Simple Mean Reversion | Page 23 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

A Simple Mean Reversion Strategy 392 replies

Pairs Trading: Reversion to the Mean 22 replies

Forward test of my new mean reversion strategy 11 replies

Synthetic hedges, cointegration, mean reversion and similar stuff 419 replies

Cointegration, Synthetic hedges, mean reversion in R, Tech Thread 0 replies

Attachments: Simple Mean Reversion
Exit Attachments

Simple Mean Reversion

  • #441
  • Dec 11, 2018 12:14pm Dec 11, 2018 12:14pm
Quoting alphaomega
Now, here comes the real question! Where the total length of price movement is longer? During range, or during trend? And why this even matters? What happens when you know precisely the total length of price movement for all time frames and all markets all the time?
During range, it matters because mean reversion profit is ideally made during range.

What happens? I'm not sure...
Total length will increase if you go into smaller timeframes, but if you calculate this for all markets and all timeframes what happens?
Your CPU will be busy, but that's not the correct answer i guess.. I'm in the blind here, I don't know what happens.
Persistence will get it. Consistency will keep it.
  • #442
  • Dec 11, 2018 12:17pm Dec 11, 2018 12:17pm
Quoting alphaomega
If the market moves from point A to point B in a straight line over time T, then what is the actual length of the movement from A to B ? If the market moves up from point A to point B and down from B to point C, then what is the actual length of the movement from A to C ? If the market moves down from point A to point B and up from B to C and down from C to D, then what is the actual length of the movement from A to D? If the market moves........ Yes, that's right. The more it moves, the longer the length becomes. Or in other words, the frequency...
Are you refering to the scaling laws from R. Olsen ?
trying to make everyday the best risk/reward decisions possible
  • #443
  • Dec 11, 2018 1:24pm Dec 11, 2018 1:24pm
Quoting alphaomega
Now, here comes the real question! Where the total length of price movement is longer? During range, or during trend?
I was puzzled by the question so I took my ticks database for EUR/USD. I measured the path length (sum of absolute tick returns) over all periods of 24 hours starting at every hour. It isn't really a range vs trend classification but the bigger the return the more I expect the price has trended during the period... I cannot see anything. Yet I'm amazed that the average length of the price path is 18980 pips per day for an average return of only 58 pips.

This is for EUR/USD since 2003
Attached Image (click to enlarge)
Click to Enlarge Name: return_given_coastline.png Size: 12 KB


(I did this quickly after work so there may be an error somewhere)
Attached File(s)
File Type: csv length.csv   3.3 MB | 551 downloads
No greed. No fear. Just maths.
  • #444
  • Dec 11, 2018 2:15pm Dec 11, 2018 2:15pm
Quoting PipMeUp
{quote} I was puzzled by the question so I took my ticks database for EUR/USD. I measured the path length (sum of absolute tick returns) over all periods of 24 hours starting at every hour. It isn't really a range vs trend classification but the bigger the return the more I expect the price has trended during the period... I cannot see anything. Yet I'm amazed that the average length of the price path is 18980 pips per day for an average return of only 58 pips. This is for EUR/USD since 2003 {image} (I did this quickly after work so there may be...

how many of 18980 pips were up ticks (in distance) versus down ticks on a one hour engulf candle (if any)? how about a day or week or month or a combination of H4+H1 or H1+M30+M5 or M30+M15+M1 or any such combinations? "Mandlebrot"
Staying in my lane...
  • #445
  • Edited 2:46pm Dec 11, 2018 2:28pm | Edited 2:46pm
Quoting Chake
{quote} During range, it matters because mean reversion profit is ideally made during range. What happens? I'm not sure... Total length will increase if you go into smaller timeframes, but if you calculate this for all markets and all timeframes what happens? Your CPU will be busy, but that's not the correct answer i guess.. I'm in the blind here, I don't know what happens.
Nothing happens!

But it will take you awhile (or forever) to understand what I mean by nothing. And why this is important. I strongly suspect very few people know about this.

For example, I started trading a long time ago and I found the most important things only in the last few months. And that's not even funny.
Now I know that I wasted many years walking in the wrong direction. Most people don't even know they are walking in the wrong direction.

You will have to go into the deep, back in history before the year 1970 to understand where and how the current financial system came to be.

The architects of the international currency markets created very fine machine! Almost perfect. Their secret "multidimensional triangular equation" was supposed to keep the world under control. And it does. To some extend. However, they made one mistake! There is a flaw in their machine. Anomaly. If you find this anomaly......

