May 19, 2003

Beat up on his last few trades on Globex, a Spec asked for insights on using stops.

**Brian J. Haag: **Hrmm ... this would seem to be a topic
that some

members of the list might not like, but I'll hazard

some thoughts anyway.

Stops should provide some (that's "some", not "total")

assurance you're going to stay in business. But they

are FAR from a sure thing, and useful only for certain

types of investors. If you're trading $1bio at a

time, stops aren't going to help.

As they are a form of insurance (which is essentially

the purchase of negative return volatility from

someone else), they're not free -- that assurance is

going to cost you in the form of decreased returns.

In my experience, they're only useful as catastrophic

protection. You have to have a method of setting them

that allows for normal market volatility. If you fail

to account for the fact that the time series is noisy

(and set your stop outside that normal level of

noise), stops will do nothing but cost you money and

cause you anguish.

Best way to do this? Count! Take the market you are

active in, measure historical volatility (taking into

account such factors as serial dependence and

seasonals), and set your stops some amount away from

your entry to allow the trade plenty of "breathing

room".

Volatility exhibits serial correlation (or

homoskedasticity if you prefer), meaning that low

volatility is followed by low volatility, and high

volatility is followed by high volatility.

Unfortunately volatility is also demonstrably

mean-reverting, making volatility prediction a science

all its own.

Volatility also has seasonal characteristics (such as

the spring weakness in the VIX).

**Hany Saad: **

Puts are the most expensive form of insurance of all.

My only use of puts is to write nakeds instead of putting a limit order. I either

keep the premium or get assigned the security I wanted to go long of in the first place.

For the globex trading, I hardly use stops specially during the low volume night

trading.This is the best time for them to get hit. Also,I`d avoid Fib.numbers for

stops.They don`t make any sense but you`ll get the odd technical trader who will

hit your level for no other reason.

If you`re not trading volume a la Mr. E and have no one to watch 24 hours a day

for you,an alternative is to set an annoying alarm at a certain price level, When

it goes off you can watch the price action with your own eyes then make a

decision instead of getting out automatically.

P.S. After you get up to watch the price action, try to avoid reading the news at any cost.

**Brian J. Haag: **

>

> --- HANY SAAD wrote:

> > Puts are the most expensive form of insurance of

> > all.

>

Yes, but this is because they are the only kind of

insurance that works every time. No chance of

slippage or non-execution of a stop. You get what you

pay for.

**Hany Saad:**

Notice Mr.Haag, I didn`t endorse the use of stops either.

The vig. of buying puts to hedge is too high for your position to be

profitable. There are better techniques than stops and puts. Getting out of the

position once your reason is not validated for one is a better strategy.Your

post on volatility makes sense as well. However, volatility stops are already

discovered by most participants. The turtles system of the eighties incorporated

volatility to stops.

**Brian J. Haag: **

--- HANY SAAD wrote:

> The vig. of buying puts to hedge is too high for

> your position to be profitable.

This would depend on the profitability of the strategy

in the first place, as well as the kind of puts you

buy and when.

> There are better techniques than stops and puts. Getting out of the

> position once your reason is not validated for one is a better strategy.

This is an interesting question, and hopefully one list members will comment on (water or wine?). When designing mechanical trading systems, I find that only

catastrophic stops (as I mentioned earlier) are useful.

>Your post on volatility makes sense as well. However, volatility stops are already

>discovered by most participants.The turtles system of the eighties incorporated volatility >to stops.

There are many things that have been "discovered" that remain sensible (profitable is another thing entirely).

**Euan Sinclair**:

I recently read an interesting working paper by Hodges,Tompkins and Ziemba

which addressed the profitability of buying out of the money options as

directional plays (either speculative or as hedges). Their conclusion,

based on the movements of the SP500, was that the movement in the underlying

practically never justified paying the offer for the option. It is possible

of course that other markets, particularly more kurtotic ones, will be

different.

They framed the situation as an example of the favourite/long shot bias,

which is well documented in horse racing studies. Basically the market-maker/bookie

won't give you a decent price for teeny options or long shots because his

risk reward is lousy.

