Simple Mean Reversion | Page 20 | 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

  • #381
  • Sep 28, 2018 4:28am Sep 28, 2018 4:28am
Quoting alphaomega
{quote} The indexes are just one factor. If the whole index is outside 2 standard deviations - take actions to reduce risk (close positions if you have to). But there are many other factors that you have to track if you want to detect the danger early! News. Obviously. Turn off the system before major news. This way you avoid almost half of the fat tails. Sometimes manual intervention is necessary. Pay attention to the spreads! Track the averages! Sudden spread spikes across the board for no reason = danger signal! Watch out for smooth trends. If...
Thx AO, nice tips!
Persistence will get it. Consistency will keep it.
  • #382
  • Sep 28, 2018 4:46am Sep 28, 2018 4:46am
Quoting alphaomega
{quote} The indexes are just one factor. If the whole index is outside 2 standard deviations - take actions to reduce risk (close positions if you have to). But there are many other factors that you have to track if you want to detect the danger early! News. Obviously. Turn off the system before major news. This way you avoid almost half of the fat tails. Sometimes manual intervention is necessary. Pay attention to the spreads! Track the averages! Sudden spread spikes across the board for no reason = danger signal! Watch out for smooth trends. If...
AO, that's plenty of filters to implement; and all make sense.
  • #383
  • Sep 28, 2018 11:51am Sep 28, 2018 11:51am
Quoting alphaomega
{quote} ...
Watch out for smooth trends. If the trend is too smooth and too fast with very few corrective moves this is danger signal!
.
Hi AO, what do you think is the best quantitative way to measure the smoothness/corrective moves of a trend?

In hindsight we can detect that by just observing the chart; but programmatically it's harder.
  • #384
  • Sep 28, 2018 12:36pm Sep 28, 2018 12:36pm
Quoting alphadude
{quote} Hi AO, what do you think is the best quantitative way to measure the smoothness/corrective moves of a trend? In hindsight we can detect that by just observing the chart; but programmatically it's harder.
Hi alpha, if I may share my thoughts on on this..
AO is mentioning model3 when he talks about smooth trends, if i look at model3 i see price is not returning to mid range.
This fact alone is not enough, maybe add a time parameter? for example: when price did not touch mid range for an x amount of time, you are in danger of being in a smooth trend?
Or: if price did make new highs/lows x amount of times in an hour you are in danger of being in a smooth trend?
To know for sure when you are in a smooth trend is only to be seen in hindsight, so these thoughts are surely not 100% correct but better be safe then sorry.
Persistence will get it. Consistency will keep it.
  • #385
  • Edited 1:40pm Sep 28, 2018 12:50pm | Edited 1:40pm
Quoting Chake
{quote} Hi alpha, if I may share my thoughts on on this.. AO is mentioning model3 when he talks about smooth trends, if i look at model3 i see price is not returning to mid range. This fact alone is not enough, maybe add a time parameter? for example: when price did not touch mid range for an x amount of time, you are in danger of being in a smooth trend? Or: if price did make new highs/lows x amount of times in an hour you are in danger of being in a smooth trend? To know for sure when you are in a smooth trend is only to be seen in hindsight,...
will check it out with statistical analysis.

I previously used a concept called "Bar Factor" to determine that market within the current day is trending or not.

for up trend:
Bar factor = (Close - Open) / (High - Low).

for down trend:
Bar factor = (Open - Close) / (High - Low).

if Bar Factor is greater than 0.80 (meaning 80% full body or more); then we are likely in an up trend.

Tried this filter but did not improve the results in the backtest.

This bar factor filter *may* be useful in predicting if today will be a ranging day or not:
If previous day's bar factor is > 0.80; then today is likely to be a ranging day.

However this fitler did not improve the backtest results.
  • #386
  • Edited 7:06am Sep 29, 2018 2:45am | Edited 7:06am
Quoting alphaomega
Anything unusual?
As a matter a fact, yes there is something unusual.
I find it a bit hard to understand how the indexes are calculated, the numbers (bid or ask) don't add up.
Take for example EUR index, you can easely see sum of bids (or asks) is above 140 while the index is @ 134 in the screenshot.
I also tried some variations adding the ! colomn to the equation but no succes.

