Serving the Quantitative Finance Community

Useful article?

Yes
2 (50%)
No
2 (50%)
 
Total votes: 4
 
User avatar
finanzmaster
Topic Author
Posts: 61
Joined: March 11th, 2007, 1:04 pm

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

October 11th, 2016, 12:49 pm

I'm telling you for the 5th time now 17% is not rare, by any standard, you can wiggle all you want.
Yes, contraty to you, I do listen to what you say and admit that formally I didn't achieve (so far) a commonly agreed critical confidence level of 95% or 99%.But commonly agreed is not an axiom. 

As to the spurious correlation, let me cite Elena Wentzel once again: "first of all, we should judge about correlation not according to the correlation coefficient but according to the common sense. So the mass and size of a bomb splitter are likely correlated, whereas the velocity of a bomber and its altimeter error are likely not" 

But if you want this glorified confidence level, here you are! :)
Given, that Einstein flips a coin, the probability to achieve  553 of 883 is about 1:50000 (we did agree on this, didn't we?)

Further, if we flip a coin and selected the sequences with Einstein's or better level, the probabily of negative return (conditioned on this post-selection) is <= (883-553)/883 = 0.373
But Einstein has once 10 consequent(!) negative returns (you can browse his trade history on wikifolio's website, start from
16.10.2014 22:09). 
And the probability to get a series of 10 losses is in this case 0.028
Since the number and order of "+" and "-" are independent, we can multiply these prpbabilites and thus obtain the critical confidence level.
 
User avatar
finanzmaster
Topic Author
Posts: 61
Joined: March 11th, 2007, 1:04 pm

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

October 11th, 2016, 12:52 pm

I'm telling you for the 5th time now 17% is not rare, by any standard, you can wiggle all you want.
Yes, contrary to you, I do listen to what you say and admit that formally I didn't achieve (so far) a commonly agreed critical confidence level of 95% or 99%.But commonly agreed is not an axiom. 

As to the spurious correlation, let me cite Elena Wentzel once again: "first of all, we should judge about the dependence not according to the correlation coefficient but according to the common sense. For example, the mass and size of a bomb splitter are likely correlated, whereas the velocity of a bomber and its altimeter error are likely not" .

But if you want this glorified confidence level, here you are! :)
Given, that Einstein flips a coin, the probability to achieve  553 of 883 is about 1:50000 (we did agree on this, didn't we?)

Further, if we flip a coin and selected the sequences with Einstein's or better level, the probabily of negative return (conditioned on this post-selection) is <= (883-553)/883 = 0.373
But Einstein has once 10 consequent(!) negative returns (you can browse his trade history on wikifolio's website, start from
16.10.2014 22:09). 
And the probability to get a series of 10 losses is in this case 0.028
Since the quantity and the order of "+" and "-" in a sample are independent, the joint probability is (1:50000) * 0.028  these probabilites and thus obtain the critical confidence level!
Last edited by finanzmaster on October 11th, 2016, 12:57 pm, edited 2 times in total.
 
User avatar
finanzmaster
Topic Author
Posts: 61
Joined: March 11th, 2007, 1:04 pm

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

October 11th, 2016, 12:53 pm

# Generating a single run and counting 
# if there are 10 consecutive negative returns 

consecutiveSuccess <- function(p, count, n) {
    trials <- rbinom(n, 1, p)
    for (i in 1:(n-count+1)) {
        if (sum(trials[i:(i+count-1)]) == count)
            return(1)
    }
    return(0)
}

# simulation starts here
set.seed(1236)
t = 100000
p = (883-553)/883
count = 10
n = 883

sum = 0
for (j in 1:t) {
    sum = sum + consecutiveSuccess(p, count, n)
}

prob <- (sum*1.0)/t
( credits to http://math.stackexchange.com/questions ... n-100-runs )
 
User avatar
outrun
Posts: 4573
Joined: January 1st, 1970, 12:00 am

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

October 11th, 2016, 1:26 pm

Looks all good.

Are now going to do cluster analysis of negative returns? What's the goal of the analysis: to show (or not) that Einstein has less clustering / runs of losses than a random coinflipper would?
 
User avatar
wanaquant
Posts: 44
Joined: May 12th, 2012, 1:19 pm

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

October 12th, 2016, 5:07 pm

Is trading really comparable to flipping coins? No way in my opinion. First, the outcome of trading a particular stock is not binary; and second, you can (to some degree) chose the outcome.

