Drawdowns

nizar · 16 July 2007

I once asked a mechanical trader if he uses all time high close as the entry for his system, as its a well-known method of entry that works. He replied "No, as its a fairly high drawdown system".

Is drawdown a function of entry, or exit, or both, or the system as a whole?

Some discussion/comments would be much appreciated.

It's Snake Pliskin · 16 July 2007

nizar said:
I once asked a mechanical trader if he uses all time high close as the entry for his system, as its a well-known method of entry that works. He replied "No, as its a fairly high drawdown system".

Is drawdown a function of entry, or exit, or both, or the system as a whole?

Some discussion/comments would be much appreciated.

I certainly believe it is a function of entry. But more so system design.

If you are entering with a new low breakout for a long system you may have drawdowns too.

nizar · 16 July 2007

It's Snake Pliskin said:
I certainly believe it is a function of entry. But more so system design.

If you are entering with a new low breakout for a long system you may have drawdowns too.

Snake can you please elaborate on the bold part.

And also what do you mean by new low breakout?

Thanks.

Nick Radge · 16 July 2007

I think it associated with the complete system, more specifically with the win rate, the win/loss ratio and in turn the mathematical probability of a losing streak.

The win percentage and win/loss ratio are inherent in any trading/investing strategy so if we take those then work for the losing streak we could get some kind of probability on drawdown size.

Therefore,

Losing Streak = ((log s)/-log((1-n)))

Where,

n = win rate
s = trade sample

Assume,

Win rate of 50%,
Trade sample of 10,000

Losing streak = 13

Using a standard risk of 2% per trade on the above will give a theoretical maximum drawdown of 23%.

nizar · 16 July 2007

Nick Radge said:
I think it associated with the complete system, more specifically with the win rate, the win/loss ratio and in turn the mathematical probability of a losing streak.

The win percentage and win/loss ratio are inherent in any trading/investing strategy so if we take those then work for the losing streak we could get some kind of probability on drawdown size.

Therefore,

Losing Streak = ((log s)/-log((1-n)))

Where,

n = win rate
s = trade sample

Assume,

Win rate of 50%,
Trade sample of 10,000

Losing streak = 13

Using a standard risk of 2% per trade on the above will give a theoretical maximum drawdown of 23%.

Thanks for that explanation Nick.

You mean 26 % (2%*13) yeh??

Nick Radge · 16 July 2007

No. Think it through carefully.

brettc4 · 16 July 2007

I will assume it is only 23% as each time you lose, your next 2% stop loss is slightly less in value than the preceeding one, and so on.

Correct? Please tell me thats correct.

GreatPig · 16 July 2007

A loss of 2% leaves 98% remaining. Do that 13 times and you have 0.98^13 = 0.769 ('^' means to the power of), close enough to 0.77.

So if you have 77% left, you've lost 23%.

GP

Nick Radge · 16 July 2007

Thats correct. Fixed Fractional allows smaller risk to be taken as you lose. In fact to lose 100% of your equity you'd need something like 140 consecutive losses, not 50.

We can also turn the equation around and say, "how much risk per trade can I use to lose n amount during a drawdown?" This is a good exercise because every one has different maxDD tolerances so if two people are trading the same system and one can withstand a 20% maxDD whilst the other can deal with 50%, what is the best position size each should use?

We can use the same numbers as before with our 50% win rate, 10,000 trades and losing streak of 13.

Therefore,

Max Risk = (1-((1-DD)^(1-LS)))*100

Where,
DD = maximum acceptable drawdown
LS = losing streak

To have a chance at a 50% maxDD we can risk 5.19% per trade or for a 20% maxDD we can risk 1.70%

nizar · 16 July 2007

THose equations are gold.

MaxDD i dont mind a high one but obviously it has to be worth it in terms of returns.

GreatPig · 16 July 2007

nizar said:
THose equations are gold.

Fools gold like that

Try:

Max Risk = (1-((1-DD)^(1/LS)))*100

Cheers,
GP

It's Snake Pliskin · 17 July 2007

nizar said:
Snake can you please elaborate on the bold part.

And also what do you mean by new low breakout?

Thanks.

If you buy high or low there is the possibility of a prolonged drawdown.

System design is paramount and understanding WHY it works as others have said too. If you take any old entry then surely that will impact on whether there is a drawdown or not. Surely the entry determines where and how much the drawdown will be (no exit or timefame in this comment) - only it will be hindsight and paradoxically is of no value until it happens.

theasxgorilla · 17 July 2007

This is a question from ignorance rather than an attempt to complicate what should be simple.

If your equity curve tracks open-equity then can you experience drawdowns due to open-equity profit retracement? Is this portfolio heat?

And if this is the case, then is it possible that the maxDD could actually be the value of portfolio heat plus the formula that Nick posted earlier, even if some lower maxDD amount was actually 'observed' during testing? If it's known that the system (keeping to the example system) can experience 13 consecutive losers and current portfolio heat is 10% then could the possible/expected maxDD in fact be 33%?

ASX.G

nizar · 17 July 2007

theasxgorilla said:
This is a question from ignorance rather than an attempt to complicate what should be simple.

