Can’t say I’m very surprised by the results. I’ve done plenty of stats similar in nature to this, although never this exact one.
What’s interesting about general axioms such as this one (less wick=more power), candle patterns in general, the trend is your friend, R:R, pivots, fibs etc. is that some of them do seem to have some evidence supporting them, while others do not at all. It makes me wonder how conventional wisdom has provided both good ideas and bad. I suspect this one may have some more merit than is shown currently, but context is the obvious filter that will make or break it and context can be very very difficult to quantify (at all/correctly).
I think that is really the problem with trading education, and more specifically the retail education. Also a lot of these things like candlesticks, trend analysis(think dow theory), fibonacci were all created before we had the ability to even quantify the markets to test them at all. So they may have worked for the creator, or it could have really been luck or the markets were drastically different then vs now. We really will never know but what really matters at the core is. Are you using tools that you actually understand, and that provide an advantage in the marketplace?
Speaking to the context argument used by a lot of TA educators is that it ends up being more of a defensive stance. The reason I say that is that context in terms of the market is that every moment in the market is unique, and that since markets are non normal you can not frame events just using “common sense”. Humans are prone to thinking in normal distributions just because thats how a majority of the world works. So you will actually never have repeating context in a market, so looking for another situation to match said context is impossible. Given that, you could even quantify exactly the context required, which again is very very impractical.
First, great work over the years and it’s all very interesting. But, other than boredom, what’s the point of looking at historical data and making assumptions from the results?
Price action and the context of the market now matter more and your system can never be impossibly rigid, other than your own money management. So why look in the past and make assumptions? Just bored?
This is directed to mostly everything after the first post, larger RR’s being more successful are fairly obvious. Many traders don’t touch anything that’s < 4:1.
It seems like you have 2 main questions here, am I doing this out of boredom and why should I make assumptions from historical data. I can address these individually.
First off the boredom, I am posting here in BP and about this subject because I do not think there is anyone else who is producing work like this. I think its a resource to the community and can maybe help people. No one else has a scientific method that has reproducible results when conducted by others. This is also how I think about and analyze the market, and i find markets fascinating. Sure, there is a bit of boredom in it because I am posting on an internet forum. But then again so are you.
Looking at your question “what’s the point of looking at historical data and making assumptions from the results?” specifically. I want to address the word assumptions here. Due to the scientific method I am using and replicable results, I am not making any assumptions. Along with that if looking at historical data and analyzing it were all assumptive. That would also negate all quantitative finance, since they use historical data for the entire field of study. So lets get to the point, as you put it. The point is to identify specific phenomenon that occurs in the market place. These phenomenon can then be used to identify edges in the market in which to ultimately profit.
I wrote in your thread already several months ago to say thank you; I kept coming back to see if there is more and lucky me…
I have used your information in several ways:
[U]I take trades on pullbacks[/U] (or at important levels) - I don’t chase trades like I used to. This means that I am now less likely to be stopped out on a retrace.
[U]I have increased my stop-loss[/U]. When I started, I used small stops and watched my money get eaten up by the noise, even though in the end my direction choice turned out to be right. My stops are now sized to give my trades a little room to breathe.
[U]I try to hold onto winning trades[/U] when they are going in the same direction as the larger trends - I noticed on your RR part of the thread that the better performing method was the one that went with the longer-term trend. This is probably the hardest thing to do but with your data, I have more confidence to do so.
I’m still not there, but I am trying hard. This quant business is way over my head but I can see the benefits and I’m able to make limited use of it. Without it I would be prodding a stick in the dark. Slowly but surely my account is growing. Note: it is still slowly…
I am glad that you found the information contained here helpful. Honestly the slow growing account phase of trader development is a great place to be. First of all you stopped the bleeding of your account, which everyone goes through when they start trading. Next off because you now have positive expectancy, you can continue to learn and trade while profiting. Consider that when you start out you pay the market for every trade you take because you have negative expectancy, yes this is the paying your dues part. But you must do so in order to learn and gain experience as a trader. Now as a trader with positive expectancy, you can continue to do the same thing. Learn and grow from trading, but are now getting paid to do the exact same thing. So enjoy it, embrace it, and most of all keep learning and growing. That’s the only way to make that slowly turn into … what ever level of performance you want it to be.
I am hugely impressed by the work you have done and your generosity in sharing. I have read through this thread once and will now go back to your analysis and study more carefully.
I am a relative newbie to Forex trading and am finding that high quality information is very difficult to find. There are many who have opinions and “systems” which often seem to be just recycled consensus opinions with no real evidence to support them. In my other career in health I am often dismayed at the lack of Evidence Based Practice. It seems to me that one can never proceed and succeed without tested high quality evidence. The evidence you have given is clear, well defined and presented in a commendably modest manner.
Its been a while but there was something nagging at me that I felt I should post. Bollinger Bands, they are a standard in almost every traders toolbox, and they are included in almost every trading platform and charting toolset there is. From MT4 to Bloomberg terminal. But we also know market prices and returns are usually non normal. So why should bollinger bands which assume a normal distribution work at all for trading in the first place? Well thats what I am here to find out.
I conducted a study taking 1 Hour bars from Oanda starting 1-1-2005 until 12-26-2014. I plotted a Bollinger Band with a 20 period look back and +/- 1, 2, 3 standard deviation bands. The moving average itself was using a simple ma. Fairly simple and straight forward test. But extremely interesting results.
