Thanks for the note. I doubt that its because of me that your enjoying your profits. But I am glad that you are. I just know my approach works for me and if it helps others that’s great. I am glad that there are people here in the community who are actually looking at my work. thx.
You do great work and i am proud to witness it.thx
I love success stories because they give many hope. Steve - could you please outline what exactly made you profitable following the nuggets gained from MeiHua. Also how long have you been profitable for now?
MeiHua,
First off awesome thread. Very interesting indeed.
However I have a question I am a little confused about. In your second post you seem to say you are making a hard 150 stop and 450 take profit. This is exactly a win/loss ratio of 3.0. However, your take average win/loss ratio is 2.943…
How is this possible? If you always stop at 150 and always take profit at 450, why isn’t your win/loss ratio exactly 3.0?
PS. I’m not accusing you of doing anything wrong, I was just curious what I was missing. How you are taking profits a factor of 2.943… higher than stops if you are always exiting at 450 vs. 150. At first I thought it was some spread thing but 0.943*150 is 8 pips which is two big for an EUR/USD spread so I think it must be something else I am missing.
Thanks.
You’re on the right path, it has to do with expectancy short fall, which can be caused by other things not just spread, execution delay vs realized signal, slippage on order (i am using market orders here so i am more accurate and reliable results), as well as no fills. However there are absolutely times where spread is larger than 8 pips, 15 pips in news, holidays, or the like is not impossible. But basically even with an automated system it is unable to achieve the exact number of 3:1 ratio. Nothing ever will be exact, so in my estimations I err on the conservative side. So even with a 2.94 (close enough to 3) ratio you still find that type of behavior.
MeiHua are you versed in market microstructure?
I have a decent grasp of it. Both from an electronic stand point and pit. I have read both Harris and Ohara’s works as well.
Testing procedure:
I used the same data and same dates as before. I used a 20 period ATR as the measurement of volatility in 5 minute increments. very straight forward. Of course this is on EUR USD.
I am going to show the volatility map of the trading day. Given by the average ATR of every 5 minute increment. Now I could have chosen a much more robust volatility estimation tools (ARCH, GARCH, SABR, HESTON etc) however those would add additional complexity and make it much more difficult for viewers at home to replicate. So we are going with tried and true ATR. However given its lagging nature as all moving averages, i have added a displacement back 10 periods in order to show a less lagged result.
Overall volatility is extremely important to any trader. It is what gives us the opportunity to capture alpha, which is what most of the people here are trying to do. Although there are traders and strategies which can run independent of directional volatility or at least attempt to neutralize it, but that is beyond the scope here.
Given that volatility must be present for us to have opportunity to make good trades. Its important to know when that will occur.
for some reason the attachment upload here at BP was not working so i had to use tinypic.
Well looking at the actual ATR, we have 2 peaks the first at 0300 and the next one at 0900. I think its fairly clear to see that after about 1000 activity is going to fall off.
But I actually don’t think that the actual ATR is as representative of whats actually going on. Which is why I have added the displaced ATR which moves back 10 periods. So it reduces the “lag” by 10 so to speak, since it takes 20 periods in order to make the calculation.
My conclusions on ATR: Given either the actual or displaced ATR, it is abundantly clear that volatility starts to enter in during late Tokyo early London, peaks around the London open then again at the New York session. After mid New York volatility tapers down to the lowest point. So looking to trade in late new York, or early Tokyo is probably not going to be as fruitful as say waiting until early London or early new York. This all ties in very well with the 1st intraday anatomy post, which shows that those are key times for highs and lows to be created. Which makes intuitive sense, volatility is necessary in order to create extremes of the day.
Firstly, interesting thread, thank you very much for sharing your knowledge, studies with us for free!
Secondly, do not get upset about newbies, who ask you to advise them when to enter, close, prove your studies etc.(they do not understand that forex is not 100% ratio win game, and keep looking for holy grail)
I think majority found this thread very interesting and understand purpose of it.
Like always pardon my English :).
Again, thanks for the latest analysis MeiHua. Have you tried looking at Asian-based pairs like the AUDUSD or the AUDJPY and their volatility profile?