If your equity curve tracks open-equity then can you experience drawdowns due to open-equity profit retracement? Is this portfolio heat?

And if this is the case, then is it possible that the maxDD could actually be the value of portfolio heat plus the formula that Nick posted earlier, even if some lower maxDD amount was actually 'observed' during testing? If it's known that the system (keeping to the example system) can experience 13 consecutive losers and current portfolio heat is 10% then could the possible/expected maxDD in fact be 33%?

ASX.G

Hi ASX.G,

I dont know the answer to your question, but this is also why I thought exits would have something to do with drawdowns (not only the entry).

Because while cash drawdowns would be due to a string of losses (hence maxDD can be derived from max. string of losses), drawdowns in open equity would be due to profit giveback, which in turn is caused by having loose stops.

Yeh??

Or are drawdowns in open equity something else??

CFD · 17 July 2007

Nick Radge said:
~~
Assume,

Win rate of 50%, Trade sample of 10,000, Losing streak = 13

Using a standard risk of 2% per trade on the above will give a theoretical maximum drawdown of 23%.

My understanding of a maximum drawdown is some what different.
The amount of drawdown your bank needs to withstand is different to the amount lost during a losing streak.

You calculate the likely longest losing streak from the strike rate in this case 13, but you may then only have, say, two or three wins before entering into another long losing streak. Your bank could have to survive a loss x 13, win x 3, loss x 10.

Now that's a drawdown and with a 50% s/r!

Nick Radge · 17 July 2007

CFD,
You're absolutely right, which is why I said this:

..then work for the losing streak we could get some kind of probability on drawdown size.

nizar · 17 August 2007

Im sure everyone who has ever designed a system has come across this dilemma.

Ok, so you have a system that has, in the past, delivered great returns, but the drawdowns are unacceptable.

How would you go about solving this problem?

I thought generally if you wanted a high return you'd have to accept a higher drawdown but for those who are proficient in systems trading know that this doesnt have to be the case.

Stevo's systems are pretty crazy, returns of 30-40%+pa and max. drawdowns in the single digits !

Any thoughts/comments/discussion would be much appreciated.

nizar · 18 August 2007

nizar said:
Im sure everyone who has ever designed a system has come across this dilemma.

Ok, so you have a system that has, in the past, delivered great returns, but the drawdowns are unacceptable.

How would you go about solving this problem?

I thought generally if you wanted a high return you'd have to accept a higher drawdown but for those who are proficient in systems trading know that this doesnt have to be the case.

Stevo's systems are pretty crazy, returns of 30-40%+pa and max. drawdowns in the single digits !

Any thoughts/comments/discussion would be much appreciated.

Any thoughts??

Temjin · 18 August 2007

nizar said:
Im sure everyone who has ever designed a system has come across this dilemma.

Ok, so you have a system that has, in the past, delivered great returns, but the drawdowns are unacceptable.

How would you go about solving this problem?

I thought generally if you wanted a high return you'd have to accept a higher drawdown but for those who are proficient in systems trading know that this doesnt have to be the case.

Stevo's systems are pretty crazy, returns of 30-40%+pa and max. drawdowns in the single digits !

Any thoughts/comments/discussion would be much appreciated.

Your best bet for developing high sortino ratio systems (annualized return over maximum historic drawdown) is simply to diversify and create as many opportunities as you can.

One example would be to diversify through trading with multiple "uncorrelated" systems. However, this does not mean you will develop several "trending" based system of different timeframe as the underlying methodology usually remain the same and gives positive correlation over the long term.

You should develop trading systems that can capture both trending periods and sideways through counter-trend strategies.

Another obvious example would be to diversify through trading with uncorrelated markets.

tech/a · 18 August 2007

nizar said:
Im sure everyone who has ever designed a system has come across this dilemma.

Ok, so you have a system that has, in the past, delivered great returns, but the drawdowns are unacceptable.

How would you go about solving this problem?

Well you have to find a way to minimise drawdowns.
The answer is in the exits.Your obviously giving back too much while allowing profit to run. Discretionary traders use a Trailing stop.
You could introduce a trailing stop after X periods---I havent done this so dont ask how to code it!

I thought generally if you wanted a high return you'd have to accept a higher drawdown but for those who are proficient in systems trading know that this doesnt have to be the case.

Stevo's systems are pretty crazy, returns of 30-40%+pa and max. drawdowns in the single digits !

Any thoughts/comments/discussion would be much appreciated.

See above. Look at exits.
If trading futures volitility would also have a large influence in this case Id be looking at timeframe.But I think your talking stock.

Crossed with Temjin ---some other good ideas there but doesnt deminish drawdown just disperses it..

Drawdowns

nizar

It's Snake Pliskin

nizar

Nick Radge

nizar

Nick Radge

brettc4

GreatPig

Pigs In Space

Nick Radge

nizar

GreatPig

Pigs In Space

It's Snake Pliskin

theasxgorilla

Problem solved... next bubble.

nizar

CFD

Nick Radge

nizar

nizar

Temjin

tech/a

No Ordinary Duck

Similar threads

Connect with us