Yeah thats what I used to have was 10 minutes but like you said if you trade higher time frames the 10 minute delay is not a big deal. I wander what pete has to say about futures and volume.
I was thinking of going with Interactive Brokers, but since Saxo is based in Denmark, just two hours from where I live, and I find their software and terms good I decided on opening an account with them.
One can always change if need be…
Where can i find the symbol name for futures of forex pairs?
I have real account with think or swim and i would love to put my eyes on real volume data to trade forex.
go to the CME website
They are the same. On the think or swim site there is a slight difference is the symbol. You will have to hit up customer support as its been at least a year since I opened there plattform.
So I have been pretty busy but I thought I would take a look at something thats used very frequently by people on the forum, TICK VOLUME. I want to compare Tick volume vs Real Trade volume. I am using 6e euro futures. Why you may ask, because it is the only related instrument with real volume and tick volume. Currently the difference between spot forex and the futures price is 2 pips. I think that is a small enough margin of error to make this a valid example, also there is arb between the 2 so they always have a high correlation minus whatever the spread is, in this case 2 pips.
So basically I used data from 2007 until today 5/15/2013 on 5 minute bars. All data is built up from tick data.
When I compare the correlation coefficient (pearson) and the R^2 value of tick volume vs real volume. You have a pretty significant problem.
Correlation Coefficient = 0.731887
R^2 = 0.535659
Basically your tick volume vs real volume (again i have to use futures as parallel since no one has total volume data on spot but the discrepancy between prices is 2 pips). You only have a medium strength correlation at .73 you would need at least >.8 to be considered a high correlation.
R^2 with a 1 meaning perfect alignment of models and 0 meaning none. we are basically in the middle. Where it sometimes explains it, sometimes doesn’t. IMHO that means coin toss. Again we would like to see high values here as well.
I know this isn’t exactly a quantified edge so to speak. But its just an elimination of a useless tool in spot. Tick volume is just not a reliable indicator. There just is no real justification to use a tool that basically only represents real volume 53% of the time and is only medium strength correlated. Thats like saying I am going to drive a car where the speedometer only works 53% of the time, and most of the time when you accelerate the speedo moves higher. But sometimes when you accelerate the speedo moves lower.
So my takeaway, is be wary of this tool. It may be helpful sometimes, maybe more often than not. But its just not as firm a tool as many here believe.
Very interesting, thanks for sharing. A correlation score of 0.73 shows that there is a relationship, albeit not a [B]very[/B] strong one. My background is statistical and modelling, and whenever I’ve hit a score like this, you start tuning and manipulating and in some cases, you’ll get that score up. For example, repeat on a 1 hour timeframe, remove mondays … I’m willing to bet that the score will move up, whether the R squared does … not sure.
The problem with tuning and manipulating is that you can make the numbers tell you anything. Sure you can get higher correlations and R^2 but thats not the purpose. Its have an unbiased mind that allows the pure number without manipulation give you the answer. I forget who said it but “if you torture the numbers enough they will tell you anything”
Totally agree but if you are looking for a statistical edge or trying to build a predictive model, that’s where manipulation comes to play. If you want an unbiased view, then ditch the manipulation. If you take out mondays, all you are saying is “is tick volume correlated to futures volume Tuesdays to Fridays”.
Also, doesn’t tick volume data vary across brokers?
Right, I totally agree with you there. If i was using tick volume to generate a strategy or an executable model there would definitely be a lot of work but it may be possible. But as with everything the more rules/filters you have the less degrees of freedom so its a double edged sword.
Regarding the tick volume because futures is a centralized market, tick volume is the same for everyone as well as real volume. in spot the difference between brokers would exacerbate the tick volume vs real volume correlation problem because each broker will have different ticks. The numbers I gave are best case scenario. In reality actually using retail spot tick volume is probably a much worse option than what is posted above.
Great thread there! but i have to sleep now. its 5am in my time GMT+8. Hope you will put more findings up!
Jack.C
Ok so far what I have learn’t from your findings is:
- Buy/sell on pullbacks
- Proper execution of R:R 3:1 or more is needed.
Unless you have a high accuracy rate 80/90% 1:1 is fine.
What other statistical data have you got laying around that I could use.