You see, it is very simple. The market is like a game. (Do you like to play games? Mind games? Logical games? Math games?)
You can play with the market by the rules (Like 99.9% of the people) and in this case you will lose every time! Just like you will lose every time if you play in the casino.

OR you can initiate a search for the back door. OR you can simply look for the weakness (the anomaly) in the system!

All you need to know is that Everything adds up to 1. Remember that! The market does not move as much as it appears! The sum of all prices is pretty much stationary and oscillates around some fixed values.

All these currencies...USD, EUR, GBP, JPY, AUD, CAD, NZD, CHF........they are simply cogs in the machine! You can't beat the individual cogs in the long run. Or it's extremely hard and mostly depends on luck.

But, if you are smart, you will target the whole machine! And in this case you have a real chance to beat the game.

And the market makers (the LP's) will hate you for this! Just like the owners of the casino will hate you when you hit them with 100K 5 times in a row.
You beat their little game every time. And in the case of the market you beat their large game every time.

I know that many people will not understand a single word of what I'm saying here. But that's OK.
  • #446
  • Dec 11, 2018 2:52pm Dec 11, 2018 2:52pm
Quoting alphaomega
{quote} Nothing happens! But it will take you awhile (or forever) to understand what I mean by nothing. And why this is important. I strongly suspect very few people know about this. For example, I started trading a long time ago and I found the most important things only in the last few months. And that's not even funny. Now I know that I wasted many years walking in the wrong direction. Most people don't even know they are walking in the wrong direction. You will have to go into the deep, back in history before the year 1970 to understand...
Good Job bro... something I have been screaming all long since I wore my "facts and data" hat. Many will NEVER understand the entire inner workings of that fantastic mathematical balance. It is REQUIRED to keep the faith in our global financial system. You are getting there.... on the right path but not possible in retail market structure with the tools we have available to us.... and mind you, in regulated markets, whatever you discover (I know where you are going with this) is not allowed under the context of "keeping markets fair and balanced" or "preventing market (making) disruptive practices". I know first hand on my first day trading futures. You can NEVER win going against market making... you have to piggy back them and use their support and leverage. "Do what Markets(makers) HAS to do to survive(make price)".... i have been telling this all along and there is a reason why I was yelling and screaming "Show me the evidence of your success" on every thread from June thru August. Long term profitability is VERY HARD in the currency market trading price based strategies. Only two ways it is possible for intradays... 1) microscalpings 2) portfolio diversifications via multi-strategies and I will stand by my research till the end of time.
It's a game designed to trick you into thinking "you are just around the corner to make it big". That day will never come for the 99.99% who are still trying this for over a decade.
Staying in my lane...
  • #447
  • Dec 11, 2018 3:39pm Dec 11, 2018 3:39pm
Quoting PipMeUp
{quote} I was puzzled by the question so I took my ticks database for EUR/USD. I measured the path length (sum of absolute tick returns) over all periods of 24 hours starting at every hour. It isn't really a range vs trend classification but the bigger the return the more I expect the price has trended during the period... I cannot see anything. Yet I'm amazed that the average length of the price path is 18980 pips per day for an average return of only 58 pips. This is for EUR/USD since 2003 {image} (I did this quickly after work so there may be...
Now that you found part of the "secret", what is your next step?

Come on, PipMeUp you are one of the smartest people in this forum. If you can't find the solution to the problem here, then what is left for the rest of the forum?

Here is what I'm thinking.
If there is so much "excess" movement in comparison to the actual open to close change, isn't this the most obvious market inefficiency?
Add the fact that the 28 pairs balance out most of the time, during the first half of the day.
And we have all the stars aligned right in front of us.

Directions, trends, news...all is irrelevant.

All we have to do is to make the correct algorithm (with the correct math) to capture small percentage of this total excess length.

All we need is something like 0.1% of the total length of all pairs.

This will result over 500 pips per day net profit.

Can you see the light?
  • #448
  • Dec 11, 2018 3:51pm Dec 11, 2018 3:51pm
Quoting alphaomega
{quote} Now that you found part of the "secret", what is your next step? Come on, PipMeUp you are one of the smartest people in this forum. If you can't find the solution to the problem here, then what is left for the rest of the forum? Here is what I'm thinking. If there is so much "excess" movement in comparison to the actual open to close change, isn't this the most obvious market inefficiency? Add the fact that the 28 pairs balance out most of the time, during the first half of the day. And we have all the stars aligned right in front of us....
I am generally weak in math (I can barely understand your SQRT(Time) function) so went the route of applying brute force. If I may add a hint. I came to the conclusion that there are only two possibilities (across 28 pairs) given unlimited capital (just for modeling purposes).
1) you either take frequent small wins during your accounting period (mine of always 4 days of week) and close the remaining in loss
2) take frequent small losses and end the week in net profit... pick your poison.

