[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [amibroker] Re: PositionSize / Capital



PureBytes Links

Trading Reference Links

Christopher:

One thing you have to keep in mind about statistics like K-ratio, standard error, and linregslope: they are based on a LINEAR regression of the equity curve, and if your CAR triples, that means that the equity curve is becoming more and more exponential in shape, due to compounding. Therefore, K-ratio and standard error are totally useless, since they assume a LINEAR equity curve. If you simply examine a plot of your equity curve with a linear regression line drawn through it, you will see what I mean. The regression is worthless in that case unless you convert the data to logarithmic format in the AA report, as Fred has been requesting, or do a more complicated non-linear regression.

Al Venosa

Christoper wrote:

Jumping in here to make one observation...

I've seen much higher returns as well when NOT using volatility based
position sizes.  But other stats/metrics really take a plunge.

For example - CAR triples while Sharpe becomes 1/5th of previous value
and K-Ratio drops to 0.04, from 0.10.  Other things like, Max Sys DD
doesn't change that much...

That's just my expereince... do other people see these type of results
as well?  How much weight can we place on K-Ratio and drop in Sharpe?

--- In amibroker@xxxxxxxxxxxxxxx, "danielwardadams"
<danielwardadams@xxxx> wrote:
>
> Al & Anthony,
> I've also seen the lower returns for volatility based versus equal
> equity position sizing in the past and didn't know what to do about
> it (assuming I wanted more positions for more diversification).
>
> I'm not sure how one would code it in .AFL, but would the following
> represent a reasonable compromise?
>
> (1) Start with an equal equity based model based on, say,  5
> positions (position size = -20). So each part of the pie equals 20%
> of total equity.
> (2) Determine actual position size within each piece of the pie based
> on volatility based sizing. So, depending on your risk parameter, one
> might use only 17% of one piece of the pie, 13% of another piece, and
> 20%, 8%, and 11% of the other pieces.
> (3) Sum the used portions of the pie (in this case 17+13+20+8+11 =
> 69%) and see what you have left. 31% in case.
> (4) Allocate the remaining cash according to the equal equity model.
> This means you get one more 20% piece of pie and only have 11% cash
> remaining.
> (5) Apply the above using your ATR based position sizing recursively
> until your cash is minimized. So if you only are able to use 9% of
> the piece of pie left in (4) you take the 11% left from that piece
> plus the 11% cash and you have 22% -- enough for another position. So
> in this case you end up with 7 positions and only 2% left in cash.
> So your cash is minimized and all your positions adhere to the ATR
> based position sizing.
>
> Like I say, I have no idea how to code it but intuitively it makes
> sense to me.
>
> Thoughts/comments?
>
> Dan
>
> (And, yes, I'm sure I'm not the first person to think of it so my
> apologies to those who have gone before).
>
> --- In amibroker@xxxxxxxxxxxxxxx, "Anthony Faragasso" <ajf1111@xxxx>
> wrote:
> > Hello Al,
> >
> > You stated:
> >
> > "the lower the volatility, the lower the risk and therefore, the
> smaller the positionsize for that stock. "
> >
> > Is this a correct assumption ? ...Would you want a larger
> positionsize on a less risk position , and a smaller position on a
> more volatile one ?
> >
> > Anthony
> >   ----- Original Message -----
> >   From: Al Venosa
> >   To: amibroker@xxxxxxxxxxxxxxx
> >   Sent: Saturday, December 11, 2004 7:53 AM
> >   Subject: Re: [amibroker] PositionSize / Capital
> >
> >
> >   Ed,
> >
> >   I, too, have confirmed many times with backtesting what you
> report, viz,, that positionsize = -x gives better performance results
> than using volatility-based MM positionsizing. The non-MM code I've
> used in the past is:
> >
> >   posqty = Optimize("posqty",5,2,10,1); // no. of stocks active at
> any given time
> >   PositionSize = -100/posqty; //equal equity model
> >
> >   I think I know what the problem is, but I have not as yet figured
> out how to solve the problem with AFL. If you use the MM-based
> positionsize statement as we have discussed (equal volatility model),
> i.e., PositionSize = -1 * C/StopAmt, and examine the tradelist, you
> will likely discover that, often, not all 5 stocks are active all the
> time. In other words, either you have idle capital earning nothing or
> you have fewer active stocks than you want. Why is this? Because some
> stocks, which might not be as volatilie as others, use up more of
> your capital to initiate a position than a more volatile stock.
> Consequently, your capital is used up before you have a chance to
> enter into your 4th or 5th stock. Instead of having 5 open positions,
> you might only have 3 because of this. Checking positionsize
> shrinking doesn't help because you'll discover you might have tiny
> positions in your 5th stock. The fewer stocks you have, the less
> diversified you are, and therefore the more risky your portfolio. The
> more risk, the higher the DDs. This problem cannot happen with the
> equal equity model since all positions are equal in size, by
> definition.
> >
> >   One possible way around this might be to increase your margin so
> that equity is expanded enough to allow full funding of all
> positions. But, again, this also increases your risk. Another way
