The Low Latency Advantage in
Online Trading
by Barry Bahrami of
Commercial
Network Services
Latency is arguably the most
important and often disregarded factor for the online trader. It
is often neglected because it sounds like a complicated computer
term, but understanding what it is can quickly and substantially
increase the trader’s earnings. Computer networks
communicate by sending packets of data through the Internet.
Latency is the time required for data to travel from the source
to destination network. It is measured in milliseconds, or
1/1000th of a second.
It is a public misconception that everyone enjoys the same
Internet. Data traveling over the Internet can be compared to
delivery vans on the road carrying packages to their recipients.
The recipients of the packages are all over the world, and so
some delivery vans must travel a lot farther than the others in
order to reach their recipients. Some experience delays due to
detours and traffic jams. Others experience lost or damaged
packages and need to restart the delivery again at the expense
of time. The delivery van that travels only a city block or two
will get to the destination first and with little chance of
damage to the packages. This is because the other delivery vans
travel hundreds or even thousands of miles farther to reach the
destination, which increases the time required to deliver the
package and likelihood of a problem or detour along the way.
With online trading, latency is even more crucial because data
must travel round trip to make a single transaction. That is, a
price quote must travel from the broker’s server to the trader’s
VPS to be processed and then the data must travel the same
distance back to the broker’s server to enter or adjust an
order. Since the market waits for nobody and will continue to
move while price quotes are in transit, the trader with the
lower latency VPS will always have a more true and real-time
reading of the market then the trader with a higher latency VPS.
If the market is volatile and latency high then the price quote
may not be “fresh” when it is finally received by the recipient.
The financial advantage of a low latency VPS over any other VPS
can be staggering to the trader. To understand why, we only need
to adjust our previous example to better reflect the real chain
of communication in an online trading scenario. That is, data
must travel from the broker to the VPS and back to the broker.
Imagine two traders; one is
located in an office two city blocks off Wall Street and the
other in a costume shop 880 miles away in Atlanta, Georgia. Both are using the
same NYC broker. For the purposes of this illustration, imagine
the price quotes and orders are sent by taxi from the broker to
the trader and back.
For the first trader in NYC, the data must only travel from the
broker down the street two blocks. That trader then sends the
taxi back another two blocks with instructions for the broker to
open or edit an order.
Although both taxis left the broker at the same time, the second
trader will not receive their same price quote before the first
trader has already sent their taxi back and entered an order
with the broker. This is because the second taxi must travel an
additional 880 miles just to deliver the price quote – the first
½ of the round trip. It must then make the return trip back –
all while the market continues to move during their journey to
the destination. The first trader will have completed many round
trips for every one made by the second trader. During periods of
volatile market activity, the price quote received by the second
trader may not even be current due to the longer
time required to deliver data. Clearly, the trader only two
blocks away from the broker has a tremendous advantage over the
other nearly 900 miles away.
Although data really travels much faster over the Internet then
cars do on the road, the distance data must travel and the time
advantages are virtually the same in real world online trading.
The result is that the trader closer to the broker will receive
price quotes and execute orders significantly sooner and with
fewer errors then any other trader, and always have the most
current representation of true market conditions.
Errors?
Yes, errors in trading. Higher latency
connections are more prone to packet delivery delays and loss.
It is a missed opportunity for the trading terminal every time
the expert advisor receives a late packet or experiences
connection problems with the broker due to the longer path the
data must take through the various networks that make up the
Internet.
The proof is in the results
If you think this is only theory then think
again. The difference can be substantial and will vary with the
strategy employed by the EA. Just how much of a difference can
be demonstrated by reviewing the same exact EA to the same NYC
broker from two different Commercial Network Services data
centers, each located on opposite coasts of the United States.
On April 26, 2009,
Commercial Network Services restarted all demo expert
advisors at US$25,000 to better illustrate the importance of low
latency to the online trader. In the most drastic example, the
Bogie-NN-v8 expert advisor in the CNS NYC data center
enjoyed a 284% larger profit than the
same EA in the CNS SDCA (US west coast) data center to the
same NYC broker. While the EA managed to earn a US$870 profit on
a US$25,000 account after only 9 days from the US west coast
VPS, the same EA with a low latency NYC VPS to the same NYC
broker earned a US$2,475 profit during the same time period –
nearly 3X more profit. The results continue to compound, and can
be viewed in real-time on the CNS
Top Expert Advisors page. The advantage in this example can
be dwarfed by scalping EA’s, such as FapTurbo and Megadroid.
