Mean Reversion Strategies

Greetings folks, I hope you are all good.

I just to ask if anyone has ever traded a mean reversion or similar strategy. That is placing a trade when you think the price will correct or revert back to the recent moving average after an initial breakout.

I am very interested to know what your current setup for a trade is if you do; at what point do you decide to enter that trade, where do you set your stop loss and take profit and what indicators if any do you use for your trade decision.

Much appreciate your input.

Haha, I’ve tried loads of them. Never stuck to them though. The problem is they are normally against trend and therefore are less reliable. I used the 100 ema or any slower moving MA. Then I would trade price back to it on any reversal setup eg head and shoulders, divergence, support resistance etc… so any reason to sell back to the MA. Stop loss at high and exit at MA. You get loads of break evens but the winners were good risk reward. I dont use it because I just didnt have the patience to wait for them, preferred day trading to get in and out.

I wouldnt recommend day trading it on a faster MA. Price usually just keeps moving up until you’ve been stopped out multiple times and then sometimes drops but not always. The worst would be price extends away from for eg 20ma and when you think its stalling it will pullback slightly and then keep going. It’s very frustrating and requires good discipline and emotional control.

1 Like

@ tradeforex077

Greetings thank you for the reply. I think I have found something that can be a starting point for a signal generator. It is a modified version of your setup and you would need to apply some of the filters you normally use to model and refine it to your needs.

It is a mean reversion trade for false breakouts.

it consists of:

  • change in 50 SMA H1: SMA50(0) - SMA50(1) (this measures the steepness of the change in the moving average t the end of the last bar): Current Change (CChng)

The Indicator Looks like this (Green line)

  • Z-value using above change compared to previous 1,000 changes in SMA50

    Taking the last 1,000 changes in H1 SMA50, find the mean change and Standard Deviation (StDev) of the changes.

The Z value is: (CChng - Mean )/ StDev - This measures how many standard deviations the current change is from the average

The Indicator Looks like this:

  • PlusMinus Directness: ((Close[0] - Close[50]) / 50) / ATR(50)

The Indicator looks like this (Purple- Red line is moving average (12) of first indicator)

  • Bollinger Band at Sigma 2 (in the main chart, there is a bollinger band at sigma 2 in black)

The Entry for Buy is:

(PlusMinusDirectnes >= AveragePlusMinusDirectness && Zscore >= -1 && Zscor <= 0.5 && Close[0] <= LowerBollingerBand)

These are situations where the price has dropped (broken out) below the lower Bollinger Band at sigma 2, but the corresponding rate of change in the moving average is low and still within -1 sigma compared to its last 1,000 values.

The price is outside its circa 97% probable range (sigma 2) but there is not a corresponding change the direction of the trend.

Take Profit: Mid point between SMA50 and current price
Stop Loss: Double the distance above

Note: the main feature that affects the win rate is the sigma value of the Bollinger Band. The higher the sigma value the higher the win rate.

You can also apply your own filters to improve the results but this seems to be a good starting template.

Note: For sell trades, use the opposite logic to the Buy entry.

Sample backtest:

AUDUSD BUY H1 2009 - 2019

Zscore >= -1 and Close[0] <= LowerBollingerBand(Sigma 2)

Win Rate = 75.29%
Break Even Win Rate = 66.67%

GBPUSD BUY H1 2009 - 2019

Zscore >= -1 and Close[0] <= LowerBollingerBand(Sigma 2)

Win Rate = circa.68%
Break Even Win Rate = 66.67%

USDCAD BUY H1 2009 - 2019

Zscore >= -1 and Close[0] <= LowerBollingerBand(Sigma 2)

Win Rate = 70.27%
Break Even Win Rate = 66.67%

USDCHF

BUY H1 2009 - 2019

Zscore >= -1 and Close[0] <= LowerBollingerBand(Sigma 2)

Win Rate = 71.59%
Break Even Win Rate = 66.67%

NZDUSD BUY H1 2009 - 2019

Zscore >= -1 and Close[0] <= LowerBollingerBand(Sigma 2)

Win Rate = 73.25%
Break Even Win Rate = 66.67%

I think there is something in the above reversion trade. But it will need some work to get it right.

Hi - I’m a contrarian by nature and use Bollingers and some ema s together with support and resistance - I’m wondering just now whether using tick charts and the “squeezing” of “dead periods” would potentially make some indicators, including perhaps BBs less susceptible to “flattening out” ?

1 Like

Wow. All these stats and filters is probably a bit over my head but it looks good. I agree it can be profitable, just gotta stick with it!

Hi thanks for the reply. Can you explain this in more detail.

I have found Support and Resistance (S&R) points tricky to work with. I did find some statistical evidence to reinforce their use but the edge improvement was marginal (1-2%).

Tying that into what you said about tick data; I think to really benefit from S&R you would have to calculate the S&R levels very accurately using the order book of your vendor.

With Price Action trading we try to interpret what is happening in the order book by looking at the ultimate changes in price from the interaction of buyers and sellers in the auction process.

It would be nice to trade using signals created by modelling inputs at a granular level of detail from the order book directly. Not HFH, but evaluating slightly longer term trends in those inputs.

I think you’re right about this. A modification might be to find reversal opportunities when the market is clearly ranging/range bound. (when price will reverse from outside Bollinger Band at sigma 2 in those circumstances)

Always found day trading lower time frames tricky because there is more noise (error) in the signals. Like how it is much easier to find a clear pattern from a trend line than the raw price action

I think higher time frames (H1+) summarise market sentiment better. For shorter time frames you would probably need to look at the order book to understand how the random timing of market orders create price changes in the shorter term, instead of looking at the cumulative result in higher time frames.