Calculating Risk/Reward Ratio
- Thread starter catmango
- Start date
Let's come back to the basic:
What is the purpose of trading and of any business in general ? it is to achieve a positive outcome on long runs, that is to say on average probability * win/loss must be > 0.5.
In statistic decision field one distingish the true value and the estimated value of parameters. The true value can never be known for sure and must be estimated by data sampling and knowledge.
So how do you estimate the probability and the win/loss ? The most common approach is to say that at any time a person without specific knowledge knows nothing about the direction of the market so that probability = 0.5. So if the probability is equal to 0.5 then the only mean to achieve a ratio better than 0.5 is to have a win/loss ratio that is consistently over 1.
I could develop and write a book so for short I will just pinpoint about what one should be careful about: in statistic field, one know that it is very difficult to achieve a sampling without biais especially in human field. So the probabilty of even 0.5 could in fact be overestimated at least for some people for as one with experience knows, market goes againts the majority so that it could even explain why a monkey could do better than some human
. Another pundit is believing that the win/loss ratio is consistent. It is consistent for a rather long period of time then can degrades severally because of market "efficiency" which is like the efficiency of a casino that let you win most of the time and then ripped you off.
So the problem is not about how to calculate the win/loss ratio, it is about how to determine a consistent win/loss ratio estimation for probability law only valid if consistency is achieved. This is a question that in Statistical Process Control in Quality field one try to answer. Many people including ingineers don't grasp probability field only calculation aspect whereas it is rather an epistemological problem, that's what Walter Shewart, Statistician of Bell telephone labs and father of Quality Engineering has criticised about statistical use in PRACTICAL industry before he came with his new approach which he often refers to as Philosophy of modern statistical control. Even today in quality enginering newcomers have nver cope with the spirit of the father and commit the same error of approach.
What is the purpose of trading and of any business in general ? it is to achieve a positive outcome on long runs, that is to say on average probability * win/loss must be > 0.5.
In statistic decision field one distingish the true value and the estimated value of parameters. The true value can never be known for sure and must be estimated by data sampling and knowledge.
So how do you estimate the probability and the win/loss ? The most common approach is to say that at any time a person without specific knowledge knows nothing about the direction of the market so that probability = 0.5. So if the probability is equal to 0.5 then the only mean to achieve a ratio better than 0.5 is to have a win/loss ratio that is consistently over 1.
I could develop and write a book so for short I will just pinpoint about what one should be careful about: in statistic field, one know that it is very difficult to achieve a sampling without biais especially in human field. So the probabilty of even 0.5 could in fact be overestimated at least for some people for as one with experience knows, market goes againts the majority so that it could even explain why a monkey could do better than some human
. Another pundit is believing that the win/loss ratio is consistent. It is consistent for a rather long period of time then can degrades severally because of market "efficiency" which is like the efficiency of a casino that let you win most of the time and then ripped you off.So the problem is not about how to calculate the win/loss ratio, it is about how to determine a consistent win/loss ratio estimation for probability law only valid if consistency is achieved. This is a question that in Statistical Process Control in Quality field one try to answer. Many people including ingineers don't grasp probability field only calculation aspect whereas it is rather an epistemological problem, that's what Walter Shewart, Statistician of Bell telephone labs and father of Quality Engineering has criticised about statistical use in PRACTICAL industry before he came with his new approach which he often refers to as Philosophy of modern statistical control. Even today in quality enginering newcomers have nver cope with the spirit of the father and commit the same error of approach.
The question is not about win/loss ratio; it's about risk/reward ratio.
Statistically, you should be able to take a small sample of consecutive trades and come up with a consistent percentage. That same percentage should be consistent with another sample of consecutive trades.