In reviewing the statements side by side, it’s easy to see how
lost packets along the way and the higher latency from the
increased distance can result in less favorable or missed
trades. Every lost or delayed packet is a lost opportunity for
the EA.
After reviewing the MT4 log details, William (Wackena) Boatright,
the author of Bogie NN-v8 and 2nd place winner of the 2007
Automated Trading Championship said:
|
I did not find
reason for difference by reviewing logs. I was
hoping to find login to server issues. If both demo
are running on same server URL, the west coast demo
did not see all of the same trade signals that NYC
had. Only reason I can think of is server connection
difference. I have not seen on my CNS vps, but on my
local PC, MT4 terminal will lose connection to
server without a log entry to ID. I manually login
again and then MT4 terminal reconnects.
Best regards,
William (Wackena)
Bogie Enterprises |
In another example, the advantage of a low
latency VPS for EA's that enter market orders can be realized.
Reviewing the following Artemis Breakout statements side by side
demonstrates price advantages for the lower latency VPS with
market orders. The Artemis Breakout EA was started on April 30,
2009 in identical VPS’s within CNS data centers on opposite
coasts of the United States. After just a few short days, the
lower latency VPS is already securing better prices on
executions over the higher latency VPS. Notice the time of the
two highlighted orders are the same, but the lower latency VPS
secured the better price for the market order. Indeed, the
profits realized from the better entry prices will add up
quickly.
ABO with NYC VPS to a NYC broker:
|
Ticket |
Open Time |
Type |
Size |
Item |
Price |
S / L |
T / P |
Close Time |
Price |
Commission |
Taxes |
Swap |
Profit |
|
7536067 |
2009.05.04
09:43 |
buy |
0.30 |
eurusd |
1.3324 |
1.3224 |
1.3424 |
2009.05.04
13:02 |
1.3254 |
0.00 |
0.00 |
0.00 |
-210.00 |
|
7548618 |
2009.05.04
13:59 |
sell |
0.10 |
eurusd |
1.3251 |
1.3250 |
1.3151 |
2009.05.04
14:18 |
1.3231 |
0.00 |
0.00 |
0.00 |
20.00 |
|
7550432 |
2009.05.04
13:59 |
sell |
0.20 |
eurusd |
1.3251 |
1.3234 |
1.3151 |
2009.05.04
15:28 |
1.3234 |
0.00 |
0.00 |
0.00 |
34.00 |
|
7611494 |
2009.05.05
09:07 |
sell |
0.10 |
eurusd |
1.3378 |
1.3377 |
1.3278 |
2009.05.05
09:22 |
1.3357 |
0.00 |
0.00 |
0.00 |
21.00 |
|
7612336 |
2009.05.05
09:07 |
sell |
0.20 |
eurusd |
1.3378 |
1.3370 |
1.3278 |
2009.05.05
09:41 |
1.3370 |
0.00 |
0.00 |
0.00 |
16.00 |
|
7627067 |
2009.05.05
12:45 |
sell |
0.10 |
eurusd |
1.3379 |
1.3378 |
1.3279 |
2009.05.05
17:23 |
1.3359 |
0.00 |
0.00 |
0.00 |
20.00 |
|
7648898 |
2009.05.05
12:45 |
sell |
0.10 |
eurusd |
1.3379 |
1.3359 |
1.3279 |
2009.05.05
17:50 |
1.3338 |
0.00 |
0.00 |
0.00 |
41.00 |
|
7652389 |
2009.05.05
12:45 |
sell |
0.10 |
eurusd |
1.3379 |
1.3356 |
1.3279 |
2009.05.05
18:01 |
1.3356 |
0.00 |
0.00 |
0.00 |
23.00 |
|
7734699 |
2009.05.06 16:33 |
buy |
0.30 |
eurusd |
1.3349 |
1.3249 |
1.3449 |
2009.05.06 17:06 |
1.3278 |
0.00 |
0.00 |
0.00 |
-213.00 |
|
7794522 |
2009.05.07
09:25 |
sell |
0.30 |
eurusd |
1.3262 |
1.3362 |
