Systematic trading | Page 63 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Trade another day through Systematic Trading 203 replies

Nonparametric Association Measures in Systematic Trading 57 replies

Anyone use Systematic from Alpari? 1 reply

A Systematic Approach to Markets 68 replies

Systematic Trading vs. Discretionary Trading 14 replies

Attachments: Systematic trading
Exit Attachments

Systematic trading

  • #1,241
  • Jan 26, 2011 2:34pm Jan 26, 2011 2:34pm
Quoting Adal
They still have reliability issues.

8 disconnects of the Quote API during todays FOMC. After almost a day with no problems.
MBTrading is and has always been a retail broker, so you probably have to eat their tech issues if you wanna play with them.

How is the data quality of the quote feed? Last time I checked (a few years ago) spreads where all over the place with extended periods of inverted spreads (sure sign that something is wrong).
  • #1,242
  • Jan 26, 2011 3:14pm Jan 26, 2011 3:14pm
Quoting BlackMage
MBTrading is and has always been a retail broker, so you probably have to eat their tech issues if you wanna play with them.

How is the data quality of the quote feed? Last time I checked (a few years ago) spreads where all over the place with extended periods of inverted spreads (sure sign that something is wrong).
I'm just storing the data as received, not converting, displaying or analyzing it.

One thing I noticed yesterday while testing the code was a Level 2 update on USD/JPY with a time 3 hours back (so the timestamp on the update was current time minus around 3 hours). All the other timestamps that I saw were basically current time. I didn't investigate it further to see if it's a bug like it seems, or if maybe it is some sort of correct edge case. The API is also very lightly documented.

Here's how the raw data looks like by the way:
Inserted Code
1|1003=EUR/USD;2006=1000000;2014=16:08:39;2015=01/26/2011
2|2024=O;2033=F;1003=EUR/USD;2016=TDFX13694100;2017=20011;2018=1.36941;2019=1000000;2020=16:08:37
2|2024=O;2033=F;1003=EUR/USD;2016=TDFX13694300;2017=20011;2018=1.36943;2019=3000000;2020=16:08:37
2|2024=O;2033=F;1003=EUR/USD;2016=TDFX13692500;2017=20010;2018=1.36925;2019=2000000;2020=16:08:39
2|2024=O;2033=F;1003=EUR/USD;2016=TDFX13692300;2017=20010;2018=1.36923;2019=7000000;2020=16:08:37
2|2033=F;1003=EUR/USD;2016=TDFX13685000;2017=20010;2018=0;2019=0
3|2022=FX;2082=330;2083=0;1003=EUR/USD;2084=2;2039=30030;2004=1.36941;2006=1000000;2013=2;2014=16:08:39;2015=01/26/2011
1|1003=CHF/JPY;2004=87.245;2006=2000000;2014=16:08:39;2015=01/26/2011
3|2022=FX;2082=330;2083=0;1003=CHF/JPY;2084=2;2039=30030;2004=87.245;2006=2000000;2013=2;2014=16:08:39;2015=01/26/2011
1|1003=CAD/JPY;2006=3000000;2014=16:08:39;2015=01/26/2011
3|2022=FX;2082=330;2083=0;1003=CAD/JPY;2084=2;2039=30030;2004=82.667;2006=3000000;2013=2;2014=16:08:39;2015=01/26/2011
1|1003=AUD/JPY;2002=82.115;2003=82.115;2005=1000000;2056=82.115;2014=16:08:39;2015=01/26/2011
2|2024=O;2033=F;1003=AUD/JPY;2016=TDFX821470000;2017=20011;2018=82.147;2019=2000000;2020=16:08:39
2|2033=F;1003=AUD/JPY;2016=TDFX821530000;2017=20011;2018=0;2019=0
2|2033=F;1003=AUD/JPY;2016=TDFX821160000;2017=20010;2018=0;2019=0
2|2024=O;2033=F;1003=AUD/JPY;2016=TDFX821090000;2017=20010;2018=82.109;2019=8000000;2020=16:08:38
1 - Level 1 update
2 - Level 2 update
3 - Time & Sales (whatever that means for Forex at MBT)
  • #1,243
  • Jan 26, 2011 3:18pm Jan 26, 2011 3:18pm
Quoting Adal
I never tried their FIX API, but due to theirs announcement I finally wrote last night a little logger of their (demo I think) ticks using their MBT Quote API that I had requested an account for last year.

