DOSI: Daily Open Strategy Improved

User smilee89 presented a promising strategy in thread https://www.forexfactory.com/thread/post/13724868 based on retracement after an exceptionally large daily candle. If the trade moves against us, averaging down is used to close at break-even.

In the original thread, several posters made suggestions how to improve the strategy, but OP preferred to keep the thread strictly about the original strategy to avoid confusion. In this thread I would like to explore some modifications to try to improve the strategy. Feel free to contribute new suggestions, backtest results, etc.

For starters, here is the original strategy on GBPCAD using minimum bar length of 110 pips, and TP of 30 pips:

Version 1.00:

1. Exact implementation of the strategy described in the original thread.

Version 1.05:

1. EntryDelayMinutes: Entries can be delayed until few minutes after daily open for brokers that temporary stop trading at the session start, default is 15 minutes (cescof, post #2)

1. BasketTakeProfitPips: Option to close basket in profit instead of breakeven, default is 0 pips (cescof, post #7)

1. MaxDrawdownPercent: Close all trades when basket is in drawdown the specified percentage of the account balance, default is 25% (bdunn, post #3)

Version 1.10:

1. MaxOpenTrades: Limit the number of open orders of a basket, default is 999 orders (bdunn, post #3)

Version 1.15:

1. MaxSpreadPips: Wait with entry or scale-in until spread is below specified limit, default is 4.0 pips (cescof, post #21)

Version 1.20:

1. TradeDirection: Use to configurate whether only Buy or Sell orders are allowed (suppress other orders), default is "BothBuySell" (bdunn, post #22)

Version 1.25:

1. BarSelection: choose how to select large bars, either SelectFixedLength or SelectPercentile (yoriz, post #31).

In SelectFixedLength mode:

- MinBarLengthPips: minimum body length of previous bar to make entry

In SelectPercentile mode:

- LargeBarPercentile: percentage of bars that will qualify as large

- NrBarsLookback: number of bars to consider for percentile calculation

Version 1.30:

1. LotSizing: choose lot sizing method, either Compounding or FixedSize (cescof, post #30)

In Compounding mode:

- LotsPer10kEquity: number of lots to use for each 10,000 equity

In FixedSize mode:

- FixedLotSize: number of lots to use for each trade, regardless of equity

Version 1.35:

1. EntryStrategy switch to choose between RetraceAfterLargeBar, TrendContinuationAfterLargeBar or RandomDirection (yoriz, post #41)

- The RetraceAfterLargeBar option is the original strategy where we assume the market will retrace after a large candle. For example, after a large bull candle we go short.

- The TrendContinuationAfterLargeBar option does the opposite. We assume the market will continue in the direction of the large candle. For example, after a large bull candle we go long.

- To test whether the entry criteria make any sense at all, I've also added a RandomDirection option that does just that. Randomly make either a Buy or Sell trade. Each time you run the Strategy Tester you will get different results with this.

Version 1.40:

1. Added MovingAverageCross entry strategy. Use FastMABars and SlowMABars to configure length of SMAs (Toppi, post #42)

Version 1.45:

1. Added stricter input parameter validation to avoid problems due to wierd, invalid values during Optimizer runs (bdunn, post #56)

Version 1.50:

1. Added configurable MagicNumber parameter to enable running multiple instances of EA on same pair (bdunn, post #69)

Version 1.55:

1. Prevent reinitialisation after chart timeframe switch
2. Try to close all open trades upon fatal error before removing EA from chart

Version 1.60:

1. Retry order execution after non fatal errors
2. Added VerboseLogging switch to log orders and positions for troubleshooting (default off)

Version 1.65:

1. Changed order of EntryStrategy options to enable skipping RandomDirection option in optimizer
2. Added RandomizationSeed input parameter to enable searching best/worst drawdown using optimizer

Version 1.70:

1. Added ConcurrentBuyAndSell trade direction that allows two independent DOSI instances trading concurrently in both directions (post #82)
2. Removed VerboseLogging switch

Version 1.75:

1. Added CloseAllTradesOnFriday, FridayCloseHour and FridayCloseMinute input parameters to force closing all trades before the weekend at the specified time. Note that the time is specified in the broker time zone (in other words: ignores the BrokerGMTOffset input parameter).

Version 1.80:

1. Added MaxOpenBuyTrades and MaxOpenSellTrades. These can overrule MaxOpenTrades to create an asymmetric configuration if you suspect the market to be bullish or bearish. If both MaxOpenTrades and MaxOpenBuyTrades/MaxOpenSellTrades are non-zero, the latter will take precedence. (bdunn, post #89)

Version 1.85:

1. Improved robustness against requotes, network hickups, etc.

Version 1.90:

1. Log trade execution (e.g. latency, slippage, disconnects, etc.) to file (yoriz, post #101)

Version 1.95:

1. Added log file rotation to limit maximum file size (yoriz, post #106)

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

WARNING: This EA is experimental and still under development.