I'm not sure if hitting the bid would be consistently profitable either

though. The bid/ask may be too wide.

**Tom Ryan: **i find this thread highly ironic (and
symptomatic) given the shift in

volatility exhibited by this morning.

the evolution of the trader to the tune of the theme from kubrick's 2001..

the horns and the big drums.... sunlight breaks on the land, the sunrise....

the salamander crawls onto the land and realizes that the land is wonderful

but dangerous. let us place some mechanical stops x% under my longs, to

protect ourselves. death from many small cuts...

realization that x in x% must be adjusted from trade to trade to account for

future volatility...the salamander evolves into the large lizard....

drum roll and horn reprise

now evolution beyond simple stop levels to avoid being such an easy meal,

the large lizard now has become the dinosaur...ruling the earth....believing

in his safety net.....look up! hey what is that streaking across the

sky........ its the meteor of 1987! crash. darkness. executions 30% under

the stop levels. extinctions.

the meek and tiny mammal emerges from the darkness....crawling up the stairs

some more, the realization that maybe stops aren't such a foolproof risk

management method afterall......

stooped over neanderthal man, the advent of fire, hey we can test these stop

levels to see if they improve our trading system. technology! counting! a

tool!

the birth of modern day trader man, standing tall and straight.

enlightenment, the leaving behind of superstition, the application of the

scientific method, no stops now required!

crescendo!

**Henry Carstens** (5/20/3)

Counting on Stops, The Mathematics of Stop Losses

To determine if you should use a stop loss with your trading system, test

the system both with and without stops and then apply the formula below:

R = E/$f + 1

Where:

R = geometric mean return

E = expectation per trade

f = optimal f

$f = largest losing trade/f

The formulas are:

f = ((($w/$l + 1) * p) - 1)/ ($w/$l)

E = f * ($w/$l)

$f = abs(largest losing trade)/f

Simplifying:

R = E/$f + 1

= (f * ($w/$l)/$f) + 1

= f^2 * $w/$l / abs(largest losing trade) + 1

The larger the resulting R the greater the compounding power of the system

and hence the better system at least in terms of compounding wealth.

(Derived from formulas in The New Money Management by Ralph Vince,

http://www.amazon.com/exec/obidos/tg/detail/-/0471043079/)

Example:

System A w/ stops

$w/$l = 1.59

f = .11

ll = -7125 (largest losing trade)

R = (.11^2 * 1.59 / abs(-7125)) + 1

= 1.0000027

System A w/o stops

$w/$l = 1.1

f = .14

ll = -11,675 (largest losing trade)

R = (.14^2 * 1.1 / abs(-11,675)) + 1

= 1.0000018

The larger resulting R from using stops means that System A will compound

capital faster using stops with the system than if we don't. As a matter of

fact, the system with stops will compound capital at a rate that is 50%

faster than the system w/o stops (.27/.18 - 1).

Conclusion:

Stops can be beneficial because compounding can be maximized by either

increased expectation per trade or reduced variance per trade. Therefore if

using a stop reduces the variance per trade more than it reduces the

expectation per trade, stops will be beneficial to the trading system.

**Tom Ryan **(5/20/3): yes but this is highly dependent
on the inputs you are using, and with all

due respect you have presented what appears to be a skewed example. in

reality don't we tend to find that stops

1. reduce variance but

2. reduce expectations

with one offsetting the other, and in many cases resulting in a system that

is worse than if we traded our patterns with specific exit times in mind.

i have always been concerned that optimal f and using stops to reduce

variance encourages too much leverage. the philosophical problem i have with

optimal f, and just about any specific stop placement strategy, is that not

only are we testing disn's of results for specific time periods, now you are

introducing a path dependence to the problem. path dependence i find highly

problematic in any adaptive type of system because:

1. any mathematical system of placing stops can be reverse engineered and

gamed, it makes you a potential target.