If you open a position, you have the choice to close it from the moment you open it. It fluctuates up and down during that time it is open. You can close it (and collect your winnings or make a loss).

Flipping a coin is very different. It's like making an "instant trade". You click "Trade" and then you have "+5%" or "-5%".
 
User avatar
finanzmaster
Topic Author
Posts: 61
Joined: March 11th, 2007, 1:04 pm

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

October 12th, 2016, 7:34 pm

@outrun
Looks all good.
Are now going to do cluster analysis of negative returns?
Of course not. You found a formal flaw in my argumentation. I found a formal argument, which eliminates this flaw. Though from common sense point of view it is somewhat stupid to distinguish an outstanding trader by a series of losses.

By the way, are you a fresh graduate or a student? I assume it because you have just recently joined the forum and mean it in no way offensive. 
Contrary, I am quite sure that if you are a student or graduate, you are top among your peers. I, myself, was top and similarly mathematically (over)fastidious (my first essays after the graduation were not on trading but on measure theory: http://yetanotherquant.com/#MeasureTheoryBook ). 
But it goes ... as we used to joke in Russia: in business nobody wants your deduction or induction ... only production! ;)
Is trading really comparable to flipping coins? 
Generally not! (well, if returns are more or less symmetric and homogenious, a binomial distribution can be a good approximation).
But flipping coins is originally not my approach. It is the vulgar adherents of the Efficient Market Theory, who talk about flipping coins (or about monkeys pressing buttons). 
And I have shown: at least some of traders like Einstein do not flip a coin but rather have an edge.
 
User avatar
finanzmaster
Topic Author
Posts: 61
Joined: March 11th, 2007, 1:04 pm

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

September 23rd, 2017, 12:50 pm

###############################################################################
Another star trader on wikifolio, from my point of view even better than Einstein on the risk-adjusted basis.
https://letyourmoneygrow.com/2017/09/23 ... wikifolio/
BTW, Einstein kept (as expected) his pace and approaches the return of 2500%!
 
User avatar
TraderWalrus
Posts: 28
Joined: March 24th, 2017, 6:54 am

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

September 24th, 2017, 1:42 pm

Showing paper trading instead of live trading records presents many issues.
Some systems are having difficulties assessing if the executions are realistic (especially with instruments with large spreads), not to mention the possibility for a person to open several virtual accounts in the system, running several uncorrelated strategies, then taking credit for the one with exceptional results over a certain period. Obviously, in reality the sum of those strategies will have random results and is very unlikely to beat the market consistently.
 
User avatar
finanzmaster
Topic Author
Posts: 61
Joined: March 11th, 2007, 1:04 pm

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

September 30th, 2017, 7:37 pm

Showing paper trading instead of live trading records presents many issues. 
The issues you mention do, of course, take place.
However, it is not so easily to open many accounts on wikifolio (unless under multiple logins but it might be easily detected), let alone, it is very time consuming to run many such trading-intensive strategies (wikifolio allows no automatic execution).
Secondly, these strategies are indeed non-scalable but for a private person with €100K it is not a problem (finally, these €100K would be turned into €2 Millions).
As to spreads, on wikifolio they are often larger even in demo-modus than by a broker that provides a direct access to Xetra, NYSE and NASDAQ. 
 
User avatar
TraderWalrus
Posts: 28
Joined: March 24th, 2017, 6:54 am

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

October 1st, 2017, 4:01 am

Super. Where do I collect my 2 million Euro then? I've already made a downpayment on a mansion so they'll really be handy as soon as possible, thank you!
 
User avatar
finanzmaster
Topic Author
Posts: 61
Joined: March 11th, 2007, 1:04 pm

Re: A case study to show by means of statistical test: this guy (it is NOT me) can systematically beat the market

October 2nd, 2017, 7:49 am

Super. Where do I collect my 2 million Euro then? I've already made a downpayment on a mansion so they'll really be handy as soon as possible, thank you!
Fairly easy!  :D
1) Callback your downpayment (remember, you need at least €100K initial capital)
2) Watch and copy HBecker on wikifolio 24*7 (well, from 09:00 to 18:00 on business days), trade history is updated in real time.
3) Open a Depot by the broker that offers cheapest fees for European stocks, to my knowledge it is DeGiro
4) Mit etwas Glück (with some luck) you will have €2.000.000 in a couple of years

Oh, almost forgot: move to a country, in which there is no taxes on capital gain... However, if you are American, you still have to pay taxes to Uncle Sam, don't you?