Everything is the same, but the results are different. Help.

New comment
 

These are the results of my strategy, both use data based on real ticks of the same currency pair and the same strategy with the same inputs and conditions, the only difference is that the real tick data from one is from Pepperstone broker and the other is from Dukascopy. Would anyone know why they show different results despite opening exactly the same positions?

This one is from Pepperstone

This one from Dukascopy


And as you can see, the positions are the same, yet the results in balance are different.

From dukascopy:


From Pepperstone:


help :(

 
Angel Sosa:

These are the results of my strategy, both use data based on real ticks of the same currency pair and the same strategy with the same inputs and conditions, the only difference is that the real tick data from one is from Pepperstone broker and the other is from Dukascopy. Would anyone know why they show different results despite opening exactly the same positions?

This one is from Pepperstone

This one from Dukascopy


And as you can see, the positions are the same, yet the results in balance are different.

From dukascopy:


From Pepperstone:


help :(

Forex and CFD's are over-the-counter (OTC) markets where each broker-dealer is its own captive market. Different prices = different results.

 

...and this is why 2 traders using the same ea can have very different results: 1 can make millions, while the 2nd trader will lose everything, even when trading the same period of time. This is nothing new.

the take away from this? always test an ea on your own broker before you put it live on your account; no matter if another trader made millions; your broker may have these same differences shown above, and you will blow your account.

 
...or trade futures contracts on a real centralized exchange like the Chicago Mercantile Exchange to avoid the debacle.
 
Michael Charles Schefe #:

...and this is why 2 traders using the same ea can have very different results: 1 can make millions, while the 2nd trader will lose everything, even when trading the same period of time. This is nothing new.

the take away from this? always test an ea on your own broker before you put it live on your account; no matter if another trader made millions; your broker may have these same differences shown above, and you will blow your account.

Thank you for responding, I agree that testing with the broker's data is the most accurate approach. The issue is that I downloaded data from Dukascopy to evaluate the strategy over longer time periods (5 years) since with Pepperstone, using the "Every tick based on real ticks" model gives me more accurate results with a demo account I have, but I can only test one year back with that model. The further back I go, the lower the history quality percentage becomes. That's why I downloaded ticks from Dukascopy, but I found that even though both execute exactly the same trades, the balances are different.
 
Angel Sosa #:
Thank you for responding, I agree that testing with the broker's data is the most accurate approach. The issue is that I downloaded data from Dukascopy to evaluate the strategy over longer time periods (5 years) since with Pepperstone, using the "Every tick based on real ticks" model gives me more accurate results with a demo account I have, but I can only test one year back with that model. The further back I go, the lower the history quality percentage becomes. That's why I downloaded ticks from Dukascopy, but I found that even though both execute exactly the same trades, the balances are different.

sounds like good reasoning; i am sure that i would have done the same thing, however, that only makes our point (punn intended haha), that like RYan said, a broker is its own market with its own price feed; and that that this is a normal issue, and that results on different brokers dont mean much, other than that the ea is obviously better run on 1 broker than the other.

 
But if you have the source code, then, you might be able to find out the main cause of the differences, and may be able to avoid much of the difference if coded better.
 
If the EA makes use of really tight trailing stops then this is probably what's causing the difference. Also different datasets usually have different spreads which also affects how trailing stops function since they are also executed at bid and ask prices. Also at times where news events occur I get  completely different high and low values for the same candles when comparing dukascopy and my broker and this is by hundreds of pips. So if you want something to be stable across all brokers then you need to learn to code a specific news filter for the strategy tester and one for a live account. The spread differences also apply to right after market close at 5pm EST. Dukascopy will have comparably smaller spreads than my broker so I take that into account by not trading until like 9pm EST. 
 
That's why there is something like "premium data" and "institutional grade data" are being sold such as LME data can be the purest form of data which is not available for free. CFD and Futures are products where people bet if it's going to up or down based on its expiry. They can't own this but place bet until this derivative is expired. Since there can be multiple players in each exchange with different views of market, price may vary. The most number of players on any exchange contribute to the chart being liquid. The less number of players in any exchange contributes to the chart being illiquid.

On what basis players decide if price may go up or down? The answer is technical analysis and where do they analyse this? The answer is Spot or Index. Spot is pure form of data where manipulation is not easy. Future and CFD can be easily manipulated. The low liquidity on any exchange means high manipulation, that is why people trade symbols  where there is good volume and liquidity to avoid issues such as slippage and benefit with tight bid ask spread. Your broker has liquidity and tightest spread or not as compared to other brokers ? Only you can find. You may often see on spot, less long wicks as compared to futures or CFD.  Index is another form of data which is based on calculations and not based on volumes. Volume is causing prices to be different from different exchanges

Every exchange has a magnet type algorithm set which never let price go too far than what it's being traded on other exchanges. When price manipulation happens, you often see a long wick. The long wick suggests that somebody tried to add too many orders and tried to push price far away but price algorithm controlled it back to original price of index. But it's rare to notice such activity on index or spot. The reason of manipulation is also because high leverage provided on CFD and Futures but in spot it's not provided. Manipulation may be the primary cause for your EA to fail on CFD and Futures.

I advise you run your EA on spot or index and perform actions on futures or CFD without worrying about price of CFD and Futures to overcome this problem.


 
Angel Sosa #:
Thank you for responding, I agree that testing with the broker's data is the most accurate approach. The issue is that I downloaded data from Dukascopy to evaluate the strategy over longer time periods (5 years) since with Pepperstone, using the "Every tick based on real ticks" model gives me more accurate results with a demo account I have, but I can only test one year back with that model. The further back I go, the lower the history quality percentage becomes. That's why I downloaded ticks from Dukascopy, but I found that even though both execute exactly the same trades, the balances are different.

Compare spreads tick by tick (or at least, second by second).

Large markup (plus swaps, comissions, fee, etc.) can kill every strategy.

 
Casey Courtney #:
If the EA makes use of really tight trailing stops then this is probably what's causing the difference. Also different datasets usually have different spreads which also affects how trailing stops function since they are also executed at bid and ask prices. Also at times where news events occur I get  completely different high and low values for the same candles when comparing dukascopy and my broker and this is by hundreds of pips. So if you want something to be stable across all brokers then you need to learn to code a specific news filter for the strategy tester and one for a live account. The spread differences also apply to right after market close at 5pm EST. Dukascopy will have comparably smaller spreads than my broker so I take that into account by not trading until like 9pm EST. 
Yes, in this strategy I use very aggressive trailing stops on each tick based on the fact that I want a higher win percentage rather than large profits. Thank you for sharing your experience. I've also noticed that there are certain cases where inputs work perfectly in both Pepperstone and Dukascopy data. I added news, spread filters and a range of hours in which it can open positions but unfortunately, if I previously had about 15 profitable variants, I ended up with only 2 because they're the only ones that worked across multiple brokers, which gives me the most confidence.
12
New comment