Why favorable R:R?

[intradaybill]

Quote from Mr J:

This statement makes no sense to me. If a return is restricted, it's by poor trading, not the risk to reward. Good returns can be achieved at all sorts of ratios, it comes down to the trader.

No, the trader does not rule the market, volatility does. Try a ratio of 1:10 in this market and see if you can make any returns.

Good returns can only be achieved, if at all, when the ratios are compatible with the relevant ATR in the trading timeframe chosen.

 

[Rabbitone]

Quote from Mr J:

If you practice sensbile capital management, the risk to reward is irrelevent. All that matters is that the trade is +ev and appropriately sized.

Reward to risk (R:R) from the automated trading perspective may shed some light on this matter. To start I agree the use of +ev along with capital management is an excellent method to track performance management especially automated strategy trading. But, a +ev number provides no guideline as to how automated strategies may perform going forward in optimizations. Whereas, reward and risk ratios provides some insight into how an automated strategy will work in live trading and indicate what the strategy is attempting to do during trading.

When I first traded a low risk reward (about 1.42:1) swing strategy about 6 years ago (a stock in the daily time interval) I got some surprises. At first the higher win rate of the strategy appeared to be like digging gold even with the losses. The losses were bigger, but what the heck the high win rate made up for it. Then it happened, the volatility shifted and those bigger losses gave me a big drawdown; one that gave me a loss overall. I traded through it and came back profitable after a stretch. But the outcome left me puzzeled. Later I compared this low risk reward swing strategy with another strategy I traded. What I found is the +ev performance was lower for the low R:R strategy.

A few years later thinking the original low R:R was a fluke I tried a different strategy with low risk reward (I couldn’t locate the R:R number in my log). This one also had a big draw down like the first low R:R. but in a different way. This time volatility went flat in the stock and it slowly built up the big drawdown. It like the first was a poorer performer.

What I found from this and other trading is using R:R as a guideline keeps me away from selecting some of the wrong settings for my automated strategies. I now track R:R, but its no big deal. It is just another characteristic of the strategy to be accumulated like slippage and commission to determine what works. Then I write them my trading plan as guidelines.

The only reason authors recommend such high R:R numbers is for beginning traders. By stating they should have 3:1 R:R they are telling the trader to trade trends. This high number will keep the trader insulated from the big draw downs and keep them in trading through congestion.

I also found R:R is associated with trading style. I trade short term and long term so my R:R are all over the map (from 1.72:1 to 3.65:1) . But the bottom line is R:R is just a guideline to tell you if the strategy setting you pick or your trade setups have a higher probability of working for your goals and objectives.

 

[ronblack]

Quote from Mr J:

If you practice sensbile capital management, the risk to reward is irrelevent. All that matters is that the trade is +ev and appropriately sized.

+ev is a metric and its value depends on R:R ratio and success rate. If Y is a function of X, you cannot say that all it matters is some value of Y>0 and X does not matter. Y depends on values X takes.

what do you think?



Ron

 

[Rabbitone]

RonBlack and Mr J.

I believe you both are missing the point to be make about EV+ and R:R. They are both excellent statistical guidelines for trading but nothing more. Traders must first learn how their strategies apply to market conditions, price direction and price volatility. Then and only then are statistics useful (see Darrell Huff “How to lie with statistics”). When you trade for several years you learn this the hard way.

Let’s look at a case that is dear to me. An automated system I’m currently trading (AAPL daily) has a for the last 13 Trades over the last 5 months had a low EV+ and poor R:R. By most Harris or EV+ standards it should be shut down because the system is currently performing marginally. However, when I review the systems longer term EV+ and the R:R for the 3+ years I have run the system the picture totally reverses. The EV+ is above the line and the R:R above 3:1 in this intermediate term trend following system (It’s part of the reason I’m retired). Am I concerned that my current system is low on EV+ and poor in R:R in the short run? No, the system is well within the expected drawdown.

What you learn is many of the wonderful linear mathematical equations (love math it was my major 40 years ago) and statistical derivatives apply in the long run but not the short. Applying strict statistical rigor in the short run may toss many a good trading strategy on the trash heap. The reason being the sample sizes are too small to draw statistical inferences.

The point is if you try and apply Michael Harris to “real’ trading in constantly shifting markets in the short term you may be in for some surprises. Why writers like Harris apply linear mathematical statistical equations to markets that require calculus or advanced calculus is beyond me. Yet we continue to be inundated with linear statistical rules of thumb.

My rule is fit the strategy to the market. I built a beautiful volatility trading system in 2007, but it performed like hell then. Using EV+ and R:R. standards it was toast. I should have never have used it again. However, in the last year that system is providing the majority of my profits. The EV+ and R:R are wonderful. I must have done my math wrong in 2007. Maybe I should read Harris again……

That’s what I think.

 

[trader3cnd]

Quote from Rabbitone:

Why writers like Harris apply linear mathematical statistical equations to markets that require calculus or advanced calculus is beyond me.

I think his approach is data mining and that is highly nonlinear way of finding systems.

 

[Mr J]

Rabbitone, I made no mention of timeframe. I simply said that as long as risk&reward isn't at an impractical ratio, it's irrelevant, and that what matters is whether the trading strategy is profitable. +ev has nothing to do with statistics, it's a mathematical edge, and it is required. It's not a guideline, it's a fact. If you know a trade is +ev and the profit is worthwhile, then you place it. If you know a trade is -ev, you don't place it. Of course, we don't usually know whether the trade is +ev or not, so what matters is that the overall strategy is +ev in the current market.

As for your case, your just playing with timeframes. The system sounds marginal or -ev over the shorterm, and +ev over the longterm. Obviously it's just having a poor run or it's no longer a profitable strategy, i.e. we don't know whether it is truly +ev (without analysing the trades). It doesn't make ev any less important, it just shows that ev is usually unknown.