Yes I have, but not fully Asian pairs or high exotics like AUDSGD. AUDUSD and USDJPY are examples of ones I have looked at. They are not as volatility inverted as you would expect but then again they share the USD base as the EURO above which gives the NY its “power” so to speak.
Full Asian pairs like SGDJPY or AUDJPY would be an interesting aside but I am not sure we would see a full on inversion either. Like moving the entire volatility map to peak in tokyo with US being the minor hump. IMHO I think the US news and EURO zone news carry too much weight but that’s just speculation at this point.
Can this strategy really be applied for real or just to inspire us to get strategy with good risk reward ratio ?
And the main point in here is about the fact that good risk reward ratio is important right ?
I am not making any trading recommendations or advocating any trading strategy. What is posted is purely for educational purposes and covered by the following disclaimer.
CFTC RULE 4.41 - HYPOTHETICAL OR SIMULATED PERFORMANCE RESULTS HAVE CERTAIN LIMITATIONS. UNLIKE AN ACTUAL PERFORMANCE RECORD, SIMULATED RESULTS DO NOT REPRESENT ACTUAL TRADING. ALSO, SINCE THE TRADES HAVE NOT BEEN EXECUTED, THE RESULTS MAY HAVE UNDER-OR-OVER COMPENSATED FOR THE IMPACT, IF ANY, OF CERTAIN MARKET FACTORS, SUCH AS LACK OF LIQUIDITY. SIMULATED TRADING PROGRAMS IN GENERAL ARE ALSO SUBJECT TO THE FACT THAT THEY ARE DESIGNED WITH THE BENEFIT OF HINDSIGHT. NO REPRESENTATION IS BEING MADE THAT ANY ACCOUNT WILL OR IS LIKELY TO ACHIEVE PROFIT OR LOSSES SIMILAR TO THOSE SHOWN.
NO REPRESENTATION IS BEING MADE THAT ANY ACCOUNT WILL, OR IS LIKELY TO ACHIEVE PROFITS OR LOSSES SIMILAR TO THOSE DISCUSSED WITHIN THIS SITE, SUPPORT AND TEXTS. OUR COURSE(S), PRODUCTS AND SERVICES SHOULD BE USED AS LEARNING AIDS ONLY AND SHOULD NOT BE USED TO INVEST REAL MONEY. IF YOU DECIDE TO INVEST REAL MONEY, ALL TRADING DECISIONS SHOULD BE YOUR OWN.
I am assuming that your referring to my 1st posts. in which I am trading random entries . Is it really logical to trade randomly? Do you believe that random entries are better than your trading skill (alpha generation)? If so then I think you need to seek professional advice and re evaluate from self directed investment.
I know its trite and a lackluster answer but you have admit that if you are beaten by a coin toss you need some guidance. The general theme of that post and the reason its there is to show that high R:R ratios have an edge in the markets, given certain guidelines as outlined in the post. In my opinion high RR ratios are important to alpha generation (your ability and skill to generate trading profits). How you utilize that information is up to you. Bottom line, yes I think that high RR ratio is important, and strategies that use high RR may have a better probability of alpha generation than ones that do not.
U completely nailed the summary of your thread.thanx.
Thank you for the excellent information! I found this thread via a link over at Forex Factory.
Interesting thread, too bad it seems to catch so little interest. MeiHua, would you mind re-posting the missing images from your first post somehow? I’m curious what your results looked like and I think some of the images expired at the hosting website you used…
Since the 3:1 and 4:1 are missing, (and apparently a few people have checked out this thread looking for it) I decided to recreate this statistic in a very crude fashion. Although the results are not identical I think the point, very clearly, remains.
First, the method:
I took EU data for the past 10 years or so (3610 days of trading). I calculated a 20 SMA of the H-L, to create my “ADR”. I set TP based on the opening price of the next day + 1x, 2x, 3x, and 4x the ADR, and SLs equal to the opening price -ADR. Then ran a macro in search of finding when the trade would terminate based on the TP level being hit or the SL level being hit.