The magic (balance or equity explosion) happens when you know which market conditions is ripe for which type of exits to incorporate for which pairs to match against which sessions and using volatility (news/session open or close) towards lift your spirits :-
Staying in my lane...
  • #449
  • Edited 4:49pm Dec 11, 2018 4:39pm | Edited 4:49pm
Quoting VEEFX
{quote} I am generally weak in math (I can barely understand your SQRT(Time) function) so went the route of applying brute force. If I may add a hint. I came to the conclusion that there are only two possibilities (across 28 pairs) given unlimited capital (just for modeling purposes). 1) you either take frequent small wins during your accounting period (mine of always 4 days of week) and close the remaining in loss 2) take frequent small losses and end the week in net profit... pick your poison. The magic (balance or equity explosion) happens when...
For now, all I know is that my 28 pair HFT system works surprisingly well most of the time. (It's a real beast and amazing to watch in real time) Unfortunately not all the time.
Occasionally the balance breaks and I have to give back almost all of the profits in order to prevent larger losses. I still make a little net profit, but this is not even close to the real potential of the system.

Now here is the real deal, If I find effective solution (the correct math) for this balance problem, then the market makers (And the retail brokers) are going to be in serious trouble.
It's possible for this to escalate very quickly even starting with small capital!

The only way for them to survive my algo it to hedge every single trade. Otherwise am going to drain them to the last penny. They will owe me money!
But we all know they can't hedge my trades. And so, I'm going to kill them with 1 million cuts! They will have to ban me from trading if they want to survive.

But for now they are safe. For now.
  • #450
  • Edited 6:02pm Dec 11, 2018 5:13pm | Edited 6:02pm
Quoting alphaomega
{quote} For now, all I know is that my 28 pair HFT system works surprisingly well most of the time. (It's a real beast and amazing to watch in real time) Unfortunately not all the time. Occasionally the balance breaks and I have to give back almost all of the profits in order to prevent larger losses. I still make a little net profit, but this is not even close to the real potential of the system. Now here is the real deal, If I find effective solution (the correct math) for this balance problem, then the market makers (And the retail brokers) are...
Your experience is VERY similar to my algo. I posted by results in the diversifief trend trading thread for april/may. It might still be there I think.

If you are doing this in mt4, they won't ban you. They will simply issue to "Not Connected' comment and fake disconnect your EA. I really hope and pray you succeed as you are only validating my hard core research...I love it !!!

EDIT: Not to discourage you... that is certainly not my intention but the biggest enemy you will encounter in creating such a balanced equation (in real live environment) is the spread widening. It destroys any sort of mathematical equation which simply cannot be modeled in demo unfortunately. I could not try this but please see if entry and exiting on midpoint of bid/ask helps with the balance. May be always use midpoint in the calculations and/or limit your algo execution to only just few hours during one active and liquid market and use session start as your starting time instead of daily open. And, finally, do not forget to check for spreads during exits. Not many know or track spread widening at exits. lol

If you really see all the "trading conditions" that span across brokers, you will see a hidden message in those terms. example: stop limit distance. That's the biggest edge they have to prevent retailers from trapping price to the micropip level by disallowing them NOT to enter too close to the current price! Second spread widening but with a twist either by keeping bid static and spread the ask or keep ask and spread the bid... depending on whether you are entering or exiting the position. Third, cutting the margin in half for hedged trades. Yes, that is a huge apparatus in their arsenal. Forth / StopOut levels, this is ALSO huge, the more they allow you to drain the account, it becomes almost impossible to recover making you lose MORE than your USED MARGIN. Biggest scam ever! We must only lose the required margin per trade and nothing more! Fifth, Charging swap by the second. Shame on you Oanda! you are draining accounts in microcents at absolutely no cost of credit for intra-day positions. Sixth, altering margin rules based on increased volatility... quite damaging for position traders and finally, Seventh/ Messing with improper FIFO implementation.

Edit2: Forgot to add two more.... Eighth / NOT checking available/free margin for pending orders. Recipe for quick margin call if you are not careful. 9) not allowing change of position size of pending orders.... another huge edge they want to prevent!

