A Real Trading Edge, Quantified: Trading and Stop Management

Beginner Questions
21

Most traders use 2 types of entry systems, those that buy on weakness AKA buying pullbacks or those that buy on strength AKA breakouts. This works in the same way for shorts, wait for a rally to sell or a break down to short. I think that we need to investigate this further. Each methodology can be profitable, if used in the right context and on the right instrument.

Lets lay down the details of this experiment first. I am using the Eur/Usd pair with data going back to 1/1/2001 to present day. I am using 60 minute bars as well. Both the timeframe and the pair I think are pretty much applicable to everyone on this board. A pullback was defined as 3 consecutive down closes, and a breakout was considered as 3 consecutive up closes to get long. the reverse for getting short, 3 up closes or 3 down closes in a row.

I then closed the trade sequentially every hour for 50 hours. Very simple. I was using 1 standard lot as the benchmark for this test. What I am looking for is over a variation of different entry and exit times(remember we shift the exit we can not be in a trade so it moves the entry as well) to have stable histograms. Lets take a look at the results.

Buy On Breakout

buy 3 bar breakout 60.jpg700×210 25.5 KB


Average of Avg Trade: -4.59
Buy On Pullback
buy 3 bar pullback 60.jpg700×214 29.9 KB


Average of Avg Trade: 17.72
Sell On Breakout
sell 3 bar breakout 60.jpg700×210 28.4 KB


Average of Avg Trade: -16.22
Sell On Pullback
sell 3 bar pullback 60.jpg700×210 23.6 KB


Average of Avg Trade: 3.12

The left chart is the % Profitable for this type of entry, the right chart is the expectancy per trade AKA the average profit per trade.

We can see here that universally across the board dealing with pullbacks vs breakouts is quite clear. Buying or shorting on pullbacks is clearly superior to trading the breakouts. To be perfectly honest with you the buying pullbacks expectancy is incredibly stable, i even ran the test again to verify its validity. I was expecting more results like the short a pullback. Where a majority of the expectancies are >0 but not universally so.

Look at the % profitable, the deviations from entry type are not that vastly different. Ranging from 44% to 54% winners. This is a key observation because it shows how when you have borderline entries how a few % in the wrong direction can be the difference between profit and loss.

How many times have you heard people tell you, Don’t Chase The Trade? This is why, on average the expectancy of waiting for price to retrace before entering is better than taking a breakout or after the market has moved a significant distance in the direction you would like to trade in.

This is in no way a trading system, or methodology. I am just putting the facts together here so people can make their own judgement calls and understand the tools in their toolbox. Most of you already have tools in your toolbox that have positive expectancy. But either don’t use them, use them improperly, or don’t even know that they are profitable in the first place and ignore them. So now that I am showing you the simplicity of things that do have positive expectancy over 12 years of data and thousands of data points.

Bottom Line: Buying weakness and selling strength(pullbacks) has a higher on average expectancy than doing the opposite. This gives credence to the Don’t chase the trade axiom.

1 Like
22

Meihua I really like your posts and this thread. Gives me a lot of perspective. Basic rundown of my own system. Determine bias. If bias is long enter when price completes a new swing low with a 150 pip TS. When price moves 100 pips in my favor move SL to BE. Take profit when bias changes or 750 pips. I’ve been wondering if entering on a swing high completes for longs would be better. Thanx for giving credence to my idea I think I’ll practice it.

As for your posts they go noticed by me and give me a lot to chew on. I enjoy reading them. Honestly tho I don’t have much input because I’m not familiar with programming or statistics. So nothing to contribute but I do like trading what you have to say. Just because there aren’t replies doesn’t make your posts undervalued.

23

[QUOTE=“MeiHua;480246”]Most traders use 2 types of entry systems, those that buy on weakness AKA buying pullbacks or those that buy on strength AKA breakouts. This works in the same way for shorts, wait for a rally to sell or a break down to short. I think that we need to investigate this further. Each methodology can be profitable, if used in the right context and on the right instrument.

Lets lay down the details of this experiment first. I am using the Eur/Usd pair with data going back to 1/1/2001 to present day. I am using 60 minute bars as well. Both the timeframe and the pair I think are pretty much applicable to everyone on this board. A pullback was defined as 3 consecutive down closes, and a breakout was considered as 3 consecutive up closes to get long. the reverse for getting short, 3 up closes or 3 down closes in a row.

