Multiple Instances & variable Max Loss

[Mickson]

Hi All,

I am really a newbie and hope not to repeat questions already covered
(true I have only skimmed through this site). I am busy making my way
through the Leverage Space Trading book and I want to compliment Ralph
on his story telling prowess. I wish to add that so many of the topics
discussed in the book cover the very questions I have wrestled with
over the last 10yrs as a fund manager/trader.

By way of background I run a long short equity fund, so here is my
question.

The way I believe I will be applying my funds return time series to
the LS model is simply with the funds net returns. I trade in a highly
correlated sector so that breaking my longs and shorts into multiple
independent instances I fear will not be a correct way to apply the
model. Lets just assume I am market neutral, I typically don't trade
with a stop loss on individual positions as when I am losing on my
longs I am hopefully making on my shorts. Therefore I typically net
everything off and just look at effectively a single instance (of
course during maintenance review I weed out relative under-performers
on both long and short side). I wonder if others have worked a way of
dealing with this type of situation in practice with the LS model. I
understand that one of the beauties with this system as opposed to
mean variance analysis is that it doesn't rely on correlation which is
subject to "miss" at crucial points. In my case I trade within only
one sector so correlation is a reality in my world, although I will
agree on a day to day trading basis correlations even though high are
unreliable and can expose the PnL to unexpected drawdown. The appeal
to me for application at the multiple instance level would be to help
me correctly size the individual positions, i.e. I typically have
about 40 positions open, not necessarily 20 long and 20 short with = $
amounts. My composition is somewhat arbitrary but usually influenced
by looking at my net beta exposure.

Next, a pressing question which may relate to my not fully
understanding the model yet.

Typically I have in mind a manual stop loss of a daily discomfort
level of a 2% DD. This is not to say that I would close out the fund
or trim positions if these level were breached, but based on my
overall objective of managing the fund I try and design the VaR of the
fund to not exceed 2%. On 98% of my more than a year of trades I have
not breached the daily DD of -2% and in the 2% of times I have it has
only been marginal. What I am leading to here is that there are times
when I feel we are in a more volatile space and therefore dial down my
total DD daily exposure so I might for a month or two only carry
exposure that would be risking about a 1% daily DD.

In this approach I have resorted to treating each day as a trade. In
conclusion I am wondering if the LS model is applicable to my style of
trading whereby I define financial ruin as a MAX DD of 15% over the
course of a year or lifetime (I think I read the length of "play" has
an effect on f.) but I wish to trade within the design of a max 2% per
day DD. I am probably complicating things with this 2% daily DD as it
could be argued it is the one side of each bet as the amount being
risked for the mean return. Which makes me start to wonder if all
these variable bets effects the optimal F. If this model will assist
me with allocating my leverage then I feel like have found my way
home :) I know I have asked many general questions. I guess I just
want some confirmation that my questions can all be answered and it is
worth the time to invest in getting a full grasp of the model.

Regards

Mick

[Brady Preston]

Mick welcome to the group.

I believe that it may help you to define trades as a period. Depending on
the type of fund your period could be monthly, daily ect... This would
remove the idea of long or short and leave you with positive or negative
equity for that given period. I see that you have stated a trade is equal to
a day. The LSP model is built on a 1 correlation assumption, assuming that
the markets will hit a 1 correlation at some period in time.

How many markets are you trading? A market can be defined in many different
ways.

If I understanding you what you would be looking for is not called max
drawdown but largest loss. You could define the criteria for a given largest
losing trade(largest losing day). Restricting the largest losing day to 2%
is easier then restricting multiple trades(days) to 15%. F moves closer to
the real optimal f as the horizon increases. Optimal f will start at 1 and
move to the left.

I could run you an example if you sent me or some data.


Brady Preston

[Michael Berman]

Hi Brady.

Thanks so much for responding. I will forward you some Data in a couple of hours.

Regards,
Mike

Sent from my iPhone