Using solely daily data, the question becomes(assuming long): if the high is greater than the TP, and the low is lower than the SL, is the trade a winner or a loser? It’s impossible to tell without looking deeper into the data. Therefore, I took the MOST conservative approach to this; when a day hit the TP, I labeled it as such. However, if the SL was also hit in that same day, I replaced the winner and it became a loser.
As you can see, I have separated the categories for all long trades, all short trades, and a for fun “random” entry (using excels =rand() function and an if statement if the random number was greater/less than .5). From top to bottom, you see the results of a 1:1 Reward to Risk, 2:1, 3:1, and finally 4:1. As the TP levels become greater, the amount of winning trades decreases. Yet the effective reward or actual amount of pips gained, increases. EU, for the past 10 years, has been increasing in price, therefore you see that the 1:1 ratio is winning (1925 wins, 1665 losses) in the long scenario. This is where the use of the “random” comes into play, verifying Mei Hua’s findings. Compare the Effective reward column with the Losses column. 1:1 and 1:2 have losses greater than the effective reward. However once you get to 3:1, the reward becomes big enough to over come the losses, even though they are increasing.
Remember that in reality the rewards are likely quite a bit higher and they are muddled here due to my conservative back test method. Hope this helps
Are you for real?
Great work, it is sooo nice to see that my work was verified by an outside source. I really had nothing to do with it but the result stands. Its just good solid science at this point. Incredible keep it up.
This quantitative analysis is considering the body of a candle vs the entire candle size (wick to wick). A majority of chart traders, and people using candle stick techniques will attest that there are bullish and bearish candles. In general it is thought that the more of the candle is body is a stronger signal. Even the babypips school is teaching this as part of the curriculum.
Basic Japanese Candlestick Patterns
The optimal is a marubozu, one where there is no wicks/shadows on either end of the candle. So I have created a test to see if this theory is correct.
Data: Using Oanda data from 10-23-2013 - 09-19-2014 on a 1 Hour chart. I have switched to this data set because some people may want to recreate my experiments and may not have the same professional level data feeds I have access to. Oanda is a very common data stream and it makes it much easier to replicate.
Testing Method: I created a ratio I am calling Candle Ratio which is very simple (Close - Open) /(High - Low), this gives you the ratio of the candle body making up of the entire candle. The value is +/- depending if it is a bullish candle or bearish candle. I then executed a trade whenever a candle was >.80 or <-.80, long or short respectively. Then held the trade for periods of 1 - 24 hours exiting at each increment. No commissions or slippage was included.
Candle Ratio Statistics
Here is the descriptive statistics and histogram of the Candle Ratio. You can see that candle sizes are non normal, its a very short and squat histogram instead of a normal bell curve. We see much higher occurrences of “bullish” or “bearish” candles than we would if it was a normal distribution. This non normality is confirmed by the kurtosis value of -1.1 and skew of -.00637. Given that our mean is very close to 0 and median is exactly 0. We have 1635 cases here as our sample size.
Trading off Candle Sizes Statistics
Here is where the rubber meets the road. This chart shows different exiting scenarios, every hour for 24 hours. The mean and median result for your expectancy for using candle bodies is negative. There is an odd amount of + expectation at longer holding periods at 20 + hours. However it is not enough to make the overall result positive. I would not consider this a stable trading relationship even if the longer holding periods were positive, only because the drawdowns of such a trade would be large and highly probable right after you enter the trade. Which would either stop most people out, or freak them out and they would exit themselves. But I think thats highly realistic, most people do not like to endure drawdown in their trade as soon as they put it on. Especially if its highly probably draw down.
Summary:
The results show that bullish or bearish candles that have 80% or more body, do not have the outcomes in general that we are expecting. Especially taking the context that most people want their trade to work right away after seeing such a bullish/bearish candle on the charts. Candle formation sizes are non normal and can not be analysed as such. So take heed and make sure if you are using candles in your trading, that they are actually producing a signal that has + expectancy. Not only that but the signals you are taking are actually related to the follow through you are looking for.