So far it logged 1.6 GB in 16 hours for the Level 2 data of the most important 13 pairs (the maximum allowed) and Level 1 data for all pairs. Fortunately a simple .zip compression will bring the size down to around 150 MB (and 100 MB with 7zip/WinRAR). Still, we are talking about 30 GB/per year compressed (and 300 GB uncompressed to process)...
I have been harvesting MBT tick data already for several years...
see http://www.forexfactory.com/showthread.php?t=72280
I have been also working with the data so i know it very well.
I'm not insane; my mother had me tested.
  • #1,244
  • Jan 26, 2011 3:24pm Jan 26, 2011 3:24pm
Quoting Adal
They still have reliability issues.

8 disconnects of the Quote API during todays FOMC. After almost a day with no problems.
That is not unusual... they quote servers leave a lot to be desired at times... especially when there is a lot of action in the market that generates lots of quote updates the server just can't keep up with delivering the all, it starts to build a backlog (which can be seen by comparing the timestamps in the quotes with the time of arrival) and once the backlog grows to a certain size the servers drops the connection like a hot potato.
Fortunately you can use several logins to connect to multiple quote server and have some redundancy.
I'm not insane; my mother had me tested.
  • #1,245
  • Jan 26, 2011 3:31pm Jan 26, 2011 3:31pm
Quoting BlackMage
MBTrading is and has always been a retail broker, so you probably have to eat their tech issues if you wanna play with them.

How is the data quality of the quote feed? Last time I checked (a few years ago) spreads where all over the place with extended periods of inverted spreads (sure sign that something is wrong).
The quality is not that great but acceptable, seems like most of the time MBT just passes the quotes (including timestamps) as they receive them from their liquidity providers. They probably have have some algo to filter out some totally bogus quotes but still some of them seem to slip through.
Anyway, i actually prefer it less filtered because i can do my own filtering and nod need to relay on the MBT filtering algo.
I'm not insane; my mother had me tested.
  • #1,246
  • Jan 26, 2011 3:38pm Jan 26, 2011 3:38pm
Quoting Adal
One thing I noticed yesterday while testing the code was a Level 2 update on USD/JPY with a time 3 hours back (so the timestamp on the update was current time minus around 3 hours). All the other timestamps that I saw were basically current time. I didn't investigate it further to see if it's a bug like it seems, or if maybe it is some sort of correct edge case. The API is also very lightly documented.
That is also nothing unusual. As i mentioned in my previous post, if you look at the stream of level 2 quotes, you can see that the timestamps are not completely ordered... my guess is that those timestamps come from liquidity providers so the depend on the clock of those liquidity provider... and those are not quite in sync.
On the other side the level 1 and T&S records are always timestamped correctly so those timestamps seem to come from MBT directly which is logical because those records contain data that is aggregated by MBT.
I'm not insane; my mother had me tested.
  • #1,247
  • Jan 27, 2011 2:29am Jan 27, 2011 2:29am
This might interest people here

http://paul.kedrosky.com/archives/20...simons_sp.html
  • #1,248
  • Jan 27, 2011 7:04am Jan 27, 2011 7:04am
Quoting RoBiK
That is also nothing unusual. As i mentioned in my previous post, if you look at the stream of level 2 quotes, you can see that the timestamps are not completely ordered... my guess is that those timestamps come from liquidity providers so the depend on the clock of those liquidity provider... and those are not quite in sync.
On the other side the level 1 and T&S records are always timestamped correctly so those timestamps seem to come from MBT directly which is logical because those records contain data that is aggregated by MBT.
This is saying something for MBT. Surely the most easy thing is to always supply/store data in GMT. If another timezone is needed simply convert on receiving/retrieving. Bet you their other providers supply data in gmt.
  • #1,249
  • Jan 27, 2011 7:50am Jan 27, 2011 7:50am
Quoting mikkom
This might interest people here
http://paul.kedrosky.com/archives/20...simons_sp.html
That is actually a really cool lecture. I'm all ready beginning to miss University...and Lie Algebras

It's definitely worth viewing the entire lecture - Simons is surprisingly humoristic.
  • #1,250
  • Edited 2:04am Jan 28, 2011 1:49am | Edited 2:04am
I have added some futures contracts to my framework and I have a small problem.. I have quite a lot of 1 min data but I don't have any bid-ask spread data.

