Machine learning in trading: theory, models, practice and algo-trading - page 159

New comment
 
Alexey Burnakov:

Why should I spell out the obvious? Well, I can open a trade with a large volume, so what? We're not in kindergarten here. It is better to admit that you do not understand what we are talking about.

You better ask Dmitry, with what intention was the question asked:

Probably he wanted to hear "1:500" and laugh. If he had a leverage of Ltd. But he does not earn much. What would it mean if I told him that a broker gives me a 1:1000? That I can open a deal for a few lots at once?
Alexey Burnakov:

You about Thomas, you about Yeremma.

What difference does it make in practice what kind of a mega-big position I can open, if, as Dr. says - and I totally agree with him - in backtests you pick a (usually small) lot so the max drawdown isn't reaching dangerous levels?

Which answer would you prefer: 1:10 or 1:1000? Will it change your mindset? Do you want to get a private consulting on brokerage companies that give 1:1000 and more?

On the contrary. I think that Dimitry asked about the 30% per annum, suggesting that you are using a very small leverage such as 1:1 or 1:10 that makes it impossible to get a high return on deposit.

 
Andrey Dik:

On the contrary. I think that Dmitri was asking for 30% per annum, suggesting that you are using very small leverage, something like 1:1 or 1:10, which makes it impossible to get a high yield relative to the deposit.

Dmitry clearly wrote that his 30-40% per annum should be divided by 10, which would lead to the interest of hedge funds.

Or else kids are sitting there genuinely bewildered, why they are not invited to a hedge fund to work.....

 

I read my broker's terms and conditions. Leverage 1:100 up to a balance of 250,000 eur. I'm okay with that.

Dmitry, you trade in a hedge fund and you are forced to 1:1? I think that they have a minimum of leverage.

Dmitry:

And the size of the lot for which you open depends on the leverage?

And the price of a pip depends on the size of the lot?

This is child's play.

I described my money management, now I'm interested in yours. Suppose you have two accounts - in both of them 10000 usd, but the first account is leveraged at 1:100, the second - 1:1000. How would you calculate the lot size for these accounts? Or you do not calculate it, but just use all available funds?

 
Dr.Trader:

Purely interesting, judging by the articles I always thought that they have at least minimal leverage, but still have.


1:2, 1:4. 1:10 max.
 
Dmitry:

Okay, if you don't understand elementary things, there's no point in arguing.

In short, to bring your percentages to the hedge fund percentages you have to divide them by about 10.

Now I understand the logic of the question. I looked through this comment...

I have 1:2 leverage on one instrument (one trade in the market). And when I talk about 30% a year (with a max drawdown on a 25% history) as the maximum I see on my models, it's on this particular leverage and one instrument. I think hedge funds give that kind of leverage.

When comparing to a hedge fund, the issue has to do with the non-linear increase in FS (and Sharpe) with increasing number of instruments and linear increase in profitability.

Here's a brief outline of the logic. Let's conservatively assume that I'm estimating a yearly return on one instrument of 20% with a 25% drawdown (constant lot). That gives FS = 8 over 10 years. This is the top of the table. Further with increasing the number of traded instruments the calculation is given for the account with increased leverage and the same leverage of 1:2. The result at the bottom right of the table is essentially an analogy for a hedge fund (or when there is a lot of money, but leverage is limited).


You can see that with an ideal portfolio (when trades don't correlate and each instrument has the same returns and drawdown), it turns out that increasing funds and leverage lead to the same profitability (20% per year for a small account with one instrument, as for a large one with 5), but to a much larger FS. In fact the FS 18 is off the scale and unlikely to be obtained realistically, but at the same time due to limited leverage there is no possibility to increase risks and profitability with acceptable drawdown (say, up to 25%).

The bottom left part of the table is showing what I can ideally use in my models, taking 56% of drawdown as acceptable drawdown and using increased leverage.

 
Alexey Burnakov: The left bottom of the table is what I can squeeze out of my models ideally...
All (models) is done in pips (fix lot) + spread + slippage, then screwed mm...
 
mytarmailS:

Everything works fine, the volume is summed up for the prices, but there is one problem, as the chart is not a tick chart but a 5-minute chart and the 5-minute chart has not one price but a range of price borders (high, low)... So, I should first divide the price into certain mini ranges of say 20 points in size and when the price gets within this mini range, I sum up the volume in it...

volume <- sample.int(101,size=100,replace=TRUE)
priceH <- round(cumsum(rnorm(100))+1000,0)        #High
 priceL <- priceH - round(cumsum(rnorm(100))+10,0)  #Low

 price_vol_df <- as.data.frame(matrix(nrow=0, ncol=2))
for(i in 1:100){
        price_levels <- seq(priceL[i], priceH[i], by=0.1) 
        for(j in price_levels){
                price_vol_df <- rbind(price_vol_df, c(j, volume[i]/length(price_levels)))
        }
}
colnames(price_vol_df) <- c("price", "volume")

volume.profile <- aggregate(price_vol_df$volume ~ price_vol_df$price, sum, data=price_vol_df)
volume.profile
plot(volume.profile,t="h" , lwd=5)

For each bar I build a vectorprice_levels from minimum to maximum price, in steps(by=0.1). And volume for each step - volume of bar/number of steps. I save it all to a table. And so on, just as it was.

very bicyclic :) prices in price_vol_df should definitely be rounded to the nearest values.

 
Dr.Trader:

For each bar I build a vectorprice_levels from minimum to maximum price, in steps(by=0.1). And volume for each step - volume of bar/number of steps. I save it all to a table. And so on, just as it was.

very bicyclic :) prices in price_vol_df should definitely be rounded to the nearest values

Thanks a lot, but the script does not work the way I thought, the levels are even less than with the first method....

I understand that I should not be tied to high prices but to do something like this

but just round the price scale, like now we have a min. move of 1 pip and we do a min. move of say 20 pips, but in every move of 20 pips there is a sum of volume that has passed within these 20 pips..... I would better draw it, I can't understand anything I've written

here's the link to the figure.http://prntscr.com/ct8kgg

I tried it 10 times

Скриншот
Скриншот
  • prnt.sc
Снято с помощью Lightshot
 
Vizard_:
Everything (models) is done in pips (fix.lot) + spread + slippage, then screwed mm...
This is without MM for purity.
 
Mihail Marchukajtes:
Unfortunately this is a failure, pairs walk on their own, entropy is another matter, this will be more interesting.
Do you also work in a hedge fund and you deliberately confuse people? What do you mean "by themselves"? Where have you seen that? And about entropy it seems that you do not understand anything at all, for the sake of science you just said, it is no more useful than the waving machine, you must have been hypnotized by the logarithm....
1...152153154155156157158159160161162163164165166...3496
New comment
Reason: