I definitely agree with you that the notion or risk:reward can be very misleading but you're examples are throwing a wrench in my gears. Are not these two strategies in opposition to eachother? Where the first would aim to capture mean reverting tendencies, the second aims to capture an expansion of volatility. Do you think the outcome would change significantly if for the first strategy you bought above 80, and sold below 20?
"If I agreed with you, we'd both be wrong."
The following user says Thank You to addchild for this post:
I don't know why you have started the thread. The avarage RR Ratio of a system is only one value of many to describe a system. Nobody would only look at the RR Ratio and say: "What a cool system". The RR Ratio as a result of a single trade can describe the risk and the reward with one number. I do a lot of statistics with these RR values and i'm really fine with that and thereby i could improve my system a lot (while taking care of the drawdowns).
Because you said "infinite" RRs. I think the RR Ratio could only be used at systems with a fixed stop loss. If the stop loss isn't fixed then you can't make your calculations. My system has a initial fixed stop loss (1R), a trailing stop loss and an initial target zone where the target trailing stop begins. Therefore i know the projected RR Ratio of a single trade at the setup bar.
Last edited by Koepisch; October 19th, 2012 at 03:58 AM.
Both systems had a 10:1 RRRatio like you said. I never build the mean of RRs because thats meaningless. I work with RRs to have a clue (with other metrics) about the risk in a system, which resulting into the max drawdowns and which defines my money management settings. If i would mix 2 systems then the risk will be hidden and concealed.
All my performance metrics are based of gained R's. Therefore i can't accidentally optimize to max. profits without considering the risk. My automated system starts to trade at larger then 1:1 RR Ratio. For descretionary trading i start never below 1:2 RRRatio. But this all belongs to your strategy.
Every hypothesis breaks down in the real world when any parameter approaches 'infinity' or '0', so any deductions made on these are meaningless the closer you approach these values in the real world. I can have a strategy that buys the ES with a target of 1 tick, and a stop loss at $0. That almost guarantees that I will be profitable on a large number of trades, but who would do that in real life? How many traders could withstand the possible un-realized drawdowns that would likely occur? What are the chances you'd get a margin call at some point which would pretty much end the experiment, and force your point of ruin?
Risk/Reward by itself is meaningless. It is a one parameter of a 'Trading Method', just like your chosen stochastics entry strategy is another parameter of that trading method, along with various other parameters.
I don't know about proof writing, sweeping books, or writing programs that know where to place stops. As to the question, let's say I have a strategy, which in backtesting says that, based on my entry method, I reach 20 tick targets 90% of the time before my stop of 10 ticks would be hit. So, I set my target to 20 ticks and my stop to 10 ticks, which is 2:1, disregarding slippage, commissions, etc. Given that scenario, I'm not sure what informed participants, market makers, and skewed bid/offers would affect my strategy, nor if there would be anything I could do about those things.
Last edited by monpere; October 19th, 2012 at 03:00 PM.