I then closed the trade sequentially every hour for 50 hours. Very simple. I was using 1 standard lot as the benchmark for this test. What I am looking for is over a variation of different entry and exit times(remember we shift the exit we can not be in a trade so it moves the entry as well) to have stable histograms. Lets take a look at the results.

Buy On Breakout
<img src=“301 Moved Permanently”/>
Average of Avg Trade: 17.72
Buy On Pullback
<img src=“301 Moved Permanently”/>
Average of Avg Trade: -4.59
Sell On Breakout
<img src=“301 Moved Permanently”/>
Average of Avg Trade: -16.22
Sell On Pullback
<img src=“301 Moved Permanently”/>
Average of Avg Trade: 3.12

The left chart is the % Profitable for this type of entry, the right chart is the expectancy per trade AKA the average profit per trade.

We can see here that universally across the board dealing with pullbacks vs breakouts is quite clear. Buying or shorting on pullbacks is clearly superior to trading the breakouts. To be perfectly honest with you the buying pullbacks expectancy is incredibly stable, i even ran the test again to verify its validity. I was expecting more results like the short a pullback. Where a majority of the expectancies are >0 but not universally so.

Look at the % profitable, the deviations from entry type are not that vastly different. Ranging from 44% to 54% winners. This is a key observation because it shows how when you have borderline entries how a few % in the wrong direction can be the difference between profit and loss.

How many times have you heard people tell you, Don’t Chase The Trade? This is why, on average the expectancy of waiting for price to retrace before entering is better than taking a breakout or after the market has moved a significant distance in the direction you would like to trade in.

This is in no way a trading system, or methodology. I am just putting the facts together here so people can make their own judgement calls and understand the tools in their toolbox. Most of you already have tools in your toolbox that have positive expectancy. But either don’t use them, use them improperly, or don’t even know that they are profitable in the first place and ignore them. So now that I am showing you the simplicity of things that do have positive expectancy over 12 years of data and thousands of data points.

Bottom Line: Buying weakness and selling strength(pullbacks) has a higher on average expectancy than doing the opposite. This gives credence to the Don’t chase the trade axiom.[/QUOTE]

Maybe I am missing something or am misinterpreting … But if buying on breakout gives average trade expectancy of +17 and buying pull backs is -5 … And selling on breakout is -16 and selling on pullback is 3… I don’t understand how it’s clear that pullbacks are better.

The sum of the averages for buying and selling breakouts is +1 … While the sum of the averages for pullbacks is -2… The difference is probably not even statistically significant and if it is, it would show that breakouts have a better expectancy by a difference of 3 pips.

24

do i understand you right, that you take any 3 consecutive candles, check if they all close higher than they opened, and then define that as a “breakout” (for a long order) ?

if so then your test is not worth much … for example: three candles, all have a range (high - low) of 400 pips, and all close with +1 pip difference (close - open)… in your setup its a breakout, but i guess we agree that it isnt.

you have to include their “velocity” and their environment into your calculations.

25

You want positive expectancies, thats entire focus of every trader. So why would you ever trade something that has a negative expectancy at all? That just makes no sense. Its clearly better because of the fact that both entries using strength/weakness is >0. as for statistical significance my error is around 1.69% given ~3500 samples per run if you multiply that by 50 for the entire test it will get very very small. So no I do not require 3 pip difference. you can +/- 1.69 % from the averages given.

26

I only used the term pullback, technically it should be buying or selling on strength vs weakness. Most people don’t use that terminology. Again because of my high sample size the individual velocity will be averaged into the equation, because the test was in the market ~80% of the time in general per each run. So the results are like averaging 80% of the entire 12 years which is about 9 years. So yes I am missing some data on the entire set on a per run basis but not on the entire 50 tests but I don’t think anyone is going to claim that a 9 year average range of an hourly bar is insignificant.

I could filter out specific conditions in which that increases or decreases the values. But this is just a broad test to determine if it has any consequence at all. It could be that be that either way is equal and there is no difference whatsoever. Environment and specific conditions are not the point of this test or any other subsequent test. I am not here to give you a system or tell you the optimal ways to implement anything. I am just telling you in wide brush strokes that very simple concepts have positive expectancy. It is that simple. If you want further information on specifics you can use this as a jumping off point and share your findings. I would love to see them.

