General Topics
Technical Topics
Brokerage Firms
Community Lounge
Site Support

# Newbie expectancy question

Discussion in 'Risk Management' started by nravo, Dec 15, 2007.

1. ### nravo

Been reading Thorp, Turtles, et-cetera, and I have a question about positioning-sizing formulas and the use of expectancy. How do you determine the win-loss percentage of your system or a particular trade? Back-testing? Guessing? If back-testing, how far back? Best software to backtest, say stocks hitting 20-day highs and lows?

2. ### bathrobe

Many systems have a win/loss percentage of less than 50% but are profitable due to letting the profits run and cutting the losers, which I am sure you have already heard. I recently had a similiar question and was directed to:

they have a 30 day trial and it is an excellent site.
If you plan on developing a trend trading system you will most likely get signals in whipsaws and have significant drawdown. I have heard these systems have not done as well lately (I can't tell you for sure, I am not a trend follower)
for more info on position sizing look into this site as well:

http://www.surinotes.com/index.cfm?CSEL=2200

I do not know what you trade so I cannot be of more help.

3. ### ronblack

Read the article "Relation of Expected Gain to Kelly Formula" by Michael Harris:

I recommend a fixed risk percent method. Expectancy is very hard to determine accurately. I use a fixed 2% risk per trade to determine position size and it does the job.

Ron

4. ### nravo

Thanks. Interesting. And here is my really really naive questions (apologies): Backtesting is a requirement for determining the % of winning trades to losing, et-cetera, right? So, for discretionary trading developing expectancy isn't really possible is it?

5. ### ronblack

Not right. What is more important is the calculation of the actual values after you employ your method in actual trading, whether it is mechanical or discretionary.

Ron

7. ### nravo

Couldn't get that link to work.

works now..

9. ### nravo

Got it; thanks. I've seem the formula elsewhere, and without the digressions. I am wondering now about back-testing for non-pattern strategies. For instance, how could one back test every stock split on NYSE, day of announcement and average gain until ex-date and EOD on ex-date. Any cheap software do that? Wealth Lab? Or is most retail back testing software made only for technical analysis and doesn't do other types of strategies?

10. ### dozu888

thorp = garbage
turtles = almost garbage.