D-VaR Position Sizing by David Varadi
Author: Eugene
Creation Date: 2/20/2010 11:33 AM
profile picture

Eugene

#1
I'm setting up to include this new money management technique, among many other improvements, in the upcoming release of MS123.PosSizers:

Introduction to D-VaR Position Sizing (Part 1)

QUOTE:
In this method we will use an incredibly simple approach:
1) take the daily returns for a given stock, index or strategy
2) compute the 5th percentile of returns (max tail loss)
3) select a budgeted risk level as a maximum daily loss such as 1% (conservative) or 1.5% (aggressive)
4) your position size is the budgeted risk level divided by the absolute value of the max tail loss
5) this position may not exceed 200%


Please help me out with the following: in this context, should max tail loss be treated separately for long and for short trades?

For example, the distribution of daily changes for 50 days can look like this: [-10.5, -5, -4.9, ... 0.25, 0.25, 3]. For a long position, the negative numbers represent max tail loss. Conversely, for a short trade, positive numbers represent the risk of price going in the opposite direction. Otherwise it doesn't seem reasonable to even look at the worst negative daily returns for a short position, because the profits are to be made off these moves.

Hope it makes sense.
profile picture

dansmo

#2
Why not make it user´s choice and provide it as an option?
profile picture

Eugene

#3
I treat the omission of such an obvious point as the author's probable long-side bias. That's why I've raised this question.

There are many options in the PosSizers, and soon, when v2010.03 comes out, there will be even more. So, generally it's the less options the better - for simplification. However, your point is good and I'll look at adding this option.
profile picture

dansmo

#4
I dont think it is because he has a long bias. I assume most of the VaR models look at confidence levels on absolute percentage changes. I dont know the author at all, but my guess would be that he just used what was widely done when thinking of VaR.
profile picture

Eugene

#5
Thank you.
profile picture

avishn

#6
Does MS123 implementation take open positions into account when calculating max tail loss?
profile picture

Eugene

#7
Max tail loss is calculated by taking the daily returns of the stock (i.e. "Period to average returns").
profile picture

avishn

#8
I'm sorry, not sure I follow... My understanding is that VaR applied to a strategy is essentially a max tail loss of the equity curve, no?
profile picture

Eugene

#9
According to author's definition above, to calculate DVaR we "take the daily returns for a given stock, index or strategy". Hence, the PosSizer takes the N-bar returns for a given security.
profile picture

avishn

#10
Could you please explain how it works for multi-symbol strategy? Which symbol returns are used?

profile picture

Eugene

#11
It's the current symbol's returns, of course. Technically speaking, currentPos.Bars in the PosSizer's SizePosition method.
profile picture

avishn

#12
Thank you. It was a dumb question. I got stuck on thinking about the equity curve's tail loss.

Should I see any skipped trades when using this position sizer while backtesting with a single symbol?

The following on SPY 5 years, D-VaR w/ separate long/short, skips about a quarter of all trades --

CODE:
Please log in to see this code.


Is it the same issue as we discussed in the other thread or I'm doing something wrong?
profile picture

Eugene

#13
Wealth-Lab is skipping trades during these long periods of time because the size, as dictated by the formula, is too large - it exceeds the equity value. For example, during 2006 the size stays at ~120% for months.

Although it may look odd, while debugging the PosSizer I saw nothing wrong about it. I manually verified the percentile and budgeted loss values to ensure they're correct.

To prove it, try to cap position size to, say, 90% of equity - you shouldn't notice skipped trades.

Actually, it's because this PosSizer is very risky -- even Mr. Varadi stresses that:
QUOTE:
5) this position may not exceed 200%