Great thread hope you continue it…
All trading is based on volatility to generate price movement. But what happens when we experience extreme volatility? This can caused by a lot of different scenarios, but usually fundamental. Our recent economic history is full of them and we are still in a current state of turmoil. So happens after these events occur? Is there any edge or interesting phenomena to take advantage of? Thats what I am here to find out.
Testing Procedure: I am using a professional aggregate data feed on EUR/USD with over 15 liquidity providers. I have bid/ask on a 1 minute bar granularity starting in 1999 to present day. I am using daily bars created from these individual minutes so thats 2880 data points per daily bar (1440 minutes * 2 for bid ask). A price shock is considered 200% of the trailing 60 day Average True Range. This has only occurred 60 times since the start of my data set. Its a fairly rare occurrence. Although this is a relative measure so if the volatility for the preceding 60 days is low it won’t take much to trigger, but its better then absolute value threshold because of the difference of trade at say .8350 or 1.5900.
OK lets explain the charts. The Shock Value is the absolute value of the price shock. Then we have the following 10 days in absolute change. So each days change is recorded individually not cumulatively. I have here charted the Average follow through of all price shocks divided by Shocks that were Positive price changes and Negative Price changes. I also have separated out the largest individual shocks.
Analysis of Average Price Follow Through:
Here we can see the average shock value is much smaller than I expected at 150. However we have to take into account times where price was in the .8400 range where that would be a huge % change. Again this is why the relative measure is important.
Lets take a look at the positive shocks only. we can see that the following days are all negative. What that means is we are seeing a retracement, over a series of 10 days. This is a huge indication of directional bias. Basically if you shorted after a Positive price shock and held for 10 days you could net ~457 pips. I mean when the picture is this clear its really easy.
Negative Price Shocks are not as clear but still indicative. We can’t always have beautiful results 
Again we see a fundamental pattern here, the retracement and having positive values. Basically we go positive for the first 4 days (days +1 to +4). Days +5 to +9 are basically a wash with +10 being a nice cap off. If you longed the day after a Negative Price shock and held for 10 days you would net on average +153 pips. Interestingly it is close to the value of the original price shock itself.
Largest Positive and Negative Individual Shocks
You can see that the Largest Positive Shock followed the tell tale signs very well. But the largest negative shock continued lower, which would be indicative of a fundamental shift. Remember guys this is just a tendency, not a law.
Bottom Line: Usually price retraces after a price shock. On the order of 150+ pips. These events don’t happen that frequently, but when they do they may present a significant opportunity.
[Disclaimer: this is for educational purposes only, this is in no way a trade recommendation or solicitation to make any financial transactions. Trading Forex carries substantial risk]
Thank you for all the hard work you put into this.
Have you ever been interested in ocean waves ?
The classic waves refer to a nice linear model which helps to predict much such as duration, strength, …
However, there are rogue waves (non linear)
"Rogue waves (also known as freak waves, monster waves, killer waves, extreme waves, and abnormal waves) are relatively large and spontaneous ocean surface waves that occur far out at sea, and are a threat even to large ships and ocean liners "
I think that extreme volatility can be compared to rogue waves and analyzed through a non linear approach.
Ok so after a major impulse wave on a daily time frame you have a chance to grab at least 450 pips on a correction?
Interesting research Meihua!
Basically, you may find a nice profit through the correction but it is difficult to determine the start of the “after” due to the extreme nature of the move.
Take a look at usdchf when it went as low as the 0.70 area.
- some people did not enter because they were afraid to sell a low
- some people did not enter because they were waiting for the confirmation of a valid bottom
- some people tried to buy at some levels but they were wrong
- triggered stop losses have contributed to accelerate the move
- people who did great were those well informed and those who entered without wondering about anything
So, if you compare extreme volatility to rogue waves, the main difficulty resides in detecting early the non linear feature of such a move
There are probably many lessons to understand from such moves. A key point is to be flexible enough to overcome fear and react instantly. This the reason why I learnt to stop thinking and make instant decisions when this kind of rare opportunity occurs.
I am not really familiar with analysis of rogue waves, though it is interesting in concept. I was not analyzing using a linear model, but if i was it would just be a huge outlier. But i guess that’s exactly what a rogue wave is?
if anyone was wondering the definition i used for what a price shock was fairly straight forward and you have plenty of time to trade it. it was past a day that was 200% of the last 60 days ATR. the following days open to close measure is day 1. so you have at minimum a few hours, if not the entire next day to get on board. This was all outlined in the original post, i just wanted to make it clear.