EDIT:
Nevermind this question, a kind member of the forum explained to me how it's calculated, thanks a lot luisinho!
Very straightforward multiply all members of the index by factor 100 except jpy pairs and then calc average.
Bit embarresed why i didn't think of this myself lol.
Persistence will get it. Consistency will keep it.
  • #387
  • Sep 30, 2018 11:05am Sep 30, 2018 11:05am
Quoting limprobable
{quote} The very nature of space-time is changed because it actually contains these changes of scale in an intrinsic and irreducible way to the ordinary space-time which is, in itself, in the physico-mathematical vision, a set of points. (...) In reality, this vision in which we represent the world in the form of points pretends to make measurements with infinitely great precision - at each small spatial interval corresponds a small time interval. However, it is quantum mechanics that tells us that it would take an infinite energy to be able to...
Unfortunately, it was complicated and difficult,
Some people simply understand it and use it
https://www.forexfactory.com/showthread.php?t=773361
be healthy and wealthy... Ceeport
  • #388
  • Sep 30, 2018 12:37pm Sep 30, 2018 12:37pm
Quoting alphadude
{quote} Hi AO, what do you think is the best quantitative way to measure the smoothness/corrective moves of a trend? In hindsight we can detect that by just observing the chart; but programmatically it's harder.
Measure the average and the maximum length of the trend corrections for the given period.
In the context of this strategy it's best to measure this length as a percentage of the range.

I give you example:

For Model 1 the average correction is above 50%. And you have at least 2 or more corrections above 100%.
For Model 2 the average correction is below 50%. But you have at least 1 or more corrections above 50%.
For Model 3 the average correction is below 25%. And the longest is below 50%.

In some extreme cases the longest correction for the whole cycle is below 30% and even below 25%. This is extremely smooth trend! But this type of cycles are very rare.

Play with the numbers, see what happens......
  • #389
  • Sep 30, 2018 1:08pm Sep 30, 2018 1:08pm
Quoting alphadude
{quote} will check it out with statistical analysis. I previously used a concept called "Bar Factor" to determine that market within the current day is trending or not. for up trend: Bar factor = (Close - Open) / (High - Low). for down trend: Bar factor = (Open - Close) / (High - Low). if Bar Factor is greater than 0.80 (meaning 80% full body or more); then we are likely in an up trend. Tried this filter but did not improve the results in the backtest. This bar factor filter *may* be useful in predicting if today will be a ranging day or not: If...
Do you have this indicator for MT4 please? I would like to look on this since the idea behind it is interesting. Thanks!

SF
  • #390
  • Sep 30, 2018 1:23pm Sep 30, 2018 1:23pm
Quoting alphaomega
{quote} Measure the average and the maximum length of the trend corrections for the given period. In the context of this strategy it's best to measure this length as a percentage of the range. I give you example: For Model 1 the average correction is above 50%. And you have at least 2 or more corrections above 100%. For Model 2 the average correction is below 50%. But you have at least 1 or more corrections above 50%. For Model 3 the average correction is below 25%. And the longest is below 50%. In some extreme cases the longest correction for the...
Alpha lets assume that we have played like these i certain day, and we close all trades when candle hits 50% like these right, for all three trades!?
Attached Image (click to enlarge)
Click to Enlarge Name: ll.png Size: 32 KB
  • #391
  • Oct 1, 2018 3:33am Oct 1, 2018 3:33am
Quoting shahed58
{quote} Unfortunately, it was complicated and difficult, Some people simply understand it and use it
Thanks for the link. It seems that they try to use just one idea "horology" but thats not imho a good way and far to be the most interesting idea developpe.

The difficulty is indeed to "translate" and use these apparently complex notions in trading. You have several options:
- understand in a global way and use these notions in the simplest possible way in trading
- search each detail and use one or two that gives you an advantage
- do not use these notions in trading if you are not familiar with it

Green pips
Bottomless wonders spring from simple rules, which are repeated without end
  • #392
  • Oct 1, 2018 9:46am Oct 1, 2018 9:46am
Inserted Video


All systems go! 8% profit on the first trading day! The testing procedure continues.....

