Chart Game: Real vs. Random


Ran across an interesting link on Steenbarger's Traderfeed blog.

http://web.mit.edu/alo/www/Papers/arora.html

Can you tell the difference between a randomly generated chart and a real one? Research has been conducted to find out if subjects could distinguish between random vs. real. Find out the results and play the "online game" yourself if ya like. The registration is quick and they don't ask for a bunch of personal info.

If anything it's entertaining ... a little bit of fun ... and perhaps a learning tool/experience in some way or another.
I'd done the practice a couple of times on the different "market names" offered when I'd initially started this topic. I only did so-so. Went back to it again today and played the "Bull" and I think I got all of 'em right (maybe missed 1).

Click image for original size
trading game thingy


Before anyone thinks this is about being the big man on campus, let me disabuse ya of the idea real quick. Initially, I'd gotten the first 8 out of 10 WRONG in training. Then I noticed something ... not about the way the chart was "unfolding" but about the chart line of its supposed price history. The wrong ones had a little more of a consistently jagged look, as if it was generated with a 5-year old using an old Etch-a-Sketch.

So, it wasn't so much about recognizing an actual price chart (at least in this particular example), it was about recognizing what was NOT a real price chart. Then it was easy and I simply didn't click on the one that looked like an Etch-a-Sketch. I wonder if this is simply a characteristic of the chart game named "Bull" or if this might be a design flaw.
Originally posted by MonkeyMeat

...Then I noticed something ... not about the way the chart was "unfolding" but about the chart line of its supposed price history. The wrong ones had a little more of a consistently jagged look, as if it was generated with a 5-year old using an old Etch-a-Sketch.

So, it wasn't so much about recognizing an actual price chart (at least in this particular example), it was about recognizing what was NOT a real price chart. Then it was easy and I simply didn't click on the one that looked like an Etch-a-Sketch...


From my post on 3/27:

"...I am just looking for certain 'price signatures' that simply don't happen in 'real' price action. I found that after a very short period of adjustment I was able to make it through an entire series and find the 'real' price action without a single miss. I wasn't looking for anything that said 'real', I was always looking for a signature that said 'not real'..."

So, you discovered what I did, a 'look' to price action that simply doesn't exist in real price action. The problem is, I'm sure the random walker researchers can now take this and apply various functions on random data to remove what we are seeing. Given enough manipulation I'm sure real action could be mimicked or modeled. But then I'm sure I can find other subtleties, and then if I reveal them, they can manipulate the data more to hide those. Probably at some point I may not be able to 'beat' the computer as it tries to disguise the differences.

No matter what they do, though, it can never 'make' the market random if it's not just because a random model can be manipulated with great effort to finally look like something real. I mean, would a mountain range be declared not to exist in the real world if a fractal generator did 10^25 iterations and then finally couldn't be told from a picture of the real mountain?

I found that all the trials had differences I was able to detect, and I was able to do basically the same thing on all series. You just need to see what those 'unreal' price action signatures are for each series.