2. volatility is a pink noise process i.e. it is anti-persistent so the

numbers we put in today based on yesterday often have no relevance to the

numbers to be experienced tomorrow.

3. in addition, when testing these stops people have a tendency to neglect

slippage from their stop level. i apologize for my silly joke yesterday on

the list and i am not trying to make a case for the derivative expert but

the proverbial meteors do strike on occasion. they have several times in my

lifetime already and it would be folly to assume that it won't happen again.

the main problem with relying on reduced variance to increase your leverage

is that you are assuming that the stops will save a highly leveraged

position from going against you, that the markets will be liquid and

continuous enough for executions near your stop. that has been the ruin of

many, and don't forget i am the one who stands accused (and guilty as

charged) of overtrading size. and even i am skeptical of this reduced

variance/increased leverage approach! Tr

**Philip J. McDonnell, private trader (5/20/3): **

With respect to using stops there is a simple rule to remember:

Stops will DOUBLE your probability of loss.

The above is a simple mathematical fact which can be derived from the fact

that price changes follow a log normal distribution to a good approximation.

Suppose you had tested a system which gave you a 60-40% win-loss ratio.

That percent loss would rise if a stop is placed.

A simple way to look at this is to consider the probability of being at or

below a certain price after a given time window. In an efficient market the

probability of an end of period price at or below the current price is about

50%. If we place a stop at the current price the chance of execution is

very nearly certainty. (Yep, 100% = 2 x 50%). A stop at the 40% mark will

have an 80% chance of execution and so on.

Simply put, once a given price is reached there is an equal chance of going

up or down. For every negative outcome which would have been below say the

40% mark there are also an equal 40% of positive outcomes which would have

been above that level.

My suggestion for limiting risk is to avoid stops and use diversification

instead. Smaller positions result in smaller losses.

**Victor Niederhoffer (5/20/3):**

Philip McConnell wrote: >>>Stops will DOUBLE your probability of loss.<<<

Is this true?

**Alex Castaldo (5/23/3):**

He is trying to use The Reflection Principle for Brownian Motion but he is

misapplying it greatly.

RP: The probability that the price will go below the level L sometime between

now and some future date is twice the probability that the price will be below

L on that same future date.

That's because of all the paths which fall through the level L half will

recover to end above L and the other half will end up further down below L on

the specified date.

So far so good.

Where he goes wrong is:

> Suppose you had tested a system which gave you a 60-40% win-loss ratio.

The correct statement would be: suppose you consider it a WIN if the price is

above a level L on some future date and a LOSS otherwise. And suppose the

level L is set so that the probability of a loss is 40%. Then by adding a stop

at the same level L you double your probability of loss to 80%.

But this is not a realistic model of a futures trading situation. We don't win

or lose based on reaching specific levels by specific dates. We win or lose if

we end up above or below where we started.

Besides with brownian motion we have an efficient market and speculation is

useless.

The math is ok but the application is not right.

**Henry Carstens (5/24/3):**

No because it assumes that the probability of loss is the same regardless of

the placement of the stop which, under the assumption of a log normal

distribution is untrue.

Use the formula below to determine the viability of a stop instead:

R = E/$f + 1

Where:

R = geometric mean return

E = expectation per trade

f = optimal f

$f = largest losing trade/f

The formulas are:

f = ((($w/$l + 1) * p) - 1)/ ($w/$l)

E = f * ($w/$l)

$f = abs(largest losing trade)/f

Simplifying:

R = E/$f + 1

= (f * ($w/$l)/$f) + 1

= f^2 * $w/$l / abs(largest losing trade) + 1

The larger the resulting R the greater the compounding power of the system

and hence the better system at least in terms of compounding wealth.

**Brian J Haag: **

--- Tom Ryan wrote:

> in reality don't we tend to find that stops

1. reduce variance but

2. reduce expectations.<

I agree with this 100%. But they don't necessarily

reduce both equally.

