How many contract to trade? Strategy allocation.

[horace]

Hi!

From the risk side point of view, I'd like to know how you fellow traders allocate contracts to a strategy.

Suppose you have a strategy that has the following stats:

Max DD: 2,500
Average loser: 116
Worst loser: 988
Pct Winner: +43%
Win/Loss: 1.68

Excluding any type of money management strategies, what should be, in your opinion, the starting equity allocated per contract.

I currently use the 2% risk per trade rule, but I imagine something better can be used.

Any help will be appreciated.

H

 

[nikke]

You dont consider the PF so much?

I would take the max DD, add a bit, and consider how big DD I am willing to go trough.

If it was a swing-for-the-fences system some fraction of Kelly would be an idea.

 

[horace]

Hi nikke,

No, I don't consider PF at all. Actually it really does not help my trading on a day to day basis.

Yes, the max DD option is a good one also, I agree.

What do you mean with the "swing-for-the-fences " arguement¿?

Thx for your reply.

H

You dont consider the PF so much?

I would take the max DD, add a bit, and consider how big DD I am willing to go trough.

If it was a swing-for-the-fences system some fraction of Kelly would be an idea.

 

[nikke]

I meant like setting aside a portion of money for it and willing to lose it all but trying to hit it real big with the system, eg. 25% of Kelly I have heard some use.

 

[horace]

Hi nikke,

I understand now. Fortunately I also understand this is a job, not a game as is the casino.

 

[nikke]

Good was just throwing out ideas

 

[forker]

I work by account percent in (your base currency of your account) divided the pips between the entry and the stop

Eg:
account capital is £150 000
Risk 2% or 3000
Stop is 60 points ( inclusive of 5 points spread either side as a buffer)
Trade is £50 per point

 

[the hare]

I suppose the first thing you need to consider is what exactly are your objectives ? Are you trying to optimise returns ? or mimimise drawdowns ? The answer to that particular question is going to have a major impact on the type of MM you end up using.

Regardless of objectives, the first thing I do is take all of the historical data for a strategy, both profits and losses, then create a bunch of simulated equity curves by picking values out at random (resampling)

In simple terms, you can imagine a situation where by luck you keep on selecting winning trades and not selecting losers, resulting in a great equity curve. The converse is also true, you could end up picking loads of losing trades, and get far greater drawdown than you've experienced in practice.

Then I apply some simple stats to find the distributions in drawdown, and the upper and lower boundaries of the simulated equity curves based on confidence intervals. That gives a starting point for the sort of drawdown that I might have experienced should things have gone tits up. It not perfect or even statistically valid but its probably better than picking a number out of thin air. It also allows me to draw a line in the sand where I'll stop trading a given strategy.

This approach tends to give a worst case scenario which limits gains, so in practice, I tend to run the position sizing a bit more aggresively on the basis that when the **** hits the fan trading automated strategies my discretionary skills are usually sound enough to claw back some of the losses until things get back on track.

 

[horace]

I understand, that is the same I use, a 2% risk per trade.

Thx

I work by account percent in (your base currency of your account) divided the pips between the entry and the stop

Eg:
account capital is £150 000
Risk 2% or 3000
Stop is 60 points ( inclusive of 5 points spread either side as a buffer)
Trade is £50 per point

 

[horace]

Nice thinking. I usually do this analysis of randomly reorderin the trades. This is done out of the box by certain software apps. The question I always bring on the table is that this analysis is not valid for mechanical strategies because I beleive that one trade is not independent from the last. So this renders this random reorder useless imho.

Monte Carlo analysis is also a good way to go following your way of thinking about the expected DD. But I like yours more (providing we accept randomly reordering trades as valid). I'll do a couple of test tomorroww and see what I get.

Thanks for your interesting reply.

H

I suppose the first thing you need to consider is what exactly are your objectives ? Are you trying to optimise returns ? or mimimise drawdowns ? The answer to that particular question is going to have a major impact on the type of MM you end up using.

