[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Futrs: Trading Stops/Risk/Reward



PureBytes Links

Trading Reference Links

Jeff writes:

> For any given trade, how do you determine the probability of various
> returns?  R Slupsky's post implies past experience, but I was wondering if
> anyone's published rigorous estimates of probability distributions
> associated with various technical indicators, events, and markets.


Seems to me the only possible way to assess the distribution of returns of
a trading system is to take a bunch of samples!  All the usual caveats
apply: gotta sample over a wide variety of market conditions, gotta get
enough data (statistics says minimum 30, I like 150 or more), gotta look
at the big picture and make sure market conditions aren't "skewed" in
some way.  

Example: ever wonder why so many systems show great results in coffee?
Surprise!  Coffee has gone through several HUGE up and down moves in the
last 5 years.  Many system make great money off that.  So if you run such
a system now, what are you really betting on?  That another huge move
occurs.  Is the probability of that really the probability of the returns
of your backtested system?  No.  (This is an easy mistake to make.)

The other obvious example is long only systems in stock indices that use deep 
stops and are tested against the last several years of stellar bull market
behavior.  They make boatloads!  And running them from here on may
get similar results...or may not.

All these are examples of sampling from a price distribution that is
"funny" and may or may not have similar "funniness" in the future.

With all this said, I still am of the opinion that no trader should ever
place a single trade, risk a dime, until they have a quantitative assessment
of the probabilities around the trade, the expected return, the probable
maximal loss, the win/loss %, the pessimistic return ratio, etcetera, and
really conclude they've got a darn good shot at making money.  That is, they
are confident their system has a positive expectation and a very good positive
expectation at that.  Such an assessment can only be based on a sample, 
acquired through careful backtesting, or rigorous simulated trading, or lengthy
paper trading (and all with factoring in realistics costs/slippage).  To do 
otherwise is guessing, or is treating trading as entertainment (i.e., playing at
a casino knowing that it's a losing proposition, because it is "fun").  (That's 
just my opinion, of course!)

As for what kind of figures are good, that's a bit subjective (although "makes
money" isn't and is a good start; most are trading systems that don't).  It all
depends just how good a set of systems you can come up with.  For 
some a return ratio (avg win * prob of win) / (avg loss * prob of loss) of
2.0 might be darn good, for others, very weak because they have systems with
ratios in the 4's or higher.  (Again, caveat: systems that will never
work in real-time can have terrific #'s based on sampled results!  Here I'm
assuming comparison of stable, reliable systems.)

I like systems with return ratios of over 3, with pessimistic return
ratios of over 2.5, with average result per trade (not average win, average
overall return across all trades) of over $800 (assuming single contract).
I also like systems with very large optimal f values (0.7 or higher), so that 
they will compound well (optf is an indication of the evenness of distribution
of wins, versus long periods of winning then long periods of losing).  Of
course I like a smaller maximal loss ever experienced, but am willing to
compromise here if the other figures are good.  Lastly, given systems with 
equal return ratios, I prefer the system with the higher percentage of winners,
because my calcs tell me they have less risk (of loss of bankroll).  Why do I 
like systems like these?  Because I have a few, they're the best I've come 
up with, and I'm only going to trade my best!  Maybe in the future I'll have 
better and my requirements will go up.  

Final word: if you don't know the stats of the system (approach, method,
whatever it is you trade, even if "discretionary") you use, you outta assume 
it's a negative expectation system (i.e., a losing system that, should you 
trade it forever, will consume every cent you own), and stop trading it and 
start measuring it.  Again, JUST MY OPINION, I believe it is NOT widely held or 
shared.


-Kevin