All of the above only applies to MT4 platform brokers.
Staying in my lane...
  • #451
  • Edited 6:19pm Dec 11, 2018 6:05pm | Edited 6:19pm
Quoting VEEFX
{quote} Your experience is VERY similar to my algo. I posted by results in the diversifief trend trading thread for april/may. It might still be there I think. If you are doing this in mt4, they won't ban you. They will simply issue to "Not Connected' comment and fake disconnect your EA. I really hope and pray you succeed as you are only validating my hard core research...I love it !!!
This is what I think:

The broker will start to initiate "disconnections" as soon as my account becomes large enough to cause problems. The other thing is that most brokers have some limit regarding the number of open trades. Usually from 200 to 500 max per account. (These limits cripple my algo)

If I put 100K in one account and my algo starts hitting them with a few hundred trades per hour they will notice this quickly. Especially considering the fact that sometimes I make up to 20% per day.

So here is the actual plan for implementation:

I select the 20-30 brokers with the best trading conditions and spreads.
Deposit small amount into each broker. Ideally below 5K. (This is the average retail deposit).
All terminals are running on one workstation and I use one of the terminals for a signal. This is where I Install my system with the interface. From the main terminal I send signals to all other terminals via trade copier. Hopefully this way the operation will survive a lot longer and I will be able to withdraw profits before they even know what is happening. It's not ideal long term solution, but I think this will work for some time. Probably long enough to pull a few mil.

Attached Image (click to enlarge)
Click to Enlarge Name: HFT Network Plan.png Size: 57 KB
  • #452
  • Dec 11, 2018 6:25pm Dec 11, 2018 6:25pm
Quoting alphaomega
{quote} This is what I think: The broker will start to initiate "disconnections" as soon as my account becomes large enough to cause problems. The other thing is that most brokers have some limit regarding the number of open trades. Usually from 200 to 500 max per account. (These limits cripple my algo) If I put 100K in one account and my algo starts hitting them with a few hundred trades per hour they will notice this quickly. Especially considering the fact that sometimes I make up to 20% per day. So here is the actual plan for implementation:...
They performed disconnects on my $1000 400:1 leverage when it reached a rapid 1200%. My experience with brokers is limited to being in the USA so this model could work if you are based elsewhere. I like the trade copier approach but know that internal memory leaks will start to creep when you reach to 200+ position count and could not run more than 4-6 terminals in parallel on the latest I7 7th gen CPU.
Keep at it man. I really hope you succeed. Dont forget to put a live TE for us wannabe millionaires to gaze on :-)
Staying in my lane...
  • #453
  • Dec 11, 2018 6:45pm Dec 11, 2018 6:45pm
Quoting VEEFX
{quote} They performed disconnects on my $1000 400:1 leverage when it reached a rapid 1200%. My experience with brokers is limited to being in the USA so this model could work if you are based elsewhere. I like the trade copier approach but know that internal memory leaks will start to creep when you reach to 200+ position count and could not run more than 4-6 terminals in parallel on the latest I7 7th gen CPU. Keep at it man. I really hope you succeed. Dont forget to put a live TE for us wannabe millionaires to gaze on :-)
Will see how it goes.

I'm currently starting to experiment with this setup with demo accounts just to see if I can run the whole operation from 1 PC.
So far 20 terminals no problem. CPU below 30% and memory a few GB max.

The key is to make the terminals light! No charts, no indicators, minimum bars in history. Only 1 chart per terminal with lightweight EA. And you keep the platform minimized!
On the main terminal I have my interface where the signals are coming from and this alone consumes around 5% CPU.
Attached Image (click to enlarge)
Click to Enlarge Name: EURCHFH1.png Size: 105 KB
  • #454
  • Dec 11, 2018 6:50pm Dec 11, 2018 6:50pm
Forgot to ask... does your algo try to optimize trade frequency? I think trade frequency will matter hugely in what you are trying to achieve and not just std deviations of price. It would be interested to know if you do (and the framework around it). Time Frequency is essential for scalping volume-centric trades in order to optimize use of available margin.

Also, when your algo does not work (as you said), look into the pairs with disproportion margin requirement and $/pip. I feel this infinite design (the global currency market game as you rightfully put it) has some protections built in between $/pip vs used margin vs pip range over speed to thaw down the equity explosion process (either upward or downward)... extremely complex shit that drove me insane for over 2+ yrs.