I include stops in historical testing in an effort to

model some logical skin-saving measures. For example,

if I have a signal to short the market today and cover

at the bell, and then Osama Bin Laden is caught with

Saddam Hussein, I don't wait until 4:00p to cover. I

get out of the way of the freight train. And I want

my testing to reflect that. So I first test with no

stops at all, then I review the specific instances,

and look to see if negative outliers were avoidable

(defined by my ability to get out of the position).

If they weren't, my logic is faulty. If they *were*

avoidable, I model with stops in.

Your mileage may vary. My attitude is this: I know

what has worked in the past, and I have a reasonable

expectation of what will work in the future. But as

long as that expectation is *only* reasonable (black

swan, anyone?), I get out of the way when I'm wrong.

The Holy Grail of trading: being able to come back to

work tomorrow.

* * *

Note: A useful discussion of risk management appears at:

http://www.seykota.com/tribe/risk/index.htm

* * *

**Nigel Davies, Southport, U.K. (5/21/3):**

Er... wouldn't you at least like to know what these exit strategies

are before plunging in with the refutation?

One of them comes from one of the most sacred bastions of the

dark realm of pseudo-science, 'Trade Your Way to Financial

Freedom' by Van Tharp. In tests with fund manager Tom Basso

Van Tharp and he claim to have found a strategy based on random

entry and having a stop at a distance of 3 x "ATR" (average true

range) from the most favourable point of the trade (ie they keep

moving it further into profitability as the trade develops favourable).

The idea appears to be to create the opposite mode of behaviour

from what are claimed to be the habits of losing traders (ie taking

small profits and let losses run).

Their claim was one of 'profitability' over most of the 10 markets

tested, and (I seem to recall) 'profitability' in all of them when they

added a 'money management stop'. The book then goes on to look

at attempts to enter the market favourably. Basso uses, I believe,

some moving averages for his entry but the message seems to be

that tea leaves are also OK as long as the exit strategies are

good...

Sorry I haven't managed to confirm or deny this stuff with some

figures - I can't simulate this stop in Metastock and haven't

managed to program it into Excel for testing there either. There is

at least a vague connection with the list as Van Tharp runs

seminars with Chuck LeBeau, who in turn runs them with Dr Elder.

And Dr Elder tells entertaining stories about escaping from Soviet

ships, which is a kind of stop-loss with regards to life under

communism!

**Peter R. Gardiner (5/22/3):**

This is a very interesting discussion, but it leaves out the critical

variable of exposure. The assertion Seykota makes that risk (not "luck"

or "payoff") can be "controlled" by "buying or selling" stock contains

within it the assumption that a stop-loss type of mechanism (or some

deux ex machine) is the "controller." But he of all people well knows

(and relates in his Schwager interview) that markets blow through stops

more frequently than we would like, therefore it remains unclear to me

how this essay addresses the critical issue of position sizing as a

function of volatility and equity base. A one contract position with a

20 point stop does not have the same risk as a twenty contract position

with a one point stop.

In addition, the "payoff ratio" must be produced by the exit

strategy of the trader. As he asserts, it cannot be controlled, but is

has to come from some rule, just as the entry does. In his example, the

entry is random (unidirectional in a random series); but we know the

exit (and therefore the payoff) cannot be.

My vote is that we discuss his article, and leave the NYMEX to

bid for all the gas on the list.

**Alix Martin (5/21/3):**

My refutation was directed at the idea of leaving stops in the globex

market during the night.

You should try to do a study on the 3 x ATR trailing stop in excel. The

30mn assigned to an Analysis Position in Power Chess should be enough

time ;)

Alix

**Brian Haag (5/28/3):**

I would only add this:

If you make a directional trade, you are assuming that

the distribution of returns, at the time you enter,

until the time you close, will *NOT* be normal. Even

if you are making the trade based on normality at a

higher scale.

For example, perhaps I want to buy the SPX every time

it closes down 3 weeks in a row and the net change is

-5% or more (I'm just making this up -- no edge is

implied on this trade). I buy because historically

when this has occurred, the market rallies 3% within

two weeks, with some sufficient measure of certainty.

Implicit in this trade is my expectation that in the

next two weeks, the SPX's distribution is not normal,

but displays positive skew and perhaps some kurtosis.

