I've implemented a C# library to download historical tick data from the Dukas data feed URLs provided for the purpose.

From information gleaned around the internet, I was able to determine the binary data (after decompressing from LZMA, and switching endianness) was in the following format per record:

int1 is milliseconds within the hour
int2 is Ask * 10000
int3 is Bid * 10000
float1 is Ask Volume
float2 is Bid Volume

I've been using TickStory to cross check my results, and noticed something strange. The millisecond values match perfectly, but for all JPY crosses and Gold (might be others, only checked these so far), the Bid/Ask prices are an order of magnitude too low when using a divisor of 10,000, and all other pairs are an order of magnitude too high. When corrected manually, they match perfectly to the data from TickStory.

For the time being I've coded in a manual check to use 100,000 or 1,000 as the divisor for the Bid/Ask prices, depending on the pair I'm getting data for, but I'm not particularly happy with the "bodge".

Can someone explain to me why the data is like this? Or if there is any official documentation I can refer to which will clear it up? I know if I'm missing the "official" reason for the problem, my code is possibly going to stop working correctly at some point in the future. I've searched several times now on Google, and have drawn a blank.

Thanks in advance

I am not sure if I follow your question 100% but, tick data's are often estimated/rounded, and not only does many brokers use different tick data pools, but even on true exchanges you will get different tick data feeds as broker provided feeds often cut corners.

Thanks for the reply, but it wasn't this specifically.

Some of the instruments in Dukas tick data feed have prices that are too high by a factor of 10. Some of them have a price that is too low by a factor of 10. Some of them are correct.

There seems to be a weird logic to it - JPY pairs, stock indices etc all are off by the same amount.

Same goes for the DAX. 120.000 Points? Not so much -
The bid/ask Volume seems to be guessed.
Also after DST started here in Germany the time was shifted by -1 hour for a couple of days.
For some days, the data ranges from 7am to 9pm, for some days from 6am to 8pm and for some from 0am to 12pm (Appears to be random);
All in all I wouldnt rely to much on their historical data to be honest

You say you wrote something in C# to Download the historical Data? Which Library did you use? Mind telling me how you did it or even sharing the Code? I'm currently implementing a Crawler in Java but Java is kinda unsuited for the job.

Regards,
Claas M.