Machine learning in trading: theory, models, practice and algo-trading - page 1143

 
pantural:

Look, Wikipedia is no substitute for WO, Wikipedia doesn't take into account the whole gamut of what happens in practice, that equity lengths\PnL can be any length, and Wikipedia implies that you only measure over a year and no other way where a clear number of daily returnees.

Read http://economic-definition.com/Other_branches_of_mathematics/Koefficient_Sharpa_Sharpe_Ratio__eto.html for example.

In fact, anyone who has done a lab at least once, should know about rationing SR

The point this user is making can be seen in the video here at 4 min 30 sec - https://ru.coursera.org/lecture/portfelnyye-investitsii/vidieo-6-schitaiem-koeffitsiient-sharpa-v-eksielie-WjPm0


Видео 6. Считаем коэффициент Шарпа в Экселе - Активные и пассивные портфельные стратегии | Coursera
Видео 6. Считаем коэффициент Шарпа в Экселе - Активные и пассивные портфельные стратегии | Coursera
  • ru.coursera.org
Video created by National Research University Higher School of Economics for the course "Портфельные инвестиции: активные и пассивные стратегии". Дорогие слушатели! Вторая неделя нашего курса посвящена изучению активных и пассивных портфельных ...
 
Rashid Umarov:

It's good that you know how to use the search. https://smart-lab.ru/blog/267416.php

Was just sitting there looking at forex, impressed by the posts above. Particularly about the Sharpe Ratio over 3.

Yes, of course it is possible to create HFT strategy and I showed its profitability somewhere in one of the threads...

BUT

As soon as the spread appears, everything goes to hell.

On the other hand, when you reduce the number of trades, it also turns out a good strategy, but again, BUT...

Oh, these Forex trend slippage, i.e. those players who opened correctly are killed by Sharpe ratio again...

I don't even know how one can get such a high coefficient...

 

And in general on the subject of the connection between HFT and the Sharpe index I found this old article http://www.long-short.pro/post/rajiv-sethi-risk-i-voznagrazhdenie-v-vysokochastotnoy-torgovle-73


Out of more than 30,000 accounts, according to the article, only 31 fit this description. But these traders dominate the market, picking up 47% of all trading volume and appearing on one or both trading sides in nearly 75% of the contracts sold/bought. And they do so with minimal directional action: the average daily portfolio is only 2% of trade volume, and the overnight portfolio of the average HFT trader is always zero.

...

HFT traders get above average returns to the fraction of risk they take on themselves. This is true in general and for every type of trader. In general, the average Sharpe ratio for HFTs, recalculated for the year, is 9.2. For subcategories: aggressive HFT traders (8.46) show the lowest risk-adjusted returns, while passive HFT traders do slightly better (8.56) and mixed firms achieve the best results (10.46). The spread is large, with an interquartile latitude of 2.23 to 13.89 for all HFT traders. Nevertheless, even the bottom return/risk for HFTs is seven times higher than the Sharpe ratio for the S&P 500 (0.31).

Rajiv Sethi: Риск и вознаграждение в высокочастотной торговле
  • www.long-short.pro
Известная статья о доходности высокочастотных трейдеров в последнее время привлекает изрядное внимание средств массовой информации. Среди авторов статьи - Андрей Кириленко (Andrei Kirilenko) из Комиссии по фьючерсной биржевой торговле (Commodity Futures Trading Commission, CFTC), который при изучении «мгновенного обвала 2010 года» применял...
 
pantural:

resample - as an obvious way to falsify the algorithm SR

I think it's time to change the algorithm quickly before anyone notices))))

IMHO - what is more good for the portfolio may be less good for an individual TS. Usually the TS has clearly defined moments of "deal closing", by which the Sharpe is calculated. The portfolios and assets do not have such distinct moments, so they can be chosen arbitrarily and not to suffer from this arbitrariness.

 
Sharpe Ratio for Algorithmic Trading Performance Measurement | QuantStart
Sharpe Ratio for Algorithmic Trading Performance Measurement | QuantStart
  • www.quantstart.com
When carrying out an algorithmic trading strategy it is tempting to consider the annualised return as the most useful performance metric. However, there are many flaws with using this measure in isolation. The calculation of returns for certain strategies is not completely straightforward. This is especially true for strategies that aren't...
 
Sharpe Ratio and Its Applications in Algorithmic Trading
Sharpe Ratio and Its Applications in Algorithmic Trading
  • www.quantinsti.com
Sharpe ratio is the ratio of the excess expected return of an investment (over risk-free rate) per unit of volatility or standard deviation. It is a measure for calculating risk-adjusted return.
 
Aleksey Nikolayev:

IMHO - what is more good for the portfolio may be less good for an individual TS. Usually the TS has clearly defined moments of "deal closing", by which the Sharpe is calculated. Portfolios and assets do not have such distinct moments, so they can be chosen arbitrarily and not to suffer from this arbitrariness.

What portfolio are you calculating there? At speculative market prices? It is necessary for once to take care to analyze the statements of issuers and compare the ratio of net assets to the number of shares (book value) with the market - speculative value. When calculating all sorts of ratios nobody, for some reason, takes into account the book value of the stock, and after all this is the basis for calculating risks - if a speculative price is traded near the book value - there is potential for growth. If it is overvalued - such shares should be discounted, but again, we look at the dynamics of reporting - if the issuer earns from ordinary activities, not from financial ones, the hell with it, such a share can be held for a while. It is necessary to take the balance price and find it on the chart of a trading instrument - put there a horizontal line and decide on the potential for growth / decline of speculative prices of the stock based on these volatilities. You are digging through garbage (in the history of speculative quotes, and even adjusted for you by your broker) and want to make a grail out of "g". Six months ago Maxim said that all this is a fudge - an advisor to educate on history. They agreed that these very expectations for growth or decline can only be predicted by a fundamental analysis of the sensitivity of trading instruments to certain indicators. Half a year has passed, and they are still pouring water about some forests and predictors, but the faces change, and the cart is still there - upside down in the river by the bank around the corner, with the wheels swapped out... there. I am disappointed.

 
A good EA has only one indicator: it earned or lost. If neither - it's a bad EA. Just open the chart - we see complete gibberish, and this is what you want to teach the EA? I want to come over here and use batons to disperse all of you.)
 
geratdc_:
A good EA has only one indicator: it has earned or lost. If it does neither, then it is a bad EA. If you open the chart, you see complete rubbish, and this is what you want to teach your EA? I want to come over here and use police batons to disperse all of you.)

I think it would be nice to have a head for any EA. Optimization is never redundant.

 
Aleksey Nikolayev:

What you are referring to is more accurately called an "Annualised Sharpe" and the "Sharpe Ratio" is exactly as it is now in MT.

For strategy performance measurement, as an industry standard, "Sharpe Ratio" is usually quoted as "annualised Sharpe" which is calculated based on the trading period for which the returns are measured.

I would not say "exactly like that", the formula itself is correct, but it should be calculated not by returns from trades, but by daily (hourly, etc.) returns. Otherwise if this number is calculated with help of trades and their significantly different amount, it is not important, for example one strategy has 0,01 Sharpe and the other has 5, it is not clear, which one is better or worse, only its sign (higher or lower than zero Sharpe) is important.)

So although pantural isn't exactly talking about classic Sharpe Ratio but still he raised an important question about it. But I personally don't prefer using the Sharpe Ratio, I prefer the ratio of profit to the maximal drawdown, as a measure of strategy performance.