As I've said, I'm far from fluent in matters

statistical, so if I've misused some terms, my

apologies. I hope that my point is clear.

**Philip J. McDonnell (5/28/3):**

Actually I believe that we don't necessarily care about the underlying

distribution for a directional trade. Rather the first order consideration

is that the expectation over all outcomes is positive.

It should be pointed out that the distributions of outcomes for a favorable

directional trade may still be normal. It is altogether possible to have a

favorable system which results in a normal distribution with a mean > 0.

Also we should note that a log normal distribution looks like a right skewed

distribution when looked at as a simple arithmetic scale. It's only when it

is rescaled on a log scale does it look like the standard symmetric bell

shaped curve.

I think your terminology is very correct & understandable.

**Victor Niederhoffer (5/27/3)**

The gentlemen seem to be having a good time theoretically discussing the proper placement of

stops. It reminds me of Metallegesellschaft who was so smart to prove that

they should sell the front month and buy the next month in oil as if their

placement of stops and fixed activities didn’t affect prices. They do ... they do.

The stops lead to opportunities for squeezes and therefore should only be used

for money management purposes in event of protection from disaster rather than

as a fixed thing.

The theoretical discussions are flawed in another way in that they don’t

answer. All discussions of this matter should be handled by

simulation. Merely take the basket of stock returns. Put them in a urn. Number

them. Then sample them with replacement randomly to see how you would do with

stops. Mr Alix from France is very good at this and we should defer to him for

an answer on any specific query keeping in mind that the use of any stops at

all would be disastrous because they would be run (except in aforementioned case).

I had a friend once. A cofounder of Princeton Review. His idea was to figure

out what the questioner wanted you to answer and then to answer accordingly. I

modified his work to take account of proper answers. Used limits and stops to

see which worked best. Assumed that my limits didn’t get filled when it hit

exactly. And the rest is in the p and l.

Assuming no autocorrelation (the trend not your friend or enemy either) and

no drift up or down, the probability that a stop at L will be hit during the

next N trading days can be calculated in two steps:

(1)Find the probability that price will be below L, N days from now

(2)Double this probability, to account for the fact that the stop is in

effect continuously, not just on the Nth day.

An example

----------

Given: We are at 935, the standard deviation of daily price changes is 9, we

will have a stop at 900 in effect for the next 20 trading days.

Question: What is the probability that this stop will be hit?

Answer:

The price in twenty days has mean 935 and standard deviation 9 * SQRT(20) or

40.25

The 900 level is (900-935)/40.25 standard deviations away from the mean, or

-0.8696

The probability of being below 900 in 20 days is NORMSDIST(-0.8696) or 0.19

Therefore the desired probability (of going below 900 sometime during the

next

twenty days) is 2*0.19 or 0.38

Conclusion

----------

This method is well known and is rigorously correct under the stated

assumptions. I wonder how accurate it is in practice? Any comments?

**Thomas F. Gross, Associate Dean**

**College of Business**

**Cardinal Stritch University (05/28/2003 9:23:23):**

Alex,

It is early in the morning, and I may be tired, I believe your math has some

mistakes. First, the distribution of price changes is probably not normal,

and I think you used a two-tailed test to calculate a one-tailed

probability.

Subject: Re: [SPEC-LIST] The probability that a stop will be hit

I think Alex has it exactly right. In a one-tailed test we are testing

whether the given outcome is at or below the given stop level L at the end

of the period (or trial). Note that a one-tailed test only considers the

cases below L at the end of the period, not those cases which hit L or lower

but later came back above L by the end of the period. These "come-back"

cases are not in the lower tail of the distribution but are usually in the

"fat" middle part of the end of period distribution. By definition these

comeback cases hit the lower stop level L at least once and thus have a

vanishing small probability of showing up in the far away upper tail (2nd

tail).