Attached Image (click to enlarge)
Click to Enlarge Name: 54654654.png Size: 10 KB
  • #393
  • Oct 4, 2018 2:29am Oct 4, 2018 2:29am
hello, alpha , interesting strategy you have here, i am looking into it,
how is the performance so far since the last update 8% ?

also, do you think backtest result is accurate for this ea? because i am getting very interesting results from 99% backtest with variable spread using tick data suite,
  • #394
  • Edited 1:55pm Oct 11, 2018 8:01am | Edited 1:55pm
Tracking and logging your account drawdown (and other metrics) in real time is very important when you trade with automated systems.
Without accurate data you can't go very far and the probability for catastrophic mistakes is too high to ignore!
You need accurate statistics in order to optimize the systems for maximum performance in relation to the risk.

Now, you can use some free third party online apps (trade explorer), but this option is very unreliable and not accurate at all!
I came to this conclusion recently after some detailed "manual" analysis on my accounts.

So last week I wrote a few different loggers in order to track and to record some account/market metrics in real time.
The difference is night and day!

The conclusion?

If you trade with automated systems - do not try to use online trade explorers for system optimization!
Make your own tools! Make you your own statistics!

Here you can test one of my new tools: Drawdown logger.

Attached File(s)
File Type: ex4 Drawdown Logger.ex4   8 KB | 664 downloads


This indicator will log your account drawdown in real time with perfect accuracy! The indicator creates separate folder for each account and separate file for each trading day. You can change the log interval. The default interval is 60 seconds.

NOTE! If you change time frames or the settings while the indicator is working - this will restart the statistics for the maximum drawdown.
Simply attach the indicator to some empty chart adjust the log interval and leave it there.
As long as the platform is connected you will have perfect logs.

Attached Image (click to enlarge)
Click to Enlarge Name: drawdown logs.png Size: 69 KB
  • #395
  • Oct 11, 2018 9:27am Oct 11, 2018 9:27am
Quoting alphaomega
Tracking and logging your account drawdown (and other metrics) in real time is very important when you trade with automated systems. Without accurate data you can't go very far and the probability for catastrophic mistakes is too high to ignore! You need accurate statistics in order to optimize the systems for maximum performance in relation to the risk. Now, you can use some free third party online apps (trade explorer), but this option is very unreliable and not accurate at all! I came to this conclusion recently after some detailed "manual" analysis...
Outstanding piece of kit. Thanks for that.
  • #396
  • Oct 11, 2018 12:59pm Oct 11, 2018 12:59pm
Very nice AO thanks a lot!

I did make my own logger too for the same reasons you pointed out, but yours is much more detailed.
Persistence will get it. Consistency will keep it.
  • #397
  • Oct 11, 2018 1:19pm Oct 11, 2018 1:19pm
Nice work AO. how is the algo performing?
  • #398
  • Oct 11, 2018 1:44pm Oct 11, 2018 1:44pm
I would have loved to see this strategy in real time as the SP500 took a dump yesterday.
  • #399
  • Oct 11, 2018 2:04pm Oct 11, 2018 2:04pm
For the people who downloaded the logger indicator - I recommend you to download again because initially I uploaded the wrong version which will not work correctly on your MT4 because of missing files.

You can download again the correct version:
Attached File(s)
File Type: ex4 Drawdown Logger.ex4   8 KB | 892 downloads
  • #400
  • Oct 11, 2018 2:17pm Oct 11, 2018 2:17pm
Quoting alphadude
Nice work AO. how is the algo performing?
The system is working great for now! Now it's OFF for some upgrades and optimization.

I'm still testing on demo and on micro live accounts with different settings and risk levels.

Probably a few more weeks.....until I find the "best" settings and then I'm going live with larger account.

Small steps! No rushing. It's better to be safe than sorry
0 traders viewing now