So what would be a good source for bid-ask spread data for futures (and why not FX too) so my simulations would be at least a bit more realistic?

(edit: next expiring/most liquid contract spread would be ok)
  • #1,251
  • Jan 28, 2011 5:34pm Jan 28, 2011 5:34pm
Quoting mikkom
This might interest people here

http://paul.kedrosky.com/archives/20...simons_sp.html
Thanks for the lead mikkom, great lecture, very enjoyable.

Cheers,

Jonas
  • #1,252
  • Jan 29, 2011 2:48pm Jan 29, 2011 2:48pm
Mikkom , I like your thread a lot and appreciate your contributions. Thank you for doing all this.
  • #1,253
  • Jan 31, 2011 1:04pm Jan 31, 2011 1:04pm
http://howtohft.blogspot.com/2011/01...nd-part-1.html
  • #1,254
  • Jan 31, 2011 3:18pm Jan 31, 2011 3:18pm
There's quite a big difference between different data feeds for futures. Zen-fire seems to be one of the most liquid feeds I've seen out there.
Kinetick is another firm who claims to have a very accurate feed, never seen their feed though.

Quoting mikkom
I have added some futures contracts to my framework and I have a small problem.. I have quite a lot of 1 min data but I don't have any bid-ask spread data.

So what would be a good source for bid-ask spread data for futures (and why not FX too) so my simulations would be at least a bit more realistic?

(edit: next expiring/most liquid contract spread would be ok)
  • #1,255
  • Feb 1, 2011 3:07pm Feb 1, 2011 3:07pm
Just wandering what time zones the automated traders use on this thread. Has anyone tested the result of their system using different timezones?
  • #1,256
  • Feb 1, 2011 3:16pm Feb 1, 2011 3:16pm
Quoting foracy
Just wandering what time zones the automated traders use on this thread. Has anyone tested the result of their system using different timezones?
I do not use any timezones, i do not use any time at all... just ticks.
I'm not insane; my mother had me tested.
  • #1,257
  • Feb 1, 2011 3:36pm Feb 1, 2011 3:36pm
Quoting foracy
Just wandering what time zones the automated traders use on this thread. Has anyone tested the result of their system using different timezones?
If one uses times in their calculations, time zone should not matter as long as it's consistent.
  • #1,258
  • Feb 1, 2011 4:02pm Feb 1, 2011 4:02pm
Quoting foracy
Just wandering what time zones the automated traders use on this thread. Has anyone tested the result of their system using different timezones?
Timezones can be important in live trading, for example, if you start trading at the London open and finish at the New York close you will be dealing with differing timezones with daylight savings changes which happen at different times. For backtesting, obviously it doesn't matter as long as the data is consistent. For the second half of the question, I assume you mean to ask will a system behave differently across different sessions? Well, of course it will, but this really depends on what you're trying to achieve.
The breaking of a wave cannot explain the whole sea.
  • #1,259
  • Feb 1, 2011 4:12pm Feb 1, 2011 4:12pm
I believe foracy might be thinking about the fact that candles will be different based on how you group the ticks into them. A candle can change from a pinbar to doji on a daily chart depending on what timezone you use since the candle will start at different times.

Since I've stopped using candles and moved to a "line chart", the charts I use look the same in all timezones - the only difference is that the time labels are shifted back or forward.

But I could get different results if I use an indicator which samples the ticks at different points - for example a 50 period "4 hour" EMA.
  • #1,260
  • Feb 1, 2011 4:46pm Feb 1, 2011 4:46pm
Quoting Adal
I believe foracy might be thinking about the fact that candles will be different based on how you group the ticks into them. A candle can change from a pinbar to doji on a daily chart depending on what timezone you use since the candle will start at different times.

Since I've stopped using candles and moved to a "line chart", the charts I use look the same in all timezones - the only difference is that the time labels are shifted back or forward.

But I could get different results if I use an indicator which samples the ticks at different...
I think you're right, that is probably what he ment. As for the line charts, they generally use close values for their data points, and consequently certain timeframes also appear differently depending on the timezone (e.g. 4h, daily). Same goes for the indicators, which you already mentioned.
The nail that sticks out gets hammered back in
0 traders viewing now