No attempt is made to remain backwards compatible between versions. Input parameters are regularly renamed or grouped differently to try to keep the ever growing list of switches and input values comprehensible. .set files of an old version might fail on the current version. Make sure you write down or take a screenshot of your optimized settings before upgrading!

Really thanks for effort.... I'm using an old dashboard to test this strategy and your support will be great...
Hope to find the right way....
Strategy seems good but dd is to consider always... First think I'd like to modify is time entries (some broker have few stop trading at the session start).
Second thing is the tp in case of grid.... I think that is useless close all trades after a big dd when a little profit is come.... I mean... If trend change after a big dd is better try to follow it till a new tp is set...
Just my 2 cents
And could you put spread set adjustabile?
Thanks
Regards

Thx for your effort to improve the strategy and a MT5 EA really work better in backtesting.

May I suggest adding two parameters to control the DD:

a. Maximum Open Trades

b. Stop Loss in Dollar Amount - Once this SL is hit, all trades will be closed.

We can't keep adding positions. Need to pull the brake when the DD reached a certain level.

Since MT5 back-tester has multi currency ability, I would warmly suggest you to add to this EA the option to choose which currencies you want to test within 1 back-test (by comma separated parameter value. i.e PairName = EURUSD,GBPCAD,GBPJPY...), and to adjust each currency to its specific MinBarLengthPips value.

What do you mean? Instead of 0:00 you want to enter at a different configurable time like 0:30? Or should we look at the spread and wait with entering until the spread is low?

What do you propose? A configurable TP in Pips? A configurable profit for all trades in the basket combined? A trailing stop?

I am a bit worried about this, because price averaging amplifies price movement. Since we are in this mode already because the market went against us, I worry we should get out as soon as possible to avoid unnecessary risk. But as with most things: let's try it out and see what happens.

Thank you for your suggestions!

Yes, absolutely. You have real tick data, simulated delays/slippage, see the actual drawdown during each trade (as opposed to after closing the trade like in MT4). And very interesting: you can test multi-currency strategies. We might want to try that later to see if we can use some pairs to compensate the drawdown of other pairs.

Sure. Although this does risk that prices keeps on moving further and the orders in your basket are too far away from the current price. Price would have to retrace a lot to have the basket in profit.

What if we close the oldest (i.e. worst) order when opening a new order (once we hit the maximum number of open trades)? We take some losses, but the open orders will be closer to current price so we require less retracement to close the basket in profit. We could even wait until the basket made so much profit that it offsets the loosing orders we closed earlier. What do you think?

This is already in the EA, and hardcoded at 25% of the account balance (as suggested in the original strategy of smilee89). I can make this a configurable input if you like.

Thank you for your suggestions!

Yes i mean configurable time ...

Yes a configurable TP different from original one when we have a grid openend (of course we have to test it but from my little experience after a big dd if trend change we could close better then a be)

And please configurable spread ( i've tried to Ea but i had just few trades even if daily range was set to 40 i suppose cause spread limit)

Thanks for your effort

Thank you so much yoriz. It is possible to convert to MT4?

I am not sure that is worth the effort at this point. I think we need to do a lot of experimentation to try out all sorts of settings and improvements. It does not make sense to do that in forward testing, as this would take years. Instead, we need to backtest to see how stable it is, what drawdowns we would have encountered in the past, etc. Backtesting this EA in MT4 is not very useful since we can't see the drawdown in MT4.

I suggest you install MT5 and play a bit with it. You will see the learning curve is limited, as it looks a lot like MT4, except for the Strategy Tester which is vastly improved.

Eventually when we settle on a nice, improved strategy, it might be useful to port back to MT4 just because that gives more choice in brokers to run the EA on.

I will download MT5 and test it, but the main disadvantage for me it's that I haven't got good quality tickdata to backtest. Could you suggest me how to get it? I have the data up to M1 from Dukascopy in .hst and export it to MT4, but I never get a good modeling of the tickdata. If you help me get it, I will help as much as I can.

As for the strategy itself, it is unacceptable to lose 25% of the equity in just 1 pair even 1 time in a decade. I think it would be interesting to find a balance in the pairs where we have the most favorable balance / dd ratio.

If we use 10 pairs for example, place ourselves in the worst scenario, what would be the maximum% willing to lose and how to recover it.

Haven't used it myself, but I read good things about Tickstory that can download tick data from Dukascopy.

Unfortunately that is hard to avoid with this strategy which doesn't have a real "edge". I tried inverting the logic, or using random buy/sell choices and still got similar equity curves. This means the strategy leans heavily on the averaging down strategy. It sort of works due to the small TP. During the day price typically fluctuates both up and down and will often hit the TP. I don't think retracement after a large candle is actually what we see here.

A strategy without an "edge" needs to rely on Martingale, Grid/Averaging Down, etc. and can go far, but will eventually fail. Not giving the strategy enough drawdown breathing room will kill the Martingale/Grid approach, but giving it too much can indeed set you back several years. Several traders have been successful for 10+ years with Martingale or Grid strategies, but they all use very conservative settings.

