Money management - relationship between profit and position size

[MRC & Co]

I suspect target exit are the best performing as there is no "give back"-----only in some cases lost "Potential" Profit.

Exactly. Has to be what Howard means.

In short timeframes, you would not want to be giving back a few % of your winnings. It would be very costly to your system.

 

[bassmann]

Nizar – I may not have been clear in my example on increased Risk following a large move so I’ll have another go.
Using the same example I have a position that moves 50% in one day. I have a trailing stop to manage the position that moves up following the move. Now I’ll re-illustrate my example.

Account size is 100k and 20k goes into the position with a 10% stop, so at entry the RISK is 2%.
The position moves 50% and is now worth 30k and account value increases to 110k. Let’s assume that the stop is still 10%, therefore new risk is (30*0.1)/110=2.7%, hence an increase of 0.7%.

I was wondering if there would be any special treatment given in the case since as if demonstrates the potential to be overexposed in a single position.
Peter 2 has indicated that Portfolio Heat is more important to his trading style than Risk per Trade and does not include open profits in the PH calc. On that basis no action would be necessary in this example.

Peter2 – Can you clarify why you don’t include open profits in the Portfolio Heat calc. Don’t you wish to protect open profits as well as initial capital in the trade?

Interested to the view from others.

Since I did not state this previously, I thought I might add that I am currently trading this system.


Regards,

Bassmann

 

[nizar]

Nizar – I may not have been clear in my example on increased Risk following a large move so I’ll have another go.
Using the same example I have a position that moves 50% in one day. I have a trailing stop to manage the position that moves up following the move. Now I’ll re-illustrate my example.

Account size is 100k and 20k goes into the position with a 10% stop, so at entry the RISK is 2%.
The position moves 50% and is now worth 30k and account value increases to 110k. Let’s assume that the stop is still 10%, therefore new risk is (30*0.1)/110=2.7%, hence an increase of 0.7%.

Bassman.

The way I see it if your system is a short term one, the stop should NOT be static when the stock moves 50% in your favour.

If the system is longer term, then your backtesting would have allowed for this increased risk.

The fixed percent risk for position sizing is based on closed equity (at least in TradeSim it is) so in this sense you are still risking 2% on the trade.

But there are aggressive position sizing methods that are based on open equity.

What's important is that you trade as closely as possible to the backtesting program that you used to design the system.

Nizar.

 

[tech/a]

Nizar – I may not have been clear in my example on increased Risk following a large move so I’ll have another go.
Using the same example I have a position that moves 50% in one day. I have a trailing stop to manage the position that moves up following the move. Now I’ll re-illustrate my example.

Account size is 100k and 20k goes into the position with a 10% stop, so at entry the RISK is 2%.
The position moves 50% and is now worth 30k and account value increases to 110k. Let’s assume that the stop is still 10%, therefore new risk is (30*0.1)/110=2.7%, hence an increase of 0.7%.

No you only calculate risk to realised capital not open profit.Peak to valley (High of the portfolio total equity to the low of the total equity within a trading period) is known as Peak to valley drawdown.
Regardless of how much open profit you have if all positions crashed to your stop your total portfolio risk still remains at x% if you liquidate then the calculations are based upon the NEW closed equity.

I was wondering if there would be any special treatment given in the case since as if demonstrates the potential to be overexposed in a single position.
Peter 2 has indicated that Portfolio Heat is more important to his trading style than Risk per Trade and does not include open profits in the PH calc. On that basis no action would be necessary in this example.

No action is correct.

Peter2 – Can you clarify why you don’t include open profits in the Portfolio Heat calc. Don’t you wish to protect open profits as well as initial capital in the trade?

Interested to the view from others.

Since I did not state this previously, I thought I might add that I am currently trading this system.

This is a totally seperate issue and one which you'll find a masterful balancing act. Giving positions enough room to correct whilst not giving back to much of open profit if the correction turns into a reversal.
For Discretionary traders this is part of their analysis skill set. For systems traders this is the setting of trailing stops to find an optimum---which will never be perfect but will encompass enough trades to strike a profitable balance. The discretionary trader will generally (if proficient) suffer less slippage of profit.