To see why the probability doubles consider the case when the price is at L

at some time m before the end of period. From that point on the price would

be expected to follow a new normal distribution centered on L and symmetric

about L. It is the symmetry of the normal distribution about its center

point which causes the doubling of the probability. Symmetry of the normal

distribution is essentially what the Reflection principle is all about.

Thomas Gross raises a fair question whether price changes are normal. Most

studies have shown that they approximate normal or log normal with a few

notable exceptions. In particular there are too many small changes which

favors market makers and contrarians. There are also a bit too many

observations out in both tails presumably favoring trend followers. All

things considered the distribution is not perfectly normal but it is very

close to it. So from a practical point of view I believe normal/log normal

distributions are a very workable way to analyze markets.

**Jack S. Fan (5/28/3 7:43 AM):**

Stop Probability: Normal Distribution

The limiting distribution of a binomial distribution is normal. That's

given. But I still don't see how you can formulate prices or returns as

a series of Bernoulli trials. Could you elaborate?

**Tom Gross, Associate Dean**

**College of Business**

**Cardinal Stritch University (5/28/2003 9:23:23)**

Jack,

I am always slow in the morning, so I may be missing your question. It

should be as simple as specifying the size of your series (n) and the

probability of a hit (p). In this case a 'hit' would be a gain. Assuming you

were more likely to have a gain than a loss, you would make p>.5; the

resulting distribution gives you the associated probability for each

outcome, from no gains in the series, to every series event being a gain.

The problem with using this system to generate probabilities for prices is

that you only have two conditions, gain or loss. The simple way out is to

use a long-term estimate of return as the value, and a similar estimate of

the likely loss.

My caveat would be that theory is easy to use, but I am not sure how well

this would work in practice. I tend to defer to people on the list who have

actually tried to test the application of statistics to the markets.

**Jack S. Fan **

**Subject: Re: [SPEC-LIST] Stop Probability: Normal
Distribution**

**Date: 05/28/2003 10:12:59**

Tom,

I'm still somewhat at a lost to see how this would work. Let's be a bit

abstract first, and define a binomial distribution so that we are all on

the same page. Let p be the probability of success (such that there are

only two possible events, success and failure). Of n independent trials,

the probability of k success is given by Bp(k,n).

Now, let is consider my understanding of your application. A gain (or

success) has a probability of 0.5 (just for illustrative purpose). Here

I'm extrapolating of your experiment, so let me know if I'm wrong. Let's

say a gain is 1%, so a 5% move would be 5 gains, so the probability is

given by B(0.5,5,n).

That's all fine and good, non-independence none withstanding. Yet, this

requires some form of calibration of n (in the pervious example, if n=7

and we wish to see the probability of 5% gain with a loss being -1%, we

need 6 success and 1 loss). Furthermore, if we were to take the limiting

distribution as n -> infinity, we see that we have a problem. That would

dictate that with a finite number of successes, and an infinite number

of failures, we want to calculate the probability a certain constant

gain, which in itself depends on both the number of successes and

failures. This is my essential problem with justifying a normal

distribution with binomial.

**From: Peter R. Gardiner **

**Subject: Stops, Stradles, and First Principles of
Market-Making**

**Date: 05/31/2003 1:09:23**

This issue of stops has long perplexed me. All of the highly

touted use their adoption as a litmus test for sanity,
proper risk

control, and even increasing returns. But if the use is
disaster

control, it would seem to me that position sizing is
cheaper, more

effective, and less taxing on the expectation of the trade.
I can barely

count on one hand, as you know, but it strikes me that a
long futures

position is transformed into a long straddle - complete with
premium

paid - as soon as the stop is entered. The upside is
truncated very

massively in exchange for the put value associated with the
stop. This

occurs through the mechanism of the barrier, or stop itself,
on the

number of possible pathways through which the ultimate
(hoped for)

expectation may be achieved. And it seems to me that this is
the case

whether the underlying is log normally distributed or not.
We know from

basic Black Shoales that the distribution -let's say its
symmetric, as

in the normal case - is produced by an infinite number of
possible

pathways, each with its different probabilities. The paths are
not

straight lines; they go everywhere; the distribution, on the
other hand,

is produced by the end point. If I agree to take your position off your

hands at price B below your entry S, what is my
compensation? It must be

the pathways at my end of the distribution, whose end points
may result

in a winner. In options, this translates into the vig. And
the other big

cost is that now not only is the number of paths reduced,
but the order

in which they occur is critical: if the paths producing my
profit don't

come before the stop is hit, I lose my bet. This is the same
as the

compound probability problem: a guy wants to buy a stock,
"knows" its

going up with 80% chance, but only 50% within the next three
months.