Here I am using the Closes above or below any standard deviation band and just calculating the percentage of closes that occurred in that area over the entire data set. I also have the expected value if the distribution of the market was based on a normal CDF bell curve.
Lets look first at the +/- 1 std dev Band results. With an expected value of 31% but an actual value of 27% would mean that the distribution of prices is highly leptokurtic. In other words its very tall and thin at the middle. Why is it thin? Well in relation to the expected benchmark less of prices are outside the 1st Dev Band then should be.
However the MA Above or Below columns show that the distribution of closes with respect to the 20 period MA is very symmetrical. I keep having results show this in the EUR/USD. Its honestly very amazing to have multiple confirmations of a phenomenon, even when it is measured several different ways.
The +/- 2 std dev band is where things start to get interesting. The fat tails of financial markets are well known and its around this area that for the EUR/USD you can actually start to see it. Our result of 6.63% is 45.7% larger than the expected value. This starts to distort the measuring power of the Bollinger Band.
With the +/- std dev band the fat tails of the market become increasingly obvious. Our result of .89% is 329% larger than the expected value. This is where the breakdown of any system using the normal CDF or probability distribution break apart. The deviations from the expected value become enormous.
With all 3 of the standard deviation bands we have very symmetrical results which does mean that this can be evaluated equally well on the long side as the short side.
My Take Away: The EUR/USD is a symmetric fat tailed market, which we already knew. But do Bollinger Bands have a place in traders toolboxes? I would say yes, because in absolute terms the difference from the expected values and the values in reality are not that far apart. Also given the information here you can adjust to take it into a count. It does show extremes, just not as extreme as you think.
Being off by a few percentage points off for an ordinary discretionary trader is not going to be that critical. If your running an algorithm or highly quantitative system then absolutely this becomes a factor, but most people are not doing that. So as long as you evaluate Bollinger Bands they can still have a place on your chart. As always understanding the limitations of your tools and knowing your market are the corner stones to trading success.
Newbies read posts #1 and #2. More than likely you will not understand them completely. So go and learn about those parts you do not understand. Ask questions about these specifics and details and not simply ask for a golden ticket to be handed to you.
Why I love it, is that this is exactly what I did at one point in my trading career. I traded the identical trades (as far as possible) on ratios of 1:1, 1:2, 1:3, 1:4, 1:5, and 1:9 to see which have optimum results. Then I altered the formula used to obtain the SL position and repeated. All the results where then compared an expectancy curve to find optimization.
This took months of work since it was all live trading not backtesting and a minimum of 100 round trips was required on each test to get a statistically significant result.
I have two trading strategies and this forms part of the core reasoning for one of them (The other is pure fundamentals with stops in the area of 300 to 500 pips, or 3,000 pips in the unique case of the USDNOK).
If you are looking for a “Holy Grail” statistical analysis will be a part of it.
I think you may be missing the main point of what these results show/imply, and other “rules” of the market that exist. Markets have two main states; moving (trending), and not-moving (flat). When markets are flat, the trader loses their easiest edge in the market: trends. When you look at the markets, they don’t move like 50 up, 60 down, 70 up, 80 down, 90 up, 100, etc. Instead, they move more like 100 up, 50 down, 120 up, 60 down, 120 up, etc. (keep in mind there are micro moves within these moves, call this factor x). In other words, one leg of the move or “wave” is disproportional compared to the previous leg (generally speaking). Thus, the trend. By using large SL and even larger TP points (both greater than factor x), we make the factor x movements disappear, and we only trade either the small leg or the big leg.
With this in mind, pretend the trend is up and we choose to short. When this occurs, the down leg will not be enough to trigger our TP, and the up leg will take out our SL. We lose. If we choose instead to long, the down leg will still be not enough to trigger our SL, and then our TP will be hit. We will lose more [I]often[/I] than we win, but we will win [I]more[/I] and come out on top in the end.
It’s a bit of a long answer, but I hope this clears things up a bit, and that if you look at your charts, you’ll see how this is true.
I think the main conclusion of your experiment might be:
High RR ratio (3:1), with stop greater than ADR, will be profitable.
This is only valid IF the stop is greater than ADR.
Some people asked (and I also wondered) whether the same principle is applicable in lower time frame (say 1H chart).
I think the answer is NO. Your experiment just proved it:
50 stop and 150 target was not profitable. And 50 is more than the ATR in 1H chart (currently it is less than 40).
So, in 1H time frame, we have 3:1 RR and stop outside the ATR (50 stop, 150 target), but it was not profitable.
The same holds true for 15M and 5M time frame, because their ATR are far less than 50.
This might suggest: In lower time frame, high RR won’t work, even if the stop is outside the noise (ATR).
Now, I think when you use ADR as the noise, maybe it is not about the noise in daily chart (in fact, the noise for daily should be ATR in weekly).
ADR is actually the noise for all of the time frames inside the day. So, whatever the time frame we trade, we should place the stop outside the ADR, which is the noise for one day.
So the final conclusion is: 3:1 RR is profitable ONLY IF the stop is outside the ADR.
Now, with stop greater than ADR, and target more than 3 times ADR, this is swing trading.
This might suggest: a swing trading or trend follower system, using high RR, and stop outside the ADR, is more likely to be profitable.
I’m still here - just in case. Just wanted to let people know that the info in this thread has proven very useful to me over the years. Even today, I pulled it up as a reference…
