Compounding

This topic is empty.

Viewing 12 posts - 1 through 12 (of 12 total)

Author

Posts
August 9, 2018 at 9:29 am #101833

DanielBaeumler
Member

A main contributor to the outstanding profits we all see in our backtests are due to compounding. In order to realize this, my approach has been so far:

Momentum system: Capital adjusted on monthly basis based on P/L from previour month. E.g. if I replace 2 positions in the portfolio, position size is calculated as: (Value of closing positions) / 2

TF system (Nick’s growth portfolio): 70% invested. At this point, I allocate equal amounts to each new open position (5% of equity). After the portfolio is fully loaded, it’s a matter of position replacement: (Capital realized from closed positions) / number of available new positions. e.g. if 3 positions will be closed, I’ll divide the realized capital by 3 for the next new buys to come.

MR/MOC: Capital adjustment on monthly basis based on P/L from previous month

I believe the compounding effect can be best/easiest realized for momentum and TF systems whereas my simplified compounding approach for MR/MOC is not matching how AB is processing it. This might by an additional reason why returns from short terms systems will not meet the backtest results (apart from all the other issues of no-fills and sk´lippage).

I would curious to hear how others realize/handle compounding.

Cheers. Daniel

August 9, 2018 at 7:43 pm #108976

SaidBitar
Member

I do it exactly as AB does it
so long term first
you have 20 positions fully invested and you sold 2 positions
the new capital is calculated like this = initial investment of the 18 open positions + the close value of the closed positions
the total i divide by 20 and this is my position size
the only issue with this that sometimes you will have some cash in the account i mean not 100% invested but 96% but the moment you start closing and opening positions you will reach 100%.
for short term every day i use the close account value from the previous day
for example if after the close yesterday the portfolio value was 50 i will use 50 as the capital for the position value for the trades of today and so on.

August 9, 2018 at 9:52 pm #108977

Nick Radge
Keymaster

Quote:

MR/MOC: Capital adjustment on monthly basis based on P/L from previous month

I recalculate every day

August 9, 2018 at 11:48 pm #108978

LEONARDZIR
Participant

Said,

I assume using your method is appropriate for a rotational momentum system with fixed position size. If you are doing volatility based position sizing it seems to me the position size for the new entry should be corrected for volatility.
This also raises the question of how amibroker is testing rotational momentum on fixed and volatility based portfolios. I notice in backtesting a portfolio of fixed position sizing the only time equal amounts of money are given to each position is at the start date. Amibroker( at least the way I have it coded) doesn’t ever seem to rebalance either monthly or upon new exits and entries monthly. I still haven’t figured out how amibroker arrives at the new entry position size. I have basically done what Daniel does with my rotational momentum system with fixed position sizes although I am not sure that is what Amibroker is testing.
I find a similar situation with risk parity sizing. Amibroker distributes money based on risk parity at the start date. It does not seem to rebalance at all. I still haven’t figured out the right amount of money to put in a new position. On a new position or positions I do look at what should be the right number of shares to enter based on risk parity but not infrequently it requires more money than I have available in the system. If that is the case than I just put the proceeds based on risk parity in the highest ranked stocks and go down the list until I run out of money. The problem with this is that over time my position sizes and sometimes the stock selection becomes out of whack with what Amibroker is testing. This has made me nervous that I might not be duplicating Amibrokers testing results in real trading.
Curious about your thoughts.
Len

August 10, 2018 at 12:34 am #108979

Stephen James
Member

Amibroker does not rebalance the positions on a monthly rotational system unless you add that to the system with CBT code – example here:-Rebalance

Without that it retains the original position as per the position size formula until an exit. Any new positions use same and will enter if enough capital allows. Position size shrinking and minimum position size are options available using the SetOption function or analyser settings to use all available cash.

August 10, 2018 at 1:04 am #108980

LEONARDZIR
Participant

Craig,
What do you mean by any new positions use “same”. Does that mean that in a fixed position mode the money from exits is divided equally by the number of new positions to determine new position size(s)? If only one position then all money from prior exit is used to determine new position size.

If that is case in risk parity model you take money from prior exit and adjust position size for new entry based on volatility? Or do you just use all the money for new entry position regardless of volatility?

August 10, 2018 at 7:45 am #108981

DanielBaeumler
Member

Len,
regarding risk based position sizing on momentum systems, I’m also struggling matching the positions sizes of AB backtests.

In order to address this, I’m rebalancing my portfolio on the 2nd instead of the 1st day in the month and then using the position sizes AB is giving me in the backtest. I’m running a backtest going 1 year back in order to consider positions which are holded several months.

August 10, 2018 at 11:21 am #108983

LEONARDZIR
Participant

Daniel,
I think that is a good idea. I had considered the same thing. I was going to wait to see what amibroker was holding. I was considering converting the value of each holding in amibroker to a percentage and use the percentage of the total portfolio for my new entries.
Clenow also suggested=summing up the inverse volatility of all the holdings and and dividing each indiviual inverse volatility by the total volatility to give a percent which can be used for position sizing. I don’t think either method is exactly what amibroker is testing but should be sufficient.

August 10, 2018 at 3:21 pm #108984

SaidBitar
Member

Actually my rotational systems are volatility based position sizing and honestly they are the ones that take a lot of time to check the new positions size.
here is what i am doing
assume i need to sell Stock A that has 12% of the portfolio and i need to Buy Stock B and as per Amibroker the position should be 40% of my portfolio in this case i assign only 12% this is because in my backtest i enabled allow position shrinking.

here are some stuff that will help you understand what AB is doing and how it calculates the position size.

then you run your backtest and you will have results like this.

if you read it you will understand how amibroker works on day to day basis for this system

August 10, 2018 at 11:16 pm #108985

LEONARDZIR
Participant

Said,
That is very helpful information.
I like Daniel’s idea of waiting until the second day to see what amibroker has exited and entered. I also like the idea of calculating the percentage of each new entry of the whole portfolio and using that percentage for position sizing. It seems a relatively simple way to emulate the backtesting results in your actual trading.

August 11, 2018 at 5:58 am #108988

JulianCohen
Participant

For volatility sizing systems I do the following.

I input the value of my open and closed profits into the account size and run Amibroker exploration to give me the top shares.

In Excel I then change the number of shares for positions that I already hold. I calculate out a value for each position of number of shares * close and then use this total sum to work out the amount of the new shares I should be buying.

I then buy from the top of the list until I run out of money.

August 13, 2018 at 9:08 pm #108989

LEONARDZIR
Participant

This discussion made me look at the backtest on my Nasdaq aggressive momo over a 10 year period. This system has been very profitable for me this year. The original intent was to have equal risk for each position using Clenow’s ATR model. This is accomplished at the start date of testing but after 10 years is problematic because there is portfolio rebalancing but no position size rebalancing. So at the end of 10 years the risks for each position are nowhere near equal and because it has a small
number of stocks with relatively low turnover there is a very significant investment in only a few stocks.
I also tested the system with fixed position sizing at the start of the test and there is less concentration in just a few stocks.
I have come to the conclusion that it will be important to backtest my models with rebalancing (both fixed and risk parity).If I were going to guess I would think the CAGR would be similar between rebalancing and no rebalancing but the drawdowns would be less. But I don’t know that unless I test.

number of stocks
Author

Posts