From theory to practice - page 741

 
Novaja:

Honest, you couldn't be more honest. If you generate like I do. Give me your row, I'll check it.

I drew you a fairer formula earlier.)

 
Yuriy Asaulenko:

This is the correct way to do it.

in E6 =CHIS(), in F6 = IF(E6-0.5>0,1,IF(E6-0.5<0,-1,0)).

Now the coin is fair.

Ok, will do and compare with mine in terms of performance.

 
Yuriy Asaulenko:

Erased the message. Didn't see that it only gives -1,0,+1. You can use it in this form as well. Zero will come in handy, at least not in the way)).

I know what you mean. In this case I'm interested in a fair coin for now.

 
Department stores with sharpened coins make decent money. Coins don't work in the forex market.
 
Novaja:
And if we imagine that all that surrounds us is not a randomness, but only a consequence of regularities, then all random processes are automatically transferred to the rank of regular, not using the word deterministic, regular, with the consequence that both SB and VR can be ranked as the same processes, with a structural difference.

SB is a completely random process and BP, if you mean the market, is not very random, just a little bit. Just the little bit that you and I and others like us contribute. Maybe someone else does, but it is not a signal, it is noise. It is noise because it is easier to describe fluctuations as random processes than to look for their causes. Randomness is an unknowable pattern. Forgive my nerdiness.

 
Novaja This is how my system earns SB, who has any say?

It's not "earning", it's the same SB as the source data.

 
Novaja Maybe this is how the equity graph should grow?

Yes, the right system has about this, and on OOS.

 
Alexander_K:

What's left to do? I don't understand...

If the process is as similar as possible to the SB, and you can't make money on the SB (as the do-gooders assure us), then what are we doing in the market?!!!

"Similar" does not mean "equal". You can make money on the differences. Yes, there are 1-2% differences. This is quite enough, they have nothing to do with the percentage of profit.

Of course, the differences lie in memory (dependence of future changes on past changes), not in distributions.

 
Alexander_K:

To prove that we are facing a process as close to the Variance Gamma Process as possible, I did a little experiment.

1. Took CLOSE price data for EURUSD for 2017.

2. Set the sliding window = a week.

3. Calculated the average value of the spread (Max-Min) for the year.

4. Calculated the average variance for the year, calculated using the formula for Variance Gamma Process.

Results:

Amazing coincidence!!!

We used to have an analogue calculating machine "Vesna" in Noginsk. We didn't do shit on it. And in Kiev - at the college there was also something analogue, I don't remember the name of it, but they did laboratory work.

 
Yuriy Asaulenko:

No go. For starters= SLU(-1;1). If >0, then 1. If <0, then -1.

You need FACTOR(0;1). If 0 then -1. If 1, then 1.