This is one of the reasons Howard Bandy suggests trailing stops arent helpful in short term systems. Price targets being the "Ideal".

 

[MichaelD]

What if there is a big move on one stock in the portfolio, say 50% in a day.. If you work on a principle of 2% risk this may increase substantially and would suggest an exit. This goes against the principle of cutting losses short and letting profits run.

The total % risk for your portfolio varies on a day-to-day basis. The 2% rule applies only at trade entry - this is the maximum amount of your initial capital you are prepared to lose with one given trade. If the trade takes off, great - you will have more than 2% of your capital between the current trade price and the stop loss, but this does not mean exit, this means move your stop up according to your pre-tested rules - a trailing stop will automatically do this.

I have done further position size testing on Fixed Position sizing and Fixed Fractional sizing. In each case I have factored brokerage and excluded outliers (best 3/worst 3 trades) and I have simulated at different risk levels.

The results on my system show that Fixed Position sizing gives better results than Fixed Fractional sizing.

These results are intriguing and contrary to all the work I have done/have read about in this area. I feel you need to figure out why this is so.

A suggestion;

- Compare systems by normalizing the drawdowns so that the drawdowns are the same across all systems - then look at the other system parameters such as profit and number of trades.

If fixed sizing is giving you twice the profits with three times the drawdown compared to fixed fractional, then in fact it is a poorer performing system despite the extra profit.

 

[peter2]

Bassmann: Protecting open profits is an individual trade management issue. As the trailing stop is above breakeven there is no risk to your starting capital with this trade. [I agree with Tech/a and MichaelD]

Just to clarify, I manage the risk in each trade primarily. Portfolio heat prevents me starting too many trades at the same time. If you start 5 trades at 2% then you will lose 10% of your capital if all trades are stopped out. What is the probability of you having 5 losing trades in a row? How are you going to handle this?

MichaelD, I agree. The only way fixed pos. sizing gets bigger profits than fixed fractional sizing is because the dollars at risk is greater (or the sample size is small).

 

[howardbandy]

Hi,



Howard, could you explain why you think this is - I'm particularly interested in whether this is the case (and why) for strategies based on scalping indexes.

thanks
Daniel

Hi Bear --

If you are using end of day data, and holding a few days, there is much less opportunity for the trailing stop level to catch up to the price, so the exit remains at the level it was set when the trade was entered. Compare this with a holding time of a few weeks, where the exit price moves up as the trade gains profit.

For short term, particularly mean reversion, trades, try testing an exit that has three components:
1. A signal. If you bought oversold conditions, and the prices rises quickly enough to give an overbought condition, use that as an exit signal.
2. A profit target. Look at the historical results to see what the distribution of the maximum favorable excursion has been following the entry signal. Set a profit target based on that. Look for a profit target that gets hit about half the time, then vary that percentage as one of the optimizable parameters.
3. A timed exit. Look at the historical results to see the distribution of profit versus holding period following your signal. Set an exit based on holding a fixed number of days based on that.

Exit the position when any of the three conditions is met.

As a second phase, you can combine all three and adjust the exits as the trade progresses. But, with EOD data and holding of just a few days, there will not be much opportunity for adjustment or improvement.

If you are using intra-day data and holding a few days, then you might try parabolic or chandelier exits based on intra-day bars.

As always -- do your testing on whatever in-sample data you wish, but only results from out-of-sample data have value when estimating likely future performance.

Thanks,
Howard

 

[WaySolid]

Hi,



Howard, could you explain why you think this is - I'm particularly interested in whether this is the case (and why) for strategies based on scalping indexes.

thanks
Daniel

My own limited testing says that the the tail you seek in no way compensates you for the give back from letting the market always take you out.

If you introduce an automatic time stop, which by definition all day traders have then targets can work particularly well.

I can't imagine how you could possibly use a trailing stop for anything related to very short time frame trading.

Just measure everything, as much time should be spent edge testing exits if not more in my opinion compared to entries.

 

[nizar]

Just measure everything,

Well said, I agree wholly.

Instead of going with popular opinion, or with what you have read, just test everything.
Its the best way.