Fitting an Optimum Stop Loss

TheBramble

Legendary member
Messages
8,394
Likes
1,170
Before I start expending hundreds of hours chasing wild geese down a blind alley (actually, that would be quite a good strategy wouldn't it!) - has anyone spent any time attempting to optimise their stop loss for any given instrument?

It is possible to determine what stop loss would have made the optimum profit for LONG positions and similarly for SHORT positions - over any selected period of time.

Optimum in that is didn't stop you out too early on good trades, but balanced that with getting you out soon enough on bad trades.

Although it's possible to do this on an instrument-by-instrument, period-by-period basis, I was wondering if any of the more mathematical types out there had a more general formula for deriving this info.
 
volatility stops are generally the best bet..

ie those based upon previous ranges.. (Average True Range)

ie if we have an erratic (erotic?) instrument (fnar fnar) then the stop needs to be wider or else we would get stopped out all too often.. on an end of day basis i like to use something in the region of 2*ATR , and preferably somewhere that coincides with major support (remember, were aren't allowed to go short anymore. LoadaDosh told us so, so resistance doesnt feature :rolleyes: )

as for intraday, then its a whole different kettle of fish, as market sentiment/noise can cause such choppy action that its a right pain. maybe in this case a fixed point (and therefore risk) stop is better, moving it to breakeven plus a bit when in profit, and then when the position goes further in your favour, pyramid in and use a trailing stop....

at least that is what i used to do before they firewalled me here at work. I wonder if they noticed :rolleyes:


i bet that doesnt help at all does it?


FC
 
No, it does help.

Factoring in volatility and using ATR sounds like a useful approach. I was just wondering if there was a rough formula that anyone had found fitted more than one instrument.
 
It depends very much on the system.

The only way is to test test test and then hope the forward testing corresponds to the backtesting.

JonnyT
 
following up on FC and just to prove I can read posh books
(Kaufman) I pass on the following. Don't ask me about it - I don't
think I even understand it :) :)

As a sound statistical measurement of risk the standard
deviation can be used to determine stop loss levels. The
Dev-Stop method devised by Cynthia Kase can be used both long
and short.

1. Calculate the true range (TR) of the past 2 trading days using
the highest high and lowest low of the 2 day period.

2. Calculate the moving average ATR of the TR using 30 periods
for intraday charts and 20 periods for daily charts.

3. Calculate the standard deviation of the true ranges in step 1
using the same period as in step 2.

4. The stop loss values are DDEV = ATR + (f*SDEV) where f = 1,
2.06 to 2.25, and 3.2 to 3.5, and where the larger values of the
pairs correct for skew and the larger numbers allow for larger
risk.

5. The Dev-Stop for long positions = trade high - DDEV; the Dev-
Stop for short positions = trade low + DDEV
 
thanks Barjon!!! lol

always good to have a simple system in a market moving quickly against you :LOL:

still, what it postulates makes pretty good sense, but again we are assuming that price conforms to standard distribution patterns ie 66%, 90%, 95% or whatever it is as we go through each SD

i seem to recall back in the dawns of time back on the Contemplating Trading Strategies board on the Fool, Martin/Hokusai298 did some research about price distribution and found that it didnt conform to normal patterns.

this probably explains why Bollinger Bands don't work as well as they should in theory.

one more thing, Barjon, thanks for making me think this morning. it is helping the hangover clear in no time. (was out late last night celebrating my exam results - pass one, fail one. Still, any excuse for a piss-up)

FC
 
ooo i didnt realise that we had swear-filters on this board??

in that case

lets have a test


**** **** ******** todmorden :LOL:
 
Barjon, looking very much along the lines I was hoping for. A useful calculation.

I'm going to check it out and see how well it does on a few of my favourites.
 
Yes, absolutley. A combination of the two would give a much higher probability of success.

Unless the entry was so well defined and executed that a stop loss was never needed...
 
barjon said:
5. The Dev-Stop for long positions = trade high - DDEV; the Dev-
Stop for short positions = trade low + DDEV

Just attempting to put this into Metastock formula and realised I don't understand what 'trade high' and 'trade low' refer to.

The lowest low and highest high of the last 2 days, previous day, today?
 
erm isnt trade high, the highest price when in the long, and therefore trade low the lowest price in a SHORT..

this is a trailing stop n'est-ce pas?



FC
 
I said not to ask questions :eek:

I think it must be be the high/low of the previous 2 days.

The source is: Cynthia Kase. "Redefining Volatility and Position
Risk" Technical Analysis of Stocks & Commodoties (Oct 1993) if
that helps in any way.
 
It's against my nature to be nice to people, I just want to exterminate everyone :devilish: But today I am unwell and sharing my DEVStop code for Matastock:

Period:=20;
f:=1.2; {2.2,3.6;}
SATR:=ATR(1);
RATR:=ATR(period);
DDEV:=RATR+(f*Stdev(SATR,Period));
Stop:=HIGH-DDEV;
If(LOW<=PREV,Stop,If(Stop>PREV,Stop,PREV));

I think this is the correct version :rolleyes: It's a stop from the long side - reverse the code for a short.

The Case book has a lot of negative comment on Amazon, I found it interesting and worth the money.

One day I'll share my super duper Stochastic with wopping low latency ...... :D

Davros
 
Thanks Davros.

Re: One day I'll share my super duper Stochastic with wopping low latency - guess we'll have to wait until your next off day -until then, carry on exterminating.

Thanks DaveJB - great research - and lots of it. If you're trying to cure me it's not working!

One thing I have established (and please let me know if I'm off track here) is that when attempting to adjust ANY automatic stoploss calculation to widen the gap (allow more play) at:- round, half and quarter points, or pivot points or support/resistance - the increased risk is NOT rewarded by staying in more winning trades longer or taking more profit.

In short, I'm suggesting you DON'T play with your stoploss calc for these points as they more often than not do throw you out - and that on balance, it's better that they do.

The few that do get through (and I've only tested 60 stocks & instruments or so) do offer large further runs, but happen so infrequently that you'd make less profitable trades overall and your drawdown would be much greater.

Does this make sense? Or am I stating a behaviour that's so obvious your're wondering why I'm bothering to state it?

What I'm saying is that hitting a round, half or quarter point or a pivot point or support/resistance is MORE LIKELY THAN NOT to reverse the current trend. And (my data) likely to do so 68% of the time.

If the above supposition is correct, it would provide a potentially fairly accurate target setting technique.

I can't shake this nagging feeling that this is all basic stuff, but I haven't read/heard/seen anywhere else.
 
Last edited:
Top