Regardless of objectives, the first thing I do is take all of the historical data for a strategy, both profits and losses, then create a bunch of simulated equity curves by picking values out at random (resampling)

In simple terms, you can imagine a situation where by luck you keep on selecting winning trades and not selecting losers, resulting in a great equity curve. The converse is also true, you could end up picking loads of losing trades, and get far greater drawdown than you've experienced in practice.

Then I apply some simple stats to find the distributions in drawdown, and the upper and lower boundaries of the simulated equity curves based on confidence intervals. That gives a starting point for the sort of drawdown that I might have experienced should things have gone tits up. It not perfect or even statistically valid but its probably better than picking a number out of thin air. It also allows me to draw a line in the sand where I'll stop trading a given strategy.

This approach tends to give a worst case scenario which limits gains, so in practice, I tend to run the position sizing a bit more aggresively on the basis that when the **** hits the fan trading automated strategies my discretionary skills are usually sound enough to claw back some of the losses until things get back on track.

 

[Adamus]

I would use 'optimal f', the fixed fraction of the money you are allocating to your system. Ralph Vince (who wrote Mathematics of Money Management) states that you should apply your largest single trade loss as one of the parameters, which is the biggest problem with the method. I don't know why, but Vince never recognises that the biggest loss is in the future. For that reason I use an estimate of what I might suffer, compromising between what might happen in a Black Swan event and a figure that will allow me some compounding.

 

[horace]

Ok, thx Adamus. In my opinion, Optimal F is a ultra agressive MM technique and does not apply to every type of accounts.

Thx

I would use 'optimal f', the fixed fraction of the money you are allocating to your system. Ralph Vince (who wrote Mathematics of Money Management) states that you should apply your largest single trade loss as one of the parameters, which is the biggest problem with the method. I don't know why, but Vince never recognises that the biggest loss is in the future. For that reason I use an estimate of what I might suffer, compromising between what might happen in a Black Swan event and a figure that will allow me some compounding.

 

[Adamus]

Ok, thx Adamus. In my opinion, Optimal F is a ultra agressive MM technique and does not apply to every type of accounts.

Thx

That's an urban myth. IMHO I have never come across a better or more widely applicable technique.

 

[the hare]

That's an urban myth. IMHO I have never come across a better or more widely applicable technique.

Are you are having a laugh ?

Whilst its true that Optimal F will enable you to hit a particular equity target in the shortest timeframe possible, it also results in the greatest volatility in returns.

There are 1001 problems with optimal F, the main one being you are dealing with non stationary time series, and the optimal F you calculated based on historical data wont be the value of optimal F in future. A simple walk forward optimisation illustrates this perfectly well.

If you are lucky, you'll underestimate the true value, and your returns will be sub optimal but with drawdowns from hell. If youre unlucky, you'll overestimate the value and your account will be history.

To claim that the agressiveness of optimal f is an "urban myth" is irresponsible nonsense (byut par for the course at T2W so thanks for the LULZ )

There is a place for optimal F but discussion about that is best restriced to a more suitable venue. You have to realise that the rate of return isnt particularly important to many traders, they are more concerned with varience, and thats why optimal F is a non starter.

 

[Adamus]

Why the outrage? And what would be the place to discuss it? This is the mechanical systems trading forum, isn't it?

I said earlier you should not use the 'largest historical loss' rather an estimate. The input parameters give you control over the leverage and the volatility. The optimal f calculation will give you a graph of risk of ruin against leverage. Then it's in your hands. You choose what suits you best.

Your argument about a non-stationary series, if you still want to make it, is not exactly useful. Of course the future will be different. What would you do instead? You need a basis for your predictions and that is your historical results.

The time series varies of course but your system must be robust and perform with at least some degree of consistency, otherwise you wouldn't want to trade it anyway.

And then you combine the system with other systems with an eye on correlation to ameliorate the volatility of returns - and just grin and bear it through the drawdowns.

I'm not trying to bash you or your opinion - I am just saying that you haven't said anything yet to convince me to discard it, or that you have something better to propose.

 

[the hare]

Your argument about a non-stationary series, if you still want to make it, is not exactly useful. Of course the future will be different. What would you do instead? You need a basis for your predictions and that is your historical results.

