Simple Mean Reversion | Page 27 | 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

  • #521
  • Jun 25, 2019 1:09am Jun 25, 2019 1:09am
Quoting alphadude
{quote} Interesting way of classification of the market. Once a market is classified in a given cycle (i.e. daily cycle), a strategy for that model is activated for the day. The challenge is how to predict the model that we are in during the day. the following is a potential quantification algorithm that should be run at each bar in the day, in order to classify which dimension we are in: 1. T = total number of bars bars since day start (i.e. using 1m bars) 2. U = count how many breakouts to the upside 3. D = count how many breakouts to the downside...
AD, using your suggestion, at which point (e.g. after 2 hours into the start of trading cycle): would you determine which dimension we're in?
  • #522
  • Jun 25, 2019 1:38am Jun 25, 2019 1:38am
Quoting carddard
{quote} AD, using your suggestion, at which point (e.g. after 2 hours into the start of trading cycle): would you determine which dimension we're in?
i think 4 hours is a good start, or perhaps use min of 50% ADR is reached during the day, then start classifying.

no point trading a very narrow session. to be avoided at all times.

george from tma thread uses the min adr, which i found to be useful for trading current day, or current session.
  • #523
  • Jun 25, 2019 3:20am Jun 25, 2019 3:20am
Quoting alphadude
{quote} i think 4 hours is a good start, or perhaps use min of 50% ADR is reached during the day, then start classifying. no point trading a very narrow session. to be avoided at all times. george from tma thread uses the min adr, which i found to be useful for trading current day, or current session.
To make sure I am understanding this right, are you saying that one could use the following ways to quantify the "state" of the market using your defined rules:

1) State of the market would be obtained during the 1st 4 hours of market open (i.e. at 00:00 hrs) - This gives me about 8 hours left to open trades (if mid cycle hour = 12)
2) Once 50% ADR has been reached (could be within 1st hour or 6th hour of market open), state of the market would be obtained

Or perhaps a combination of both, whichever comes first?
  • #524
  • Jun 25, 2019 3:25am Jun 25, 2019 3:25am
Quoting carddard
{quote} To make sure I am understanding this right, are you saying that one could use the following ways to quantify the "state" of the market using your defined rules: 1) State of the market would be obtained during the 1st 4 hours of market open (i.e. at 00:00 hrs) - This gives me about 8 hours left to open trades (if mid cycle hour = 12) 2) Once 50% ADR has been reached (could be within 1st hour or 6th hour of market open), state of the market would be obtained Or perhaps a combination of both, whichever comes first?
Either of the 2 conditions is fine; for example; if the market reached a 50% ADR in the first hour; it is up for something sinister; see for example 24th June 2016; during Brexit vote count.

After you have started classifying the market; you need to re-do the classfication all day long, at every 1minute bar.

If the market changed from trending to ranging; trade it accordingly; basd on your algo for that type of market.

We don't know how the market will behave; sometimes it changes from ranging to trending, and vice versa during the day.

Some quants uses pattern recognition machine learning algorithm to classify the market regime intraday; based on past pattern pre-selected by the user for the AI to learn from it.

