This EA takes advantage of statistical distributions and the mean-reversion properties during the London Session open plus takes advantage of Law of Large numbers to do compounding due to stable long-run win/Loss ratio. It also takes small bites, mostly 5 to 20 pips. If you use only the default settings with compounding enabled it has very stable Win/Loss ratio of  9:1 . i.e On average you will win 9 trades before you loose 1 trade. 

When trading with this EA i don't concern myself with 1 or 2 loosing trades in a row since it has a stable Win/Ratio and that at the end of each month/ year i will have a good net profit.

The law of large numbers, or LLN for short, ^[1] is a theorem from statistics. It states that if a random process is repeatedly observed, then the average of the observed values will be stable in the long run. This means that as the number of observations increases, the average of the observed values will get closer and closer to the expected value.

NOTE:  In FX trading , Compounding works well only if you have an EA with long-run stable win/loss ratio and requires patience because its a long-term game but you also need to have confidence in the Win/Loss ratio of an EA.

Compounding profits and Law of Large Numbers : Using statistics analysis , if you find generalized optimal parameters with a very stable Win/Loss ratio , e.g 9:1  then we can enable compounding of profits (Note compounding  your profits is not Martingale and this EA does not use martingale ) .

i have decided to divide the compounding and LLN analysis into 3 phases with respect to number of trades :

Compounding phases and example:

image 1:

Assuming no withdrawals in 10 years :

Starting Capital = USD$ 100,000

Reinvested Profits/capital = USD$  100

Number of Months  = 120 ( 10 years ) 

Monthly Profits %  =  4.5 %

Balance after 120 months = USD$ 20,111,857

Image2:

Assuming a monthly withdrawal of 0.5% of profits for 10 years :

Starting Capital = USD$ 100,000

Reinvested Profits/capital = USD$  100

Number of Months  = 120 ( 10 years ) 

Monthly Profits %  =  4.5 %

fixed monthly withdrawals = 0.5%

Effective Monthly Profits = 4.0%

Balance after 120 months = USD$ 11,340,412

Total withdrawals in 120 months = USD$ (20,111,857 - 11,340,412 ) = $ 8,771,445

Using a sample of  5000 trades

1. Ramp up phase == Slow growth Phase ( first 40% of trades = 2000 )

2. Acceleration Phase == Medium growth Phase ( next 30% of trades = 1500)

3. Exponential Phase == Fast growth Phase   ( last 30% of trades = 1500)

Here is a snapshot of the backtesting results from 2010/01/01 - 2020/12/31 using Default settings with Compounding factor of 2 enabled. When performing backtesting please use 1-Minute OHLC for modelling  and TimeFrame of M5 . The EA does not use ticks but only Bars . All calculations are performed after the close of previous bar or at bar Open and therefore using ticks is not required if you have clean 1-Minute OHLC data. Statistical distribution on 1-Minute OHLC is enough for the EA to perform various calculations for parameter values for trading signals :

The AlgocepGrid EA Version 3.9  backtest analysis with and without profit compounding. NOTE: This EA  by default  DON'T  use GRID trading. its disabled and not recommended. i don't use grid trading myself. 

Three factors are very important in Profit compounding while using an EA 

1.  Compounding period .

2.  Stable average long-term Win/Loss Ratio (9:1 or 90% winrate) and also average long-term Profit Factor above 1.1

3.  Low Equity Drawdown ( Less than 25% --Same as most Hedge funds )

4.  Initial Compounding capital equilibrium (OPTIONAL) == Amount of starting capital that will enable you to jump over the Ramp up phase and start directly at Acceleration Phase or Exponential phase of profit compounding. 

Below i have attached the BACK TEST performance  for the past 16 Years, 11 years and 2 years  of AlgocepGrid EA. You can see that from the backtest analysis , the EA meets the requirements of a good compounding routine.

 16 Years profit compounding both after every Win and Loss:                                         16 Years profit compounding after every Win only:                                         16 Years with NO profit compounding:                      

11 Years profit compounding both after every Win and Loss:                                         11 Years profit compounding after every Win only:                                         11 Years with NO profit compounding:     

2 Years profit compounding both after every Win and Loss:                                         2 Years profit compounding after every Win only:                                         2 Years with NO profit compounding: 

In order to replicate the backtest results above it assumes that your back testing initial deposit is USD $ 100,000 with Leverage of  1:50 . You can also use other initial deposits amounts and leverage values during your back test. 

incase you are using an account with $100 or $ 1000, or $ 10000 and you don't want to run the backtest again then you can simply divide it by 100,000 and then multiply by the net profits and you will have similar values 

e.g  Starting with BACKTEST initial deposit of  $ 1000 and in 16 years of profit compounding  , you should get  (1000/100000) *15,075,878 =  $ 150,758 as Net Profits:

The backtest example calculation in this immediate above comment  is not correct since it also depends on a few things .Ignore it and just run the backtest with your initial back test deposit 

Why is it that many FX traders believe you need tick-by-tick data for back testing every EA Strategy to get accurate results ?

If your strategy calculations don't not require actual tick-by-tick calculations , You don't need tick-by-tick data for back testing. This notion that you need tick-by-tick historical data for back testing every strategy for it to be accurate is completely useless and debunked.

One of the main reason that most EAs or strategies require tick-by-tick historical data is because they tend to use a ton of indicators for the strategy. 

If your EA or Strategy does not use indicators or tick-by-tick historical data for actual signal calculations  then you completely don't need tick-by-tick historical data . The 1-Minute OHLC (M1) is more than enough . Note : also that with MT5 , ticks that are missing are synthetic and automatically  generated. 

Example:

AI based Strategies  ===>    Mostly need tick-by-tick so that you can perform binning of the data to create various types of bars suitable to the Deep Machine Learning algorithm  in use.  Mostly, i  will use very clean tick-by-tick historical data for AI strategies only.

Standard EAs that use a ton of indicators  ===> End up needing tick-by-tick data for accurate back testing in most cases.

Strategies/EAs that DON'T use Indicators or tick-by-tick calculations to generate trading signals  or only calculate their trading signals during new Bar Open or Previous Bar Close ==> DON'T need tick-by-tick historical data for backtesting.  Use 1-Minute OHLC or Higher e.g M5 , M15 , M30 or H1

I hope this helps newbies.