In other words I should be able to take my first 200 trades and my last 200 trades (for any sample > 400 trades) and calculate the risk/reward ratio for both sets of trades ... and I should be relatively consistent.
rolegario,
Ambushillbilly did a good job ...or the best job of describing that 3 to 1 ratio...I think he is pretty much right on ....the only thing I might add is that....when you are looking to trade, you want to find a set up which will give you the most profit with the least amount of risk....sometimes it is possible to have even a better ratio....so in essence then the real thing you are trying to do is find those trading set ups that will have the best ratio you can find.
Ambushillbilly did a good job ...or the best job of describing that 3 to 1 ratio...I think he is pretty much right on ....the only thing I might add is that....when you are looking to trade, you want to find a set up which will give you the most profit with the least amount of risk....sometimes it is possible to have even a better ratio....so in essence then the real thing you are trying to do is find those trading set ups that will have the best ratio you can find.
The later "How many contracts or shares." as a percentage of equity.
If you have a working system, Tharp may set you back a few months but it's nessary.
It's not that Simple, there is more to it. Adding to an established trend is a very easy thing to do. Pyramiders may be takeing the same risk for marginaly less profit per contract or share with easier profits. Pyramiders may not be seeking the most profit with the least risk but easier profits.
-ooO-(GoldTrader)-Ooo- 
If you have a high success rate then 2:1 will do fine.
Just a side point.....When massaging the numbers, I have noticed that by increasing the stop verses decreasing the target will increase the win rate and net more bottom line faster. This is on the ES and laborously and methodically manipulating the numbers the old fashioned way on an Excel spreadsheet.
I have seen systems with a high winrate ....up above 80% with an upside down risk/reward ratio....like 0.18:1 (target/stoploss). But who would trade like this....as it would require perfect execution and a quick trigger finger or IB bracket trader.
Michael B.
Just a side point.....When massaging the numbers, I have noticed that by increasing the stop verses decreasing the target will increase the win rate and net more bottom line faster. This is on the ES and laborously and methodically manipulating the numbers the old fashioned way on an Excel spreadsheet.
I have seen systems with a high winrate ....up above 80% with an upside down risk/reward ratio....like 0.18:1 (target/stoploss). But who would trade like this....as it would require perfect execution and a quick trigger finger or IB bracket trader.
Michael B.
I think we tend to beat some of these concepts into the ground sometimes in search of an edge? In short term trading I've heard people assign the term risk/reward to exactly what the first response was to this thread: stop out at 1 handle with the hope of making at least 3. Or hoping to look back on the day and find that your winners outpaced your losers by this much?
I don't think the term fits here but its used anyway. The correct place for the term is in much longer term plays like those undertaken by hedge funds. In event-driven type situations you look at the RISK/REWARD. Notice RISK comes first. Therefore, with an expectation of an event (merger, spinoff, dutch auction etc ...) at say 90% probablility of it happening; you are willing to RISK $3 to make $1 in profit or REWARD. With a 90+% probablitlty a 3:1 risk/reward would be an excellent investment for a longer term play.
That is really the ONLY correct usage of the term.
I don't think the term fits here but its used anyway. The correct place for the term is in much longer term plays like those undertaken by hedge funds. In event-driven type situations you look at the RISK/REWARD. Notice RISK comes first. Therefore, with an expectation of an event (merger, spinoff, dutch auction etc ...) at say 90% probablility of it happening; you are willing to RISK $3 to make $1 in profit or REWARD. With a 90+% probablitlty a 3:1 risk/reward would be an excellent investment for a longer term play.
That is really the ONLY correct usage of the term.
I use R/R of 3. When I got 1R of profit I move stop to BE and wait for my target or BE. This gave 22% wins, 44% BEs, and 34% losses. Overall the Profit Factor = 2. On the surface, very good right? But the thing is my avg loss is 0.1, my avg win is 0.3 I daytrade SPY. so my avg. trade is 0.22*0.3+0.44*0-0.34*0.1 = 0.066+0-0.034 = 0.03. So my avg trade is +0.03 on SPY. But commission/trade is 0.02. So on the surface I got a great little system but when we add commissions it goes to hell. What should I do?