however; I have gone that route and spent lots of time on it; and beleive me it is not worth the hours. It's better to use simple classification algorithm; like the one mentioned above.
  • #525
  • Jun 25, 2019 3:40am Jun 25, 2019 3:40am
Quoting alphadude
{quote} Either of the 2 conditions is fine; for example; if the market reached a 50% ADR in the first hour; it is up for something sinister; see for example 24th June 2016; during Brexit vote count. After you have started classifying the market; you need to re-do the classfication all day long, at every 1minute bar. If the market changed from trending to ranging; trade it accordingly; basd on your algo for that type of market. We don't know how the market will behave; sometimes it changes from ranging to trending, and vice versa during the day....
Have you tried successfully implementing your classification algorithm on AO's mean reversion method?
  • #526
  • Jun 25, 2019 6:38am Jun 25, 2019 6:38am
Quoting carddard
{quote} Have you tried successfully implementing your classification algorithm on AO's mean reversion method?
I didn't have a chance to implement it or test yet
  • #527
  • Jun 25, 2019 12:23pm Jun 25, 2019 12:23pm
I've been looking for an ADR indicator like this one. I like to work within statistics.
Attached Image
  • #528
  • Jun 25, 2019 8:08pm Jun 25, 2019 8:08pm
Quoting alphadude
{quote} I didn't have a chance to implement it or test yet
I gave this a go last night. I had to modify the percentages though, to make them more realistic.
I tried variations of implementing the quantification algorithm - for e.g., using it as an additional filter for entries or exits. Couldn't seem to find an additional edge because of this. Maybe there's a better way for me to utilize this.
  • #529
  • Edited 10:00pm Jun 25, 2019 9:46pm | Edited 10:00pm
What did you modify them to?
  • #530
  • Jun 26, 2019 3:38am Jun 26, 2019 3:38am
Quoting JaxPacific
What did you modify them to?
Tested a few ranges.
I noticed that the percentages do not differ too much. I increased the 20% to 25% and the 1% is no longer relevant. The difference between the 2 percentages is more important, I realised.
  • #531
  • Jun 26, 2019 5:57pm Jun 26, 2019 5:57pm
Yeah, 50 seems high unless you're taking a small target with a big stop loss, but then that topic had already been brought up here I see, back in the first few pages.

You might have to explain more how you used the difference.. but I was going to type more on my earlier comment, but held back to see what you said first..

Anyway.. That is why I posted the ADR % concept in that picture. I believe the best way to go.. added filter also or not, is to take a percent of the ADR and then take a move of that. If 90% of days are over 64 pips and 50% are over 108.. that sure tells me something. Simple filters should take care of the bouncing points, then enter the trades at the extremes, 2 positions, one for the near term ADR target, move stop to BE and then the other for a larger percent of the range. I think I would shoot for 65% of the time for probability sake, and then take half of that move from the HOD or LOD plus bounce filter as the entry. I don't care where or when it bounces as well.. as long as I am already in the trade in the opposite direction, and it's finally going to make the move to the larger percent of the ADR. So trend or range doesn't matter.

Anyway, what percentage difference worked the best? you can PM me if this is all hush hush or something.
  • #532
  • Jun 26, 2019 8:02pm Jun 26, 2019 8:02pm
Quoting JaxPacific
Yeah, 50 seems high unless you're taking a small target with a big stop loss, but then that topic had already been brought up here I see, back in the first few pages. You might have to explain more how you used the difference.. but I was going to type more on my earlier comment, but held back to see what you said first.. Anyway.. That is why I posted the ADR % concept in that picture. I believe the best way to go.. added filter also or not, is to take a percent of the ADR and then take a move of that. If 90% of days are over 64 pips and 50% are...
It doesn't matter. There is no additional edge. It could be good for other uses I'm not sure if it adds incremental value to AO's strategy.
  • #533
  • Jun 26, 2019 8:38pm Jun 26, 2019 8:38pm
Quoting carddard
{quote} It doesn't matter. There is no additional edge. It could be good for other uses I'm not sure if it adds incremental value to AO's strategy.
i did quick test, and you are right. it does not seem to help at all
  • #534
  • Jun 26, 2019 11:05pm Jun 26, 2019 11:05pm
I think you may be right. Every time I go to try and make His system work, I end up with.. pretty much a whole new system. That's happened like 3 times. I know for sure the original system wouldn't work as is.. or even with that one small tweak. Question is.. how far away from it do you guys want to get and still call it the same system?

This was the latest. I used volume waves, m and w formations, and an MTF ATR grid set at 14per and 20%. Not exactly The same system.. But I guess I kind of put them all in the same category.

