According to the Sharpe ratio - page 6

New comment
 
Aleksey Nikolayev:

Sharpe will equal infinity if all trades have the same profit - this is only possible if they correspond to a sequence of deposits at the same interest. I would say that the physical meaning of Sharpe is proximity to a constant interest deposit - the bigger it is, the closer it is.

In your example, the Sharpe would be the same because you get a multiplication of a random variable by a constant. The mean and RMS would be multiplied by the same number, which would be reduced by being in the numerator and denominator.

That's exactly what I wanted to hear. (emphasis added). So after all, the size of the deposit has nothing to do with the calculation of the coefficient.

Then how to understand it (?):

"I would say that the physical meaning of Sharpe is the proximity to a constant interest deposit - the bigger it is, the closer it is."

 
Renat Akhtyamov:

That's exactly what I wanted to hear. (emphasis added). So the size of the deposit has nothing to do with it after all.

Then how do you understand this (?):

"I would say that the physical meaning of Sharpe is in its proximity to a fixed interest deposit - the bigger it is, the closer it is."

What is meant is that when you deposit in a bank with constant interest, the Sharpe will be infinite regardless of the amount invested and the specific interest value. The Sharpe for CU is always finite, but the larger it is, the more the operation of our CU is like depositing money in a bank at fixed interest.

 
Aleksey Nikolayev:

It means that when investing in a bank with constant interest, the Sharpe will be infinite regardless of the amount of money invested and the specific interest value. Sharpe for TC is always finite, but the bigger it is, the more our TC will work like depositing money in a bank at interest.

Here I completely agree.

I read articles about this coefficient at the time when there was no controversy.

As I recall, everyone was unanimously saying: the greater the profit as a percentage of the amount of money invested in the deal, the higher the coefficient.

But time changes views, so let it be as it is now.

PS

And if you take into account the time of the deal, it's a different coefficient, talking about turnover.

 
Renat Akhtyamov:

And if you take into account transaction time, that's a different ratio, talking about turnover.

In my opinion, it is closer to "annualised Sharpe", which is absolutely necessary when moving from an individual TS to their portfolio.

 
Maxim Romanov:

Now I was checking how my self-adaptive robot can tune to a known signal consisting of a mixture of sine waves. But that's not the point, I got a great result and remembered about the Sharpe Ratio and looked what ratio is shown in the tester.

So with a perfect yield chart, Sharpe is 0.82! At the same time the drawdown of funds is 972$ and profit is 406000$. It's not even close to 1. But the point is that the test is on a harmonic series and it is impossible for a robot to fail there, but anyway according to the widely known criterion Sharpe must be greater than 1, the strategy looks bad.

This chart has coefficient 0.82


Let me tell you a secret, I have a Sharp over 4. There is also a table on the monitor, in it , the risk of losing 10% of the depo is <0.01, for that you need to make an infinite number of trades. Here are the facts, didn't make it up.
For those who say it does not show anything, it is not true. They just turn a blind eye to the imperfection of their TS, and you cannot cheat Sharpe
 
Sprut112:
Let me tell you a secret - my Sharp is bigger than 4. There is also a table on the monitor, in it , the risk of losing 10% of the depo is <0.01, for that you need to make an infinite number of trades. Here are the facts, didn't make it up.
For those who say it shows nothing, this is not true. They just turn a blind eye to the imperfection of their TS, and Sharp cannot be fooled.
When I had a signal there was the same probability in the table too.
But I don't agree with this assessment, because I calculated risks by hand based on peculiarities of that algorithm.
And in my example I cited, it was impossible to crash because the series is harmonic and if there is one profit, the robot will always react to it in the same way.
 

Here's an example where my sharpe was 0.82. Obviously, the robot won't lose any more and the probability is 100%. Nevertheless the ratio is below 1 and the fact thatSprut112 hasmore than 4 sharpening confirms the low meaningfulness of this ratio. It is clear that any robot can fail in the real market, while it will never fail in the harmonic series if it has shown profit. And so it turns out that the robot with Sharp 4 trading on the real market is more reliable than the robot with 0.82 trading on the harmonic set, which is obviously not true.

 
Maxim Romanov:

Here's an example where my sharpe was 0.82. Obviously, the robot won't lose any more and the probability is 100%. Nevertheless the ratio is below 1 and the fact thatSprut112 hasmore than 4 sharpening confirms the low meaningfulness of this ratio. It is clear that any robot can fail in the real market, while it will never fail in the harmonic series if it has shown profit. Thus, the robot with 4 sharps trading on the real market is more reliable than the robot with 0.82 trading on the harmonic series.

I wonder what and how this "picture" was "drawn"?

 
aleger:

I wonder what and how this "picture" was "painted" with?

It's just the sum of 20 sinusoids. Then I put it in custom symbols from excel. I wanted to see if the robot could adapt to such a simple signal or not.
 
Maxim Romanov:
It's just the sum of 20 sinusoids. Then I put it in custom symbols from excel. I wanted to see if the robot could adapt to such a simple signal or not.

Thank you. It looks like a normal graph, so I was a bit puzzled.

1234567
New comment