27

i really like the statistics your sharing here meihua, but i’m not really understanding how to use this to enter trades?? some help pls, thanx

28

Basically the trade is done using the Risk - Reward ratio which means that you need to see if the risk you are taking is worth the reward from such a trade and this has to be taken into account for all such trades you need to do :slight_smile:

29

Hi MeiHua, this is damn useful information. Please keep continuing if you have the time.

How do you perform your analysis? Is it all automated or do you do it by hand? Or a combination of both?

30

[QUOTE=“MeiHua;480305”]

You want positive expectancies, thats entire focus of every trader. So why would you ever trade something that has a negative expectancy at all? That just makes no sense. Its clearly better because of the fact that both entries using strength/weakness is >0. as for statistical significance my error is around 1.69% given ~3500 samples per run if you multiply that by 50 for the entire test it will get very very small. So no I do not require 3 pip difference. you can +/- 1.69 % from the averages given.[/QUOTE]

So you wouldn’t buy on pullbacks then since the expectancy is -5 pips?

31

Good catch that is an error in my report. I couldn’t see what image i am posting under because it just shows the code in the post. I am editing it immediately, thanks.

32

It its totally automated, either in my software platform that i code myself, or in the statistical package R. I don’t believe in manual backtesting as theres to much fuzziness so to speak. All of these experiments and tests can be replicated following the procedures I outlined in each experiment. If I did it by hand it would be unreplicatable. I am trying to take the scientific paper route so to speak.

33

It is always better to do automatic back testing as that gives us the required results.

But in the end it all comes down to how much Risk - reward you are having :slight_smile:

34

Actually no that’s a false statement. The only thing that matters is expectancy. You can have low risk reward and high win rates, or low win rates and high risk reward. Both methodologies create winning systems.

35

okay meihua, i think you miss my question on last page but hey its okay… maybe it was dumb question i don’t know lol

but its good because i thought of somthing else. was looking at your chart, and small stop means more noise and good chance of getting stop out (and loss of course). so what i learned, and maybe would help you is if you just don’t use the stop loss. it help me sooooo much more and hardly ever lose now. probably better than using big stop anyway, like 100 pips, because then you can just close the trade when it is losing and probably won’t be a winner anyway.

see what i am saying??? or no, sorry if im not explaining my idea right :stuck_out_tongue:

36

I am assuming you mean that this is the post your referring to. I am not here to give you entries and exits. I thought I was pretty clear on that. I am just revealing tools that most people already have in their toolbox that have positive expectancy. You decide how you want to do with this knowledge.

Regarding the second post. not using a stop loss is just ludicrous in my opinion. I would always use a stop loss period. If my trade is down 100 pips its just going to be a loser anyway? making generalizations like that is totally incorrect. Stops and Targets are totally system dependent. If i have a weekly swing system if a single trade is down over 300 pips its not going to be a big deal. If I have a 1 minute scalping system 30 pips probably way too much. Making these assumptions in trading is going to cost you money. Each system and market have different characteristics. Also closing trades early because you think its going to be a loser kills consistency without understanding your MAE.

37

the base of this research…I have used daily data of the EUR/USD pair from 1/1/1999… using 1 minute bars

I never found any reliable data. Where did you get yours? To my knowledge, it’s impossible to get the minutes bars for the last year. Unless you collect it yourself. And from which broker?

Are-you making money?

38

ok first let me say sorry, i know you point isn’t give traders when to buy/sell stuff like that, but even though all this charts and statistic look really helpful i’m just not really sure how to use it. i see you write “you decide” but since i just don’t know, it is hard!!

for my system i like to use 1hour and 15minute chart, so i just know if it going say 100-125 pips wrong, it is just a loser

39

I have a professional data feed and access to historical data banks from exchanges around the world. I can select brokers or get aggregate. This study here is using FXCM historical prices.

I am a full time trader, enough said.

40

If you read the post more carefully you would have discovered I used the average range of the last 14 bars to determine the baseline. So you can do that same study on the 1 hour bar and come up with your own answer for your own time frame. This way its adaptive to current market conditions and you did your own homework.