Yes, we should select a wide variety of non-correlated pairs to avoid multiple pairs going into drawdown at the same time. Backtesting will give a feel for what drawdowns, and what equity growth to expect.

Posted version 1.05 (see post #1)

1. EntryDelayMinutes: Entries can be delayed until few minutes after daily open for brokers that temporary stop trading at the session start, default is 15 minutes (cescof, post #2)

1. BasketTakeProfitPips: Option to close basket in profit instead of breakeven, default is 0 pips (cescof, post #7)

1. MaxDrawdownPercent: Close all trades when basket is in drawdown the specified percentage of the account balance, default is 25% (bdunn, post #3)

Thanks really appreciate but for backtest I think it's necessary set spread as adjustable...
Regards

Yes, that is very powerful feature. Looking forward to see how a larger set of EAs running in parallel makes the equity curve more smooth (hopefully).

Is it not possible to just select only the desired pairs when starting the Strategy Tester? Or will MT5 then feed ticks for all 28 symbols?

Not sure how to do that in a way that we can run the MT5 Optimizer. A comma-separated list won't work. Having a long list of placeholder parameters MinBarLengthPips1, MinBarLengthPips2, etc. feels a bit clumpsy and error prone if we accidentally think a parameter maps to a different symbol.

Perhaps we should use alternating parameters like this:

Symbol1=EURUSD

MinBarLength1=170

Symbol2=GBPCAD

MinBarLength2=110

etc.

What do you think?

Thank you for your suggestion!

Can you elaborate? How should we react to varying spread?

Sorry, maybe i've some problem with my mt5, because backtest show only few trades, and if i change min candel size i've trade just for a couple of days and then backtest stop...
Could you try GBPUSD from01/01/2021 min bar lenght to 60?
Thanks
regards

what's wrong??

Yes.
This will do the job and maybe adding TP param for each as well.

Symbol1=EURUSD

MinBarLength1=170
Symbol1TP = 50

Symbol2=GBPCAD

MinBarLength2=110

Symbol2TP = 30

Hi mate:

Again, thanks for your great work and valuable time.

While I still backtesting your EA Version 1, you already released a new version with new features. That's fast and responsive. Respect!

I have 1 question and 2 Suggestions in mind at the moment:


Question -

May I know what is the different between TakeProfitPips vs BasketTakeProfits?

We could have 1 trade and closed in profit vs 10 trades.

I will consider 1 trade as 1 basket as well, same as 10 trades in 1 basket.

So, mate, how the TakeProfitPips vs BasketTakeProfits works differently?

Suggestions -

1. As you know, in Metatrader 5, there is no “Common Tab" (unlike in MT4) which allows you to select Buy Only, Sell Only, or Buy and Sell.

Can you add user input parameters that allow user to select trade/backtest on Buy Only, Sell Only, or Buy and Sell? I hope you understand this clear.

This allow one to have a direction bias on certain pairs as the market condition keeps changing, and experiment/backtest if we need different MinBarLengthPips for an upward & downward market condition.

And, another example will be, if we have Sell trades in an clear upward trend market condition, we might need lower TP or higher SL.

Mate, we don't know any of these info/setting for each trading instrument, and it will be very useful & handy to have these user input parameters to backtest and improve the overall trading results, hopefully.

2. As previously mentioned, pls kindly add an user input "Maximum Total Open Trades".

This allow us to manage the risk in better way, together with the "Maximum Drawdown Percentage".

Mate, we try to earn some alpha (money) while managing the risk. And most probably will trade multiple pairs in portfolio basis.

Let's see if the Total Opened Trades reach a certain max level, the price will retrace & settle the clear direction, and the price will re-bounce and hit our TP, without we risking more trades/leverage.

Together with all above mentioned user inputs, and with the advantage of MT5 platform for multi pairs backtesing feature, we might able lower the drawdown, leverage needed and risks, either for 1 single pair or multi pairs trading.

Yoriz, if anything you need further explanation from me, do let me know. I am here to help & share my thoughts, and hope we all can improve the EA with better trading results.

Once again, mate, many thanks for your valuable time.

Respect,
Bill

The TakeProfitPips parameter is from the original strategy. It sets the TP level of the first order opened. The BasketTakeProfitPips parameter was added on request of cescof to not close the basket at breakeven, but wait a bit longer before closing it. In short, if you have only one open trade, the TakeProfitPips parameter is used, if you have two or more trades open, the BasketTakeProfitPips parameter is used.

How should that work? Should the EA just ignore Sell/Buy opportunities and move on, or should the EA force to Sell/Buy orders even if the strategy suggests the opposite?

Sure. Added MaxOpenTrades parameter to version 1.10. See post #1 for the download link.

Thanks for your contributions!

If the backtest stops, then the EA must have logged an error message. What does it say in the "Journal" tab?