every tick based on real ticks - MT5, huge variance in results
-
A question if I may, I have been running tests on MT5 for my strategy using 'every tick'. I was getting amazing results on a 1m 'trend following' strategy in the NY session, with my test showing circa $50K profit in a year (on $100k). I then switched to 'every tick based on real ticks' and got a $90k loss. Can it really be that different?
I understand that the affect depends on the type of strategy. I was originally trying to snipe entries and would need tick data, but I've changed it so that all of my conditions are at candle close. But I still get -$90k using 'real ticks', whereas OHLC is +$25k. Can anyone give me any more insight? Is the MT5 tester totally useless for testing scalping strategies therefore?
-
Those results are perfectly expected. Any back test other that based on real ticks is a waste of time leading to false results. Sorry for the bad news.
-
Ah, so I now gather, back down my hole, I'll cancel that order for a new audi
I have now got myself a nice set of live data to test on, courtesy of eareview.net and Dukascopy.
What confuses is me is that I still get vastly different results between 1min OHLC and 'every tick based on real', one shows a loss of $5k and the other a loss of $90k over a 6 month period on a $100k bal. This is despite all my conditions being on candle close. It's not even close. what am I missing?
-
@martymoon120 There have been a number of threads on this, one that I recently found which was very helpful.
Essentially, what's happening is the testing software makes your 1M OHLC backtest look a LOT better than real results because there is a flaw in the software that interpolates what happens between those 4 values (OHLC). So, in most cases when you have an open order the testing software 'gives' you a perfect entry, right at your asking price, it also gives you a perfect exit, even if it doesn't have the data near your TP or SL.
The best way to explain this is like this: If have have an M15 candle, that is 15 pips in total size (high to low). There are 4 values associated with those 15 pips, OHLC. Let say you placed a trade at the open of that candle and had your SL 5 pips from the entry and a TP was also 5 pips from the entry. Using OHLC data the backtesting software will likely tell you that your TP was hit and not your SL, even though the SW does not have the data to definitively tell you. It's just randomly generating data inside that 15M candle between the OHLC.
Some parameters can be found using 1M OHLC, but anything that dictates TP or SL (Trailing stops) should be optimized and testing on Tick data.
-
@cmuphyfx Many thanks, understood. So for my purposes testing on MT5 is a waste of time. So you suggest Tick Data Suite, right?
-
@cmuphyfx ah you mean download the data from TDS and import it into MT5 and do the test on there...? TDS don't have a testing platform do they? Or do you test on something else?
-
Test on real ticks, unless you are just testing something, and want to quickly visually see what's happening.
-
I typically use 1M OHLC data to find rough parameters, and then use tick data for final optimization with trailing stops.
-
But you do this on MT4/5 (using TDS data)? Not some other platform...? And am I right in saying there isn't a software/platform that will run through all the different combinations of my strategy? For example, what I working on right now, I don't know whether to have my SL at 50, 75, or 100 pips, or trail it, whether to start it behid a candle, an ema etc etc. that's before I even get started on adding in/taking away a few Conditions. It can easily run into 100's of back tests.
Do you more experienced guyss do this all manually? Or is there some platfrom to help?
Right now I'm flitting between fxDreema, my folder system and excel, getting abit unwieldy
-
I use the strategy tester in MT5 for all my data - it gets the data from the broker where I also run the EAs. Data is very important - I've found optimizing from broker to broker is VERY different and should be avoided. If you test and tune on a specific brokers data, you should also run the EA there.
-
@martymoon120 MT5 has a strategy tester where you can optimize parameters. Sounds like exactly what you need to do. Seems like you just need to optimize on how many pips to place the SL and where to start a trailing stop. I see 4 variables there that need to be optimized on... RR, trailing stop start, size, step. You can easily do this in MT5
-
I don't know what type of market you are operating in, but I will give some examples of Forex as an example of some observations and opinions:
Opinion ---> When you obtain similar results (similar and not precisely the same, closer both in capital curvature and in values and quantity of operations) in OHLC, EACH TICK and EACH REAL TICK, then you can be sure that your next step is to put it on a REAL account for at least 1 week, and the following week you compare whether the backtest entries are similar to those on the real account. After carrying out these tests, if the information is close then you have an EXCELLENT EA. Other than that, the rest is a waste of time.Observation ---> Most EA's Scalpers (who look for very few TP points or who use a short Trailling give positive results in OHLC and a super-negative result in each real tick, because OHLC as mentioned above by someone else It does not have slippage and a series of delays and particularities that brokers have, so for quick Scalpers, always try to follow the rules in testing (Each Real tick) do not waste your time with OHLC.
OHLC is more used for non-scalper strategies, or Day trade, it is more used for Swing Trade, strategies that can be positioned for days and have a TP and a Fixed SL without Trailling and size in long pips, in these cases it is worth testing in OHLC because Spread or Slippage and all the brokers' nonsense will not interfere.
Finally, always keep in mind that before developing your EA, think about the entire context in which it should work and even think and study the characteristics of brokers, for example, Scalpers work very well at ICMARKETS, TICKMIL, EXNESS , but on FBS, ROBOFOREX or FTMO proprietary tables Scalpers catch high slippages, so there is a need to know very well what you are planning and (where to plan to run).
Study about Market Makers, Plugin, Slippage, Hyperactivity, Physical distance from liquidity provider to broker and Spread and with this, try to find out which factors most harm your strategy and avoid brokers that make it easier for these factors to hinder you.I have helped in something.
-
@cmuphyfx @OntradingX really helpful both of you, learnt alot, many thx.
-
Hello friends, when you pay, how long does it take for your purchase plan to be activated?
-
@OntradingX is correct about brokers. A patricular EA of mine could be classified as a scalping strategy and works very well on ICMarkets data, but when I take the EA and backtest on FTMO data feeds, it falls apart. My solution to this is to use a trade copier so I am not crushed by the difference in market conditions.