> might be dynamically setting your risk to fit the volatility of each
> stock individually (the lower the volatility, the lower the risk and
> therefore, the smaller the positionsize for that stock). However,
> this changes your model so that you no longer have equal
> volatility/equal risk (getting closer to the equal equity model). So,
> the problem remains unsolved for the moment. I have not had time to
> devote to cracking this problem yet, but some day I hope to do this.
> If you have any ideas, I'm all ears.
> >
> >   Al Venosa
> >
> >
> >   ed nl wrote:
> >     Thanks for your effort Al. It is very clear,
> >
> >     In one of my earlier posts I posted
> >
> >     // money management block
> >     stopLoss = Ref(bbb*ATR(20),-1);
> >     // trade risk
> >     tr = IIf(Buy,(stopLoss / BuyPrice),stopLoss / (ShortPrice +
> stopLoss));
> >     // renormalisation coefficient
> >     rc = 0.02 / tr;
> >     // positionsize
> >     PositionSize = rc * -100
> >
> >
> >     it actually gives the same result as your:
> >     PositionSize = -2.0 * IIf(Buy,BuyPrice,ShortPrice) / stopLoss
> >     except for short positions. Exact the same it would be if I
> use: tr = IIf(Buy,(stopLoss / BuyPrice),stopLoss / (ShortPrice));
> >
> >     Unfortunatelly I do not get better results this way. Using just
> a simple PositionSize = -10 still gives somewhat better results.
> >
> >
> >
> >     rgds, Ed
> >
> >
> >       ----- Original Message -----
> >       From: Al Venosa
> >       To: amibroker@xxxxxxxxxxxxxxx
> >       Sent: Saturday, December 11, 2004 4:19 AM
> >       Subject: Re: [amibroker] PositionSize / Capital
> >
> >
> >       ed nl wrote:
> >
> >         Al,
> >
> >         but how do you implement the risk factor now?
> >
> >         ed
> >       Ed:
> >
> >       Let us suppose you have established your risk as 1% (i.e.,
> the maximum you are willing to lose on a trade). Let us also suppose
> your initial equity is $100,000. So, if the stock you buy (or short)
> goes down by the amount based on your system, you lose only $1000,
> keeping you in the game. Now, let us say you defined your volatillty-
> based stop in terms of 2*ATR(20), which you incorrectly assigned to
> the variable TrailStopAmount. I say 'incorrectly' because the
> TrailStop in AB was designed to mimic the Chandelier exit, which is
> basically a profit target type of stock (it hangs down like a
> chandelier from the highest high since the trade was initiated, if
> long). I don't think you want the TrailStop to be your money
> management stop. Rather, the MM stop is the max stoploss, defined as:
> >
> >       StopAmt = 2*ATR(20);
> >       ApplyStop(0,2,StopAmt,1);
> >
> >       So, if your stock declines by 2*ATR(20) from your entry, you
> exit with a 1% loss. Let's take an example. Stock A is selling for
> $40/share. It's ATR(20) is $1/shr or 2.5% of 40. Your stop amount is
> 2*ATR(20), which is $2/shr. How much stock do you buy? You simply
> divide your risk, $1000, by 2*1, which is 500 shares. This amounts to
> an investment of $40/shr * 500 shrs or $20,000. All of this can be
> coded in one simple line of AFL plus the 2 lines above defining the
> MM stoploss:
> >
> >       PositionSize = -1 * BuyPrice/StopAmt;
> >
> >       where -1 is 1% of current equity (0.01 * 100,000 or $1000),
> BuyPrice = $40/shr, and StopAmt is 2. Keep in mind that a negative
> sign means 1% of CURRENT equity, which means compounded equity, not
> just a constant initial equity of $100,000. If you carry through the
> above math with your renormalization coefficient notation, you wind
> up with the exact same answer.
> >
> >       One more thing. When you place your order, assuming you are
> trading with EOD data, you do not know what the buyprice is until you
> buy the stock, which is the next day. So, what most traders do is
> base their positionsize on the closing price of the night before the
> entry. Therefore, to place an order in the evening to be filled in
> the morning at the open, your positionsize statement would actually
> be:
> >
> >       PositionSize = -1 * C/StopAmt;
> >
> >       where C is the closing price on the night before you buy. So,
> if you use the code SetTradeDelays(1,1,1,1), then the above formula
> is OK. However, if you use SetTradeDelays(0,0,0,0), then you have to
> ref the C back a day.
> >
> >       This is probably more information than you were asking about,
> but I hope it helps.
> >
> >       Cheers,
> >
> >       Al Venosa
> >
> >
> >
> >
> >   Check AmiBroker web page at:
> >   http://www.amibroker.com/
> >
> >   Check group FAQ at:
> http://groups.yahoo.com/group/amibroker/files/groupfaq.html
> >
> >
> >         Yahoo! Groups Sponsor
> >               ADVERTISEMENT
> >             
> >       
> >       
> >
> >
> > --------------------------------------------------------------------
> ----------
> >   Yahoo! Groups Links
> >
> >     a.. To visit your group on the web, go to:
> >     http://groups.yahoo.com/group/amibroker/
> >      
> >     b.. To unsubscribe from this group, send an email to:
> >     amibroker-unsubscribe@xxxxxxxxxxxxxxx
> >      
> >     c.. Your use of Yahoo! Groups is subject to the Yahoo! Terms of
> Service.





Check AmiBroker web page at:
http://www.amibroker.com/

Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html





Check AmiBroker web page at:
http://www.amibroker.com/

Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html



Yahoo! Groups Sponsor
ADVERTISEMENT
click here


Yahoo! Groups Links