I suppose I'd agree that its not useful, but it is a fact that sadly wont go away. I was mainly responding to the suggestion that optimal f wasnt a agressive strategy. With the exception of betting your whole equity on the outcome of the first trade I cant really think of anything more agressive

The issue for me, is that if you do trade optimal f in it truest sense, and your sitting on the right hand shoulder of the curve, then the smallest change in market conditions can tip you over the edge from optimum returns to zero.

The minute you start trading at fractions of f then surely its not really much different to trading any other fixed fractional method at x% where x satisfies your own personal risk and volatility tolerances ?

Or are you perhaps suggeting that you calculate F on an ongoing basis, and then trade a fraction of F (I'n not sure that would work either as the new value of f is as likely to go up as down).

 

[Adamus]

Sorry, had to re-edit this.

Optimal f is no more aggressive than a gun - it's what the person does with it that counts, as you recognise, but that doesn't make it 'not really much different to trading any other fixed fractional method'.

By putting your P and L thro optimal f, you end up with a chart showing returns (a sort of notional number representing your capital after a very long period of time) on the y axis against fixed fraction on the x axis. The chart goes up, over the brow of the hill and then at bigger fixed fractions, it tails off down to zero, i.e. your point of guaranteed ruin.

The top of the hill is optimal f.

Any fraction greater than that is pointless. You wouldn't raise your risk of ruin for less rSorryeturn.

Any fraction less than that is your choice based your estimate of the consistency of the returns and the robustness of the system. And your estimated greatest single loss on any one trade, although that is an input to the calculation, not a chart axis.

 

[horace]

Sorry to be late here.

I'm afraid the point here is not wherether optimal-f is risky or not. At least not in this post. Optimal-f is simply useless in real word trading, no matter what.

Theory and paperwork support whatever the point is in favour of this method. Truth is you will never see a sucessfull trader trading if for a long time because sooner or later you'll lose all your money. All of it.

Thanks for posting your ideas.

Sorry, had to re-edit this.

Optimal f is no more aggressive than a gun - it's what the person does with it that counts, as you recognise, but that doesn't make it 'not really much different to trading any other fixed fractional method'.

By putting your P and L thro optimal f, you end up with a chart showing returns (a sort of notional number representing your capital after a very long period of time) on the y axis against fixed fraction on the x axis. The chart goes up, over the brow of the hill and then at bigger fixed fractions, it tails off down to zero, i.e. your point of guaranteed ruin.

The top of the hill is optimal f.

Any fraction greater than that is pointless. You wouldn't raise your risk of ruin for less rSorryeturn.

Any fraction less than that is your choice based your estimate of the consistency of the returns and the robustness of the system. And your estimated greatest single loss on any one trade, although that is an input to the calculation, not a chart axis.

 

[Adamus]

Sorry to be late here.

I'm afraid the point here is not wherether optimal-f is risky or not. At least not in this post. Optimal-f is simply useless in real word trading, no matter what.

Theory and paperwork support whatever the point is in favour of this method. Truth is you will never see a sucessfull trader trading if for a long time because sooner or later you'll lose all your money. All of it.

Thanks for posting your ideas.

Hi Horace,

you have my very focused attention. Why will I lose all my money? I assume you are a fairly rational and intelligent person, and you have obviously discovered something about optimal f that escaped me entirely.

If you wish to help but don't wish to have any more discussion of optimal f on your thread here (I assume that's why you don't go into any detail), please send me a private message.

Thanks.

 

[horace]

Hi Adamus,

I'm very sory if my response sounded a little bit rude. It was not my intention and probably was because my english is not as good as it should.

Optimal-F is a methor that only works on paper. If you are lucky enough to survive your first 6 months, your will finally get this 90% DD that is 10% probable. Or else you'll need to be trading such a big amount of contracts that your trading method will be rendered useless by brokers and slippage.

It's just a matter of backtesting and taking a close look at numbers.

Thanks for your help.

H

Hi Horace,

you have my very focused attention. Why will I lose all my money? I assume you are a fairly rational and intelligent person, and you have obviously discovered something about optimal f that escaped me entirely.

If you wish to help but don't wish to have any more discussion of optimal f on your thread here (I assume that's why you don't go into any detail), please send me a private message.

Thanks.