I have written several posts in the past regarding the issue of broker dependency in Forex trading and why this general lack of a central exchange makes the evaluation of automated trading strategies tremendously difficult as we are bound to get different statistical characteristics for each different broker we use. Through the past 3 years of trading amongst various brokers I have noted that these differences are not small or negligible, even for strategies that trade on time frames as high as the daily reason why understanding the factors that generate these differences is so important. On today’s post I am going to tell you which are the main aspects that cause results amongst different brokers to be so different and also give you some hints on how they might be eliminated or at least reduced.
People often believe that the differences caused by using brokers A and B in Forex trading is only relevant if the strategy used is vulnerable to things such as spread widening which is the most common way in which people become aware of the fact that not all brokers are made equal. However upon a closer look and long term evaluation of different strategies we can see that these differences are quite important for ALL strategies and not only those that trade on the lower time frames or are affected by spread widening (although the problems are obviously tremendously magnified for such strategies). What are the things that in practice cause brokers A and B to be so different ? Here are the most influential factors I have noted regarding broker dependency:
1. Broker market opening time. When your broker opens up for the week makes a long term difference regarding the results you’re bound to get in the longer term. The fact is that having one hour more or less in the beginning of the week can make a dramatic differences since any price-action or indicator-calculated value will be affected by that extra or missing bar. Certainly systems that calculate some sort of gap measurement or breakout range over this period will be the most affected but even if your system trades a simple 200 period MA hourly close in the long term this additional bar will cause a “shift” which will influence signals.
2. Existance of DST and changing times. This issue is related to the above because when brokers change DST they do not simply shift the hour of their candles but they also add or eliminate a candle from the total number of candles during the week. If a broker changes its DST time it usually means that it will take away or put in an hour in Friday and remove or add an hour to Sunday trading. This means that a broker that has a DST change will have a shifting of candles done throughout the year which may make its results different from what you expect. The way in which DST changes are done will inevitably cause variability between two brokers with different values as time moves on as these changes introduce distortions EVEN if you correct time (since candles are added or eliminated when the changes are carried out).
3. Day ending time. One of the things that affects results the most amongst different brokers (especially if some measurement of time is used) is the precise time when the day ends. The problem here is generated because the overall shape of the daily charts – and the indicators calculated within them – will change according to where your broker decides a day should end. If you add to this the fact that some brokers even introduce a full additional Sunday candle the problem becomes even more complicated as the overall distortion in indicator values on the daily charts becomes even more prominent. Although most brokers end their day around the same hour just a one hour difference between two brokers on the time where their days close can cause significant differences in results if any measurement concerning the daily candles is used.
4. Spread variations. Brokers have different spreads and this is no secret to most people. However when you measure the way in which spreads vary amongst brokers you will see that the variation is very randomly distributed and brokers tend to widen their spread at different times and by very different amounts. This is another reason why obtaining reliable statistical characteristics for scalpers is so difficult, even if you have tick data with Bid/Ask spread values for one broker the next one is most likely going to be completely different. Such is the different amongst the Bid/Ask spread variation between brokers that the lower time frame data is almost never the same (very rarely do candles match precisely) implying that large differences in short term time frames and systems with profit targets close to the spread are to be expected.
5. Liquidity. This affects the filling of orders and therefore it affects the exit and entry points of your strategies through slippage or in the case of none market makers through the availability of someone filling your market orders at the price levels you desire. Often liquidity issues will cause variations in the long term as just a small 1-2 pips variation in just 1% of entries can cause at least a few trades to have opposite outcomes from a long term statistical perspective (that slight “off” entry leads to the fact that you will not be able to exit at the intended levels on a number of trades). Liquidity therefore also plays a significant role because it affects the exact entry and exit points of your strategies.
As the above shows it becomes obvious that having the same results amongst all the different brokers will be quite impossible but it is also true that certain things can be done to reduce the effect of all this dependency. Several options exist – like using a feed as “central” and distributing it to other brokers – but such approaches suffer from additional dependency-generating complications such as “lag” and price differences between brokers. Perhaps the best thing in the end might be to stick with a broker for which you have long term data which you can trust and then use only this broker and avoid others. Since you know the precise statistical characteristics for this broker then dependency issues are not something you will have to worry a lot about. A broker like Dukascopy which offers this data and the possibility to later migrate into a FIX protocol Currenex link might be an extremely good choice for the long term development of strategies. However several tactics to reduce dependency amongst regular brokers such as dithering, fuzzy logic and daily chart corrections through re-building of charts based on NTP time stamps might be ways of making the code execution more reproducible at the expense of system complexity.
If you would like to learn more about my work in automated trading and how you too can learn to develop and evaluate your own automated trading systems please consider joining Asirikuy.com, a website filled with educational videos, trading systems, development and a sound, honest and transparent approach towards automated trading in general . I hope you enjoyed this article ! :o)