I am a computer programmer and so I am able to number crunch a lot of data and now that I can use API’s I can find a lot of accurate data very easily. I ran some stats on 68 currency pairs and this is what I found. Most of it we already know as I posted similar information a while ago…
Based on the Daily timeframe although it does also hold true for other timeframes.
Statistically the chances of the next candles direction being up or down is a 50/50 split. This is also the case no matter what the previous (up to) 4 candles did. So if we have 4 up candles in a row then it is still 50/50 that the next candle will be up.
With automated systems each currency pair gives different results and the best system for each pair is different.
The best system for each pair changes over time in a random way meaning that it is not possible to predict when it changes or how it changes. The only way to know that it is changing is to watch for the profits to start going down. Once profits start falling there is no way to predict which system will start to work better.
A backtest of a completely random buying/selling algorithm produced both profits and losses and behaved just like any other backtested system. Without labelling it you would not be able to pick its results out from the rest of the systems. Having said that a random system does not win or lose as much as a specified automated system, it hangs around in the middle of the pack not producing spectacular gains or losses.
My backtesting went back 1000 candles and took over 12 hours to run on a 12 core machine. I may try a run with 5000 candles which is the maximum I can obtain from the API but I suspect the results will be the same.
I used Moving Averages, Bollinger Bands, RSI, MACD and even candlestick patterns in various combinations for the various automated systems.
My conclusion is this, the charts are random, they have trends and channels and everything else we have read about but when you get down to it they are still random in nature. No system of any kind works 100% in all conditions due to the random nature of the charts. Systems of any kind (not just automated) will work and not work on various pairs in various timeframes and various market conditions. Certain systems will give you an advantage for a while until they do not anymore. The best way to handle the markets is to be able to take the losses while you find the system that works for a while then get enough gains to cover the losses when the system stops working and you have to look for another system again.
One more interesting thing. I wrote a neural net to try and figure out the charts using the previous 5 candlesticks as inputs and attempting to predict the next candlesticks closing price. What was interesting about it was that the best fit it could come up with involved taking the closing price of the last candlestick and just outputting something close to that. It produced a graph that was a lot like a moving average very heavily weighted to the latest candlestick and one day behind the actual data.
Your statistics are no doubt accurate but there is a disconnect between the criteria for your imaginary trades and those that traders would use to run actual trades.
For example, a trend-following strategy on the D1 time-frame would only be initiated when price is actually trending. Are you able to run a simple test using such a set-up? I suspect the results will be on the positive side.
A strategy that uses moving averages for trend spotting and then only buying/selling when the trend is moving in the right way produces both wins and losses based on the combination of moving averages used and the currency pair. Sticking with one currency pair and using a strategy like that works for a while but then it fades and new parameters for the moving averages has to be found. From what I have seen no strategy works for too long before it needs to be replaced or changed even if it is the same basic idea with different values for the indicators.
The data suggests that systems give you an advantage for trading for a while but they need to be constantly assessed because no system lasts for very long on any currency pair before it no longer gives you an advantage. The trick is to find the new system/parameters that work before you lose too much money and before it changes again.
Thank you for this post. It supports everything I have learned over 10 years about price action, and none of the content has caused me to question what I have spent a lot of time building. I am thankful for that.
This is the takeaway from your analysis:
What was interesting about it was that the best fit it could come up with involved taking the closing price of the last candlestick and just outputting something close to that. I use previous closing price as baseline, as I am sure many others do.
I have just written a Forex bot and what my analysis has shown me is that the strategy does not matter as much as having the bot adapt to market conditions. It now looks at the past 24 hours worth of trading data and sees which currencies have performed well with the chosen strategy and which ones have not. It prioritises them according to their profitability so if a currency is having a bad 24 hours with regards to the strategy it is not likely to trade it. I may update it so that it reduces the trading possibility down to zero, I will see what happens.
thanks for the reply i was nt meant to sound flippant
i just found this on forex factory
iscussions about the randomness or predictability of the financial markets are mostly useless. it has more to do with one’s “belief”. believers of the opposing poles may produce their “proofs” but they are usually subjective and can not be scientifically proven.
one can categorize financial markets as of a complex adaptive system. these systems have the following main attributes:
they exist in leaderless, dynamic environment; their participants interact with each other by following the simple rules and create self organized patterns.
just like a flock of starlings.
my thought how to be profitable in this game is understanding and following the basic rules, having statistical edge and managing the trades accordingly.
The fact that markets has randomness is the reason they exist.Though certain traits or patterns reoccur out of the randomness
@tommor - Pick any strategy, back test it against several currency pairs (I used 68) and you will find that some of the back tests come out with a profit and some come out with a loss for the same strategy.
[quote=“igillman, post:1, topic:675546”]
No system of any kind works 100% in all conditions due to the random nature of the charts.
hi,
can you explain what you mean please
Pick a system, any system. Back test it on all pairs you can find and see if they all show a profit. No system will show a profit with all pairs because the charts are random enough that what works for one does not work for all. Also, pick a system, any system and back test it using last years candles on all pairs you can find. Now back test it using the year previous to that one, then the year prior to that etc… You will find that the same system on the same pair produces a profit some years and a loss on other years.
so your conclusions are no systems work “100 %” in all condition s due to the random nature of the charts.Why would you expect them to work 100 %. Regarding the above paragraph again you say not all pairs shows profits or the annual p/l some years positive others negative {traders are known to adapt their strategies over time} ect .Ok so you have an argument , if you believe some retail traders have a positive P/L over say a 5 year period, where is their edge?
you not shown any example s to back your argument ,for example ,how do you determine your first entry, you queried tommor system he has a criteria before enters a trade
I think he has a valid argument though. That no system works all the time. Isn’t that why we have different strategies for breakouts, reversals, ranges & trends? To me, it just says that there’s no holy grail, which is in line with the stuff we read anyway.
Personally I don’t think it’s completely random but I can’t prove it. As long as seasons affect commodities (agriculturals) and demand for goods (oil in winter) there is cyclical behavior that may be outside the scope of this kind of analysis.
Not to mention bank manipulation. It is not conspiracy theory. All the largest banks have been fined multiple times in the USA for interest rate manipulation, commodities manipulation, and other nefarious activities. Yet they have the full support of western governments as they conspire to rob working people blind. That’s not going to continue in future, and the sooner the masses adopt cryptography as the way to go, the more quickly the manipulation will recede.