EV applies over all timeframes.

"The reason being the sample sizes are too small to draw statistical inferences."

Just because we can't determine ev, doesn't mean it doesn't exist. It exists for every trade and every sample, so ev always applies.

"I built a beautiful volatility trading system in 2007, but it performed like hell then. Using EV+ and R:R. standards it was toast."

You weren't using "+ev" at all, you were using an estimated range of ev. You didn't know the ev. Perhaps at the time it was not +ev, or perhaps it was +ev but the negative side of variance.

You seem to be misunderstanding the difference between estimates of ev, and ev itself.

 

[Mr J]

Quote from ronblack:

+ev is a metric and its value depends on R:R ratio and success rate. If Y is a function of X, you cannot say that all it matters is some value of Y>0 and X does not matter. Y depends on values X takes.

what do you think?



Ron

I meant the ratio is irrelevant as far as determining which trades are placed, assuming that ratio is acceptable for the trader.

My point is in reply to the suggestion that trades should have a certain R:R range to be profitable. My point is that it is not the range that matters, but whether the trade has perceived +ev, i.e. whether R:R and the strikerate are a profitable combination.

Perhaps a better way to make my point would have been to state that R:R itself does not determine the profitability of a trade.

 

[Rabbitone]

Quote from trader3cnd:

I think his approach is data mining and that is highly nonlinear way of finding systems.

When I hear people, like Harris, use the term Data Mining in the context of trading systems I want to puke. I retired as a Database Administrator supporting Data Mining on large scale IBM Data Warehouses. With the SQL software wonders like Hummingbird we used to build SQL that would scan a terabyte of data with 200 to 250 million transactions. From this scope of work statistical inferences about pattern recognition could be gleaned. Putting many of these runs together this pattern recognition was called business intelligence.

I am not aware of Harris or any one else building business intelligence on this scale. When I hear every tick for the last 40 years for every stock has been data mined and the patterns check for accuracy then I will give his method some credence.

Patterns are a linear approach. Don’t get me wrong trading patterns are wonderful. I love the approach. Sorry, but they are not data mining. But, they are the primary area I use in all my systems. I have traded with them for more than 20+ years. However, they require the markets to directly support them in a linear fashion. Without the correct market characteristics patterns and strategies will fail. The EV+ and R:R must support them in their linear fashion.

 

[Rabbitone]

Quote from Mr J:

Rabbitone, I made no mention of timeframe. I simply said that as long as risk&reward isn't at an impractical ratio, it's irrelevant, and that what matters is whether the trading strategy is profitable. +ev has nothing to do with statistics, it's a mathematical edge, and it is required. It's not a guideline, it's a fact. If you know a trade is +ev and the profit is worthwhile, then you place it. If you know a trade is -ev, you don't place it. Of course, we don't usually know whether the trade is +ev or not, so what matters is that the overall strategy is +ev in the current market.

As for your case, your just playing with timeframes. The system sounds marginal or -ev over the shorterm, and +ev over the longterm. Obviously it's just having a poor run or it's no longer a profitable strategy, i.e. we don't know whether it is truly +ev (without analysing the trades). It doesn't make ev any less important, it just shows that ev is usually unknown.

EV applies over all timeframes.

"The reason being the sample sizes are too small to draw statistical inferences."

Just because we can't determine ev, doesn't mean it doesn't exist. It exists for every trade and every sample, so ev always applies.

"I built a beautiful volatility trading system in 2007, but it performed like hell then. Using EV+ and R:R. standards it was toast."

You weren't using "+ev" at all, you were using an estimated range of ev. You didn't know the ev. Perhaps at the time it was not +ev, or perhaps it was +ev but the negative side of variance.

You seem to be misunderstanding the difference between estimates of ev, and ev itself.

Mr J this gets better by the minute…

+ev has nothing to do with statistics, it's a mathematical edge

So if

Expectancy EV = (% winners * Avg win) + (% losers * Avg loss)

Then the averages and % are not statistics in expectancy. I though an average was a statistic. Boy I learn something new every day. And If I divide %winner by %loser wouldn’t I get something like an R:R. Well that ones out too I guess.

As for your case, your just playing with timeframes. The system sounds marginal or -ev over the shorterm, and +ev over the longterm.

Compared with

ev, doesn't mean it doesn't exist. It exists for every trade and every sample, so ev always applies

Now you got me confused? One minute the EV is irrelevant because of time frames and the next minute it always applies? Nice slight of hand.

You weren't using "+ev" at all, you were using an estimated range of ev. You didn't know the ev.

and

You seem to be misunderstanding the difference between estimates of ev, and ev itself.

Now I’m dizzy. So please explain to us when does Ev stop being insignificant and start being significant? Do you mean if a test of significance gives a p-value lower than the á-level, the null hypothesis is rejected? When does this happen when trading a strategy? If I optimized over 80 trades how many trades must I do in live trading before EV means anything significant? What happens If I go broke first?

Real examples please no more hypothetical clap trap.

 

[Mr J]

Perhaps I'm wrong, but I'm pretty sure ev we're talking about probability here, not statistics.

"Compared with"

Did you actually read that paragraph? I never said it was irrelevant at all in the first quote. In fact, I stated that it had not lost any importance.

"So please explain to us when does Ev stop being insignificant and start being significant?"

As above, I never said ev was insignificant.

"If I optimized over 80 trades how many trades must I do in live trading before EV means anything significant?"

This has nothing to do with my posts. You're referring to perceived ev, while I'm refering to the true ev. You must not have understood my posts, or I must be misunderstanding your's. Your taking this discussion to a completely different place that has nothing to do with my original point, which was that a particular R:R ratio is not needed to make good trades.