Quoting alphaomega
{quote} Will see how it goes. I'm currently starting to experiment with this setup with demo accounts just to see if I can run the whole operation from 1 PC. So far 20 terminals no problem. CPU below 30% and memory a few GB max. The key is to make the terminals light! No charts, no indicators, minimum bars in history. Only 1 chart per terminal with lightweight EA. And you keep the platform minimized! On the main terminal I have my interface where the signals are coming from and this alone consumes around 5% CPU. {image}
Yep. tried all that. CPU will churn up with volatility so something to keep in mind. IS your EA one multipair, multiTF EA running on each tick? mine required that I run on every maintenance on every tick and entries/exits only on open close of M1 candle (M1 being the universal time element). The entire entry signal generation process would take over 20+ seconds across 28 pairs.... perhaps yours is more lightweight and optimized codeset. Mine is just copy/paste/make it work type programming skills :-
Staying in my lane...
  • #455
  • Dec 11, 2018 7:01pm Dec 11, 2018 7:01pm
Also investigate CHF when things don't make your algo work smoothly. I think CHF plays a unique role in striking equilibrium (flight to risk situations). I could not figure this one out. There is something about it in the design that makes CHF unique. I have found even strength indicators to be unreliable for CHF pairs (after the peg was lifted)
Staying in my lane...
  • #456
  • Dec 11, 2018 7:02pm Dec 11, 2018 7:02pm
Quoting VEEFX
Forgot to ask... does your algo try to optimize trade frequency? I think trade frequency will matter hugely in what you are trying to achieve and not just std deviations of price. It would be interested to know if you do (and the framework around it). Time Frequency is essential for scalping volume-centric trades in order to optimize use of available margin. Also, when your algo does not work (as you said), look into the pairs with disproportion margin requirement and $/pip. I feel this infinite design (the global currency market game as you rightfully...
Trade frequency is proportional to the volatility. (Most of the time) But I also try to optimize the system in such way that it never reaches the limit for max open trades. My standard limit is no more than 15 open trades per pair at any time. And based on this rule I calculate my margin and trade sizes so that even if I have max open trades in all pairs I still have another 50% available margin. In other words I never use more than 50% of the margin.

My system is not tick based. It's based on multiple millisecond timers! This way the CPU load is almost constant and the periods of high volatility does not affect the performance of the system too much.
  • #457
  • Dec 11, 2018 7:59pm Dec 11, 2018 7:59pm
Quoting alphaomega
{quote} Trade frequency is proportional to the volatility. (Most of the time) But I also try to optimize the system in such way that it never reaches the limit for max open trades. My standard limit is no more than 15 open trades per pair at any time. And based on this rule I calculate my margin and trade sizes so that even if I have max open trades in all pairs I still have another 50% available margin. In other words I never use more than 50% of the margin. My system is not tick based. It's based on multiple millisecond timers! This way the CPU...
Got it. Because of reversion to mean, a boundary is required for you to cap. Mine was mostly divergence from mean so it was far most aggressive in using available margin. If I had a free magin be one available even for one microlot, it would take the signal and trigger partial close when margin gets depleted squeezing every little bit of buying power made available to me I would also go in idle mode during low liquid times by closing trades that approach Breakeven ( not those in profit or loss) to free up margin and accommodate for the next volatile period (and there is one always happening around d the corner...i.e. do what market have to do to survive. Markets won't exist if there is no volatility :-)

Anyways, good discussion bro. Keep us posted of your efforts. I will most likely resit this model when I retire to keep my mind occupied from collecting dust..lol
Staying in my lane...
  • #458
  • Dec 11, 2018 9:44pm Dec 11, 2018 9:44pm
Quoting alphaomega
Now, here comes the real question! Where the total length of price movement is longer? During range, or during trend? And why this even matters? What happens when you know precisely the total length of price movement for all time frames and all markets all the time?
I would say it depends. It depends on how far price moves from the mean in a trend and for ranging it depends on how many times it crosses back and forth across the mean.
The trend is what kills me though with this method of reversion to the mean. In a trend it can be a long, long time before it gets back to where the first trades were made. Maybe years.
  • #459
  • Dec 11, 2018 10:49pm Dec 11, 2018 10:49pm
Quoting BlackStack
{quote} I would say it depends. It depends on how far price moves from the mean in a trend and for ranging it depends on how many times it crosses back and forth across the mean. The trend is what kills me though with this method of reversion to the mean. In a trend it can be a long, long time before it gets back to where the first trades were made. Maybe years.
Recycle trades daily. Close all trades and pending @ 23:30 or 00:00
  • #460
  • Dec 11, 2018 11:54pm Dec 11, 2018 11:54pm
I understand that this can help! They have weekly settings!
Attached Image (click to enlarge)
Click to Enlarge Name: Screenshot10.png Size: 87 KB
Attached File(s)
File Type: pdf Trading_With_Market_Statistics.pdf   557 KB | 1,074 downloads
File Type: ex4 VWAP.ex4   21 KB | 542 downloads
0 traders viewing now