Result - an 80% becomes a 40% bet. The stop must reduce the
value of the

trade every time on average, but exactly how much in any
individual

case, we cannot know.

The other
element which is the killer comes from Osborne.

The structure of proper market-making activities assures
that once the

stop is entered, theoretically the entire distribution (sum
of all

possible pathways, normal or not) must change, for all the
reasons he

describes in his market-making section. This is required by
rational

market-making, and since irrational market-making is
penalized with

extinction, it is a guarantee.

Therefore, the only real way to achieve downside protection

is to be small enough when you are really wrong; and to
achieve great

profits, to be big when you are right. (Will Rogers and Mark
Twain,

traders.) I guess that's where the counting comes in. One,
two, three...

**Philip J. McDonnell **

**Subject: [SPEC-LIST] Stops, Straddles, and Market-Making**

**Date: 05/31/2003 14:11:09**

Peter Gardiner made the incisive observation that making a
position trade

with a stop below it is analagous to taking a straddle
position. His

insight stimulated me to realize that an alternative to the
use of stops is

simply to take a call position. A long call is equivalent to long

stock/futures with protective put (give or take some
interest), but simpler

to execute.

A long position trade with a stop will eliminate the paths
that lead lower

than the stop as well as the comparable number of paths
leading to a

comeback. A call
slightly differs from this in that it allows the trader to

ride out any downside excursions during the life of the
option. Thus all

the comeback paths are still available if desired. A call does offer a

guaranteed limit on loss equal to the premium paid. This is
in contrast to a

stop which is not guaranteed to be executed at your price.

Of course you pay something for these benefits to the call
writer. You pay

daily as the call premium slowly erodes away. Because of time erosion

buying a call won't improve your probability of an eventual
profitable

trade. But it may
improve the odds that you avoid bankruptcy.

I would rank the various money management techniques from
best (#1) to worst

as follows:

1. diversification
/ small position sizes - best & cheapest

2. call purchase /
(position with protective put)

3. position trade
with stop - backtested with stop!

4. position trade
with stop - Not backtested with stop!

Note that backtesting stops requires OHLC daily data not
just a simple look

to see if the close was below your stop. You need the open data to account

for gap opens which trade through your stop. The high - low data tells you

if you got stopped out on an intraday move. It's a bit trickier than it

might seem.

**Faisal:**

At: 5/30 19:08

SURPRISED THAT THERE IS SO MUCH DEBATE ABOUT CALCULATING THE PROB OF HITTING THE

STOP. HAVING A STOP IN PLACE IS REALLY LIKE OWNING A KNOCK-OUT CALL (FOR LONGS)

OPTION POSITION WITH THE STRIKE AT ENTRY PRICE AND A KNOCK-OUT LEVEL (BARRIER,

OR STOP) BELOW THAT. THE FORMULA FOR CALCULATING THE PRICE OF THIS OPTION, AND

AS A BY-PRODUCT, THE PROB OF HITTING THE BARRIER ARE QUITE WIDELY AVAILABLE.

ALTHOUGH I'M NOT AN EXPERT ON THIS SUBJECT, IT SEEMS TO ME THAT THIS QUESTION

SHOULD BE EASILY SETTLED. IF YOU TYPE "KNOCK-OUT OPTION HELP" ON BLOOMBERG,

THERE IS EASY ACCESS TO A PAPER BY DR. ERIC BERGER OF BERGER FINANCIAL RESEARCH

LTD. WITH THE NECESSARY OPTION PRICING FORMULAE.