Strictly or blindly keying off of the high or low doesn't make sense to me. So we can add PA, volume, and bounce TA type filtering a few different ways and change the furthest from the original system. Or.. I still think the easiest is a hedge sort of formula. All I would need to know is the accuracy rating of the system before/between the huge draw down days. Then I think we could Use a hedge system, limit scaling in to like 3.. have B/E point and make sure the streaks outpace he hedge losses and the hedge captures the big move days. Take some tweaking, but that is the only way I see it happening based on pure math without going down the pair correlation path that the others seem to have traveled.
Attached Image (click to enlarge)
Click to Enlarge Name: example.png Size: 58 KB
  • #535
  • Jun 27, 2019 9:42am Jun 27, 2019 9:42am
I did some statistical analysis of price movement after a breakout of x pips from the daily range. Negative values means the price retraced on average. Positive values means the price continued in the same direction of the move.

here is the result for EURUSD; from 2011 todate.

Some observations from the chart:
- There is a negative correlation between the day's range and retracement; which means; the larger the value; the more likely it will retrace
- On average; EURUSD reverses course after a 150 pips of daily range; which is interesting to know. This means EURUSD tends to continue the move if the price range is below 150 pips.
- As the day range approaches 200 pips; there is an edge there; the average reversal movement is 20 pips
Attached Image (click to enlarge)
Click to Enlarge Name: T2818_sc_20190627-153853 eurusd.png Size: 113 KB
  • #536
  • Jun 27, 2019 9:51am Jun 27, 2019 9:51am
AUDUSD has higher reversal rate compared to EURUSD

GBPUSD is a breakout beast; don't attempt to fade it!!!! You are better off trading breakout strategies on GBPUSD especially after 200 pips range

GBPCAD is even worse than GBPUSD; never fade that monster!

EURCHF is also bad for fading

NZDUSD is perhaps the best for fading so far
Attached Image(s) (click to enlarge)
Click to Enlarge Name: T2818_sc_20190627 audusd.png Size: 114 KB
Click to Enlarge Name: T2818_sc_20190627 gbpusd.png Size: 120 KB
Click to Enlarge Name: T2818_sc_20190627 gbpcad.png Size: 116 KB
Click to Enlarge Name: T2818_sc_20190627 eurchf.png Size: 106 KB
Click to Enlarge Name: T2818_sc_20190627 nzdusd.png Size: 110 KB
  • #537
  • Jun 27, 2019 9:55am Jun 27, 2019 9:55am
Quoting alphadude
AUDUSD has higher reversal rate compared to EURUSD GBPUSD is a breakout beast; don't attempt to fade it!!!! You are better off trading breakout strategies on GBPUSD especially after 200 pips range GBPCAD is even worse than GBPUSD; never fade that monster! EURCHF is also bad for fading NZDUSD is perhaps the best for fading so far {image} {image} {image} {image} {image}
excellent charts

you're sharing gold here...
  • #538
  • Jun 27, 2019 9:57am Jun 27, 2019 9:57am
Quoting alphadude
AUDUSD has higher reversal rate compared to EURUSD GBPUSD is a breakout beast; don't attempt to fade it!!!! You are better off trading breakout strategies on GBPUSD especially after 200 pips range GBPCAD is even worse than GBPUSD; never fade that monster! EURCHF is also bad for fading NZDUSD is perhaps the best for fading so far {image} {image} {image} {image} {image}
Have you tested this with tick data ? or at least taking the spread of each pair into account?
  • #539
  • Jun 27, 2019 10:05am Jun 27, 2019 10:05am
Quoting DavidRP
{quote} Have you tested this with tick data ? or at least taking the spread of each pair into account?
I just used the daily bar OHLC. Pretend you enter at the breakout price, and exit at EOD. No transaction cost involved in the analysis.
  • #540
  • Jun 27, 2019 10:05am Jun 27, 2019 10:05am
Quoting alphadude
{quote} I just used the daily bar OHLC. Pretend you enter at the breakout price, and exit at EOD. No transaction cost involved in the analysis.
good enough!
2 traders viewing now