eliminating almost all of the unknown factors) and a small change in the factors in this experiment does not result in a big change in the outcome (in fact the change is almost linear) so just by watching the previous experiment and watching the current one you can make a gross estimate of the outcome (because the system will behave similarly).
You have been using this information all along to make your estimate and relying on the fact that the system is NOT chaotic, even you realised this in your comments, you just didn't noticed.

Err, no. I contend that this is the whole point of the experiment itself. I did notice, I even explained that this was the point. I have done so again in this thread. The question was always whether or not enough randomness had been removed from the system. As you have just pointed out, the system cannot be entirely random if we are to glean information using the wisdom of crowds. Some form of predictability must ensue. This experiment is designed specifically to see if the wisdom of crowds works in this instance. Your "cheat", as it were, is actually teh entire point of the experiment.

If he in the other hand had spined the wheel (even if slow) and he freezed the video a couple of seconds near the beggining I bet the guesses would be more random.

It would be entirely random (not counting for slight bias due to manufacturing error) and would have made the experiment pointless.

 

[Master_Ghost_Knight]

The matter of fact is I did understood the point of the experiment squawk, aparently you are the one that didn't.

If you replay the clip you will hear that he clearly mentioned that the point of the experiment is to show that extremely elaborated chaotic systems can be predicted by the "wisdom" of the crowds

Actually he didn't. He stated the following.

I wonder, for example, how it applies to basic physics...
very basic newtonian stuff
Can chaotic newtonian systems, that would be a nightmare to work out mathematically, be predicted by the wisdom of crowds

Until here so good, everything else you said from this point onwards you just made it up.
You just fited whatever excuse you could find to excuse him. Sorry things don't work that way.

Even the quote you just taken, precisely described what I have said and yet you used it as counter argument:

I wonder, for example, how it applies to basic physics...
very basic newtonian stuff
Can chaotic newtonian systems, that would be a nightmare to work out mathematically, be predicted by the wisdom of crowds [?]

I don't know if you have realised or not, but you are trying to excuse is ineptitude, and I have to add that whatever you do to try excuse him from his ineptitude is completly irrelevant because he has already done it, if you are saying that he didin't meant what he has said then the only person that can correct this is him and he alone because you are not him to know what he meant.

It would be entirely random (not counting for slight bias due to manufacturing error) and would have made the experiment pointless.

Well in order to achieve his point he would have to do it that way, if you think that then the experiment would be stupid in that case I agree, but that is not my fault, he was the one that set it up.

 

[Squawk]

Until here so good, everything else you said from this point onwards you just made it up.
You just fited whatever excuse you could find to excuse him. Sorry things don't work that way.

I made a notable inference from the experiment which you continue to ignore. You have already aluded to the fact that such an experiment is deterministic. For the wisdom of crowds to apply it must be possible to make some inference from the information available. In this case we are given bits of information that MIGHT make an inference possible. We are shown how the ball behaves when it strikes the wheel. We are then shown the point at which the ball strikes the wheel on the second spin.

I contend that althought chaotic, the location the ball finishes after striking the wheel WILL be described accurately a probabilyt distribution, that each number is not equally likely to be the result once the strike point is known. That is, that the system is not "chaotic enough" to prevent inference. The question is whether or not the human eye is good enough to approximate this based on information gleaned from a single prior spin.

I don't know if you have realised or not, but you are trying to excuse is ineptitude, and I have to add that whatever you do to try excuse him from his ineptitude is completly irrelevant because he has already done it, if you are saying that he didin't meant what he has said then the only person that can correct this is him and he alone because you are not him to know what he meant.

I noted his methods, contemplated the experiment itself and decided I like what he's doing. I think there is useful information to be gleaned from it. I wouldn't have posted the video if I thought otherwise, and I wouldn't be defending it now. My only contention before this started was that he might not have removed enough randomness. You seem to be arguing the other way.

It would be entirely random (not counting for slight bias due to manufacturing error) and would have made the experiment pointless.

Well in order to achieve his point he would have to do it that way, if you think that then the experiment would be stupid in that case I agree, but that is not my fault, he was the one that set it up.

After reading this I still contend you don't get it. You seem suggest that he should use the wisdom of crowds to model a situation in which all outcomes are equally likely (from the point of view of the observer). This would be an entirely pointless venture, is clearly not what he is doing and is a complete irrelevance to this thread. It would be doomed to failure from the start.

 

[Master_Ghost_Knight]

If the problem is semidefined then it is not chaotic period.
Thrust me the rest of us know what we are talking about, you are literaly being decieved with cheap magic tricks, wake up.

 

[Squawk]

If the problem is semidefined then it is not chaotic period.
Thrust me the rest of us know what we are talking about, you are literaly being decieved with cheap magic tricks, wake up.

Lol

 

[nasher168]

He says in the video to not try to work out where it's going, thus removing the "wisdom of crowds" effect. If people were allowed to actually guess based on the position and direction etc of the ball, then we could maybe get an effect occurring. Also, I fail to see how one could get a mean result from this like with the beans in a bottle, as the numbers are all fairly randomly laid out around the wheel.

 

[MRaverz]

23 red

I doubt this is going to get anywhere though, jelly beans in a box is different to numbers on a roulette wheel.

 

[Squawk]

He says in the video to not try to work out where it's going, thus removing the "wisdom of crowds" effect. If people were allowed to actually guess based on the position and direction etc of the ball, then we could maybe get an effect occurring. Also, I fail to see how one could get a mean result from this like with the beans in a bottle, as the numbers are all fairly randomly laid out around the wheel.

He didn't say that, he said not to model it. Why do you think he showed the first spin if not for you to make an inference?