1.3162 |
2009.05.07
13:06 |
1.3333 |
0.00 |
0.00 |
0.00 |
-213.00 |
|
7810434 |
2009.05.07
13:06 |
buy |
0.10 |
eurusd |
1.3339 |
1.3340 |
1.3439 |
2009.05.07
14:49 |
1.3359 |
0.00 |
0.00 |
0.00 |
20.00 |
|
7821351 |
2009.05.07
13:06 |
buy |
0.20 |
eurusd |
1.3339 |
1.3347 |
1.3439 |
2009.05.07
14:58 |
1.3347 |
0.00 |
0.00 |
0.00 |
16.00 |
|
7825195 |
2009.05.07
15:22 |
buy |
0.30 |
eurusd |
1.3348 |
1.3248 |
1.3448 |
2009.05.07
15:36 |
1.3267 |
0.00 |
0.00 |
0.00 |
-243.00 |
ABO with west-coast VPS to NYC
broker:
|
Ticket |
Open Time |
Type |
Size |
Item |
Price |
S / L |
T / P |
Close Time |
Price |
Commission |
Taxes |
Swap |
Profit |
|
7536069 |
2009.05.04
09:43 |
buy |
0.30 |
eurusd |
1.3324 |
1.3224 |
1.3424 |
2009.05.04
13:03 |
1.3254 |
0.00 |
0.00 |
0.00 |
-210.00 |
|
7548624 |
2009.05.04
13:59 |
sell |
0.10 |
eurusd |
1.3251 |
1.3250 |
1.3151 |
2009.05.04
14:18 |
1.3231 |
0.00 |
0.00 |
0.00 |
20.00 |
|
7550436 |
2009.05.04
13:59 |
sell |
0.20 |
eurusd |
1.3251 |
1.3234 |
1.3151 |
2009.05.04
15:28 |
1.3234 |
0.00 |
0.00 |
0.00 |
34.00 |
|
7611496 |
2009.05.05
09:07 |
sell |
0.10 |
eurusd |
1.3378 |
1.3377 |
1.3278 |
2009.05.05
09:22 |
1.3357 |
0.00 |
0.00 |
0.00 |
21.00 |
|
7612337 |
2009.05.05
09:07 |
sell |
0.20 |
eurusd |
1.3378 |
1.3370 |
1.3278 |
2009.05.05
09:41 |
1.3370 |
0.00 |
0.00 |
0.00 |
16.00 |
|
7627069 |
2009.05.05
12:45 |
sell |
0.10 |
eurusd |
1.3379 |
1.3378 |
1.3279 |
2009.05.05
17:23 |
1.3359 |
0.00 |
0.00 |
0.00 |
20.00 |
|
7648896 |
2009.05.05
12:45 |
sell |
0.10 |
eurusd |
1.3379 |
1.3359 |
1.3279 |
2009.05.05
17:50 |
1.3338 |
0.00 |
0.00 |
0.00 |
41.00 |
|
7652382 |
2009.05.05
12:45 |
sell |
0.10 |
eurusd |
1.3379 |
1.3356 |
1.3279 |
2009.05.05
18:01 |
1.3356 |
0.00 |
0.00 |
0.00 |
23.00 |
|
7734587 |
2009.05.06 16:33 |
buy |
0.30 |
eurusd |
1.3350 |
1.3250 |
1.3450 |
2009.05.06 17:06 |
1.3278 |
0.00 |
0.00 |
0.00 |
-216.00 |
|
7794538 |
2009.05.07
09:25 |
sell |
0.30 |
eurusd |
1.3264 |
1.3364 |
1.3164 |
2009.05.07
13:06 |
1.3335 |
0.00 |
0.00 |
0.00 |
-213.00 |
|
7810433 |
2009.05.07
13:06 |
buy |
0.10 |
eurusd |
1.3339 |
1.3340 |
1.3439 |
2009.05.07
14:49 |
1.3360 |
0.00 |
0.00 |
0.00 |
21.00 |
|
7821324 |
2009.05.07
13:06 |
buy |
0.20 |
eurusd |
1.3339 |
1.3347 |
1.3439 |
2009.05.07
14:58 |
1.3347 |
0.00 |
0.00 |
0.00 |
16.00 |
|
7825194 |
2009.05.07
15:22 |
buy |
0.30 |
eurusd |
1.3348 |
1.3248 |
1.3448 |
2009.05.07
15:36 |
1.3261 |
0.00 |
0.00 |
0.00 |
-261.00 |
It is also
interesting to note in the ABO trades dated 09:25 and 13:06, the
higher latency VPS appears to have secured the better price.
However, this is really reflecting the movement of the market
during the time required for the signals to reach the
higher latency VPS and make the return trip back
to the broker. In the trade dated 09:25, the market
continued to move up with the trade, and so the higher latency
VPS entered the order after the lower latency VPS.
This demonstrates how the market will continue to move during
the extended time the data must travel round trip to the trading
terminal and broker. In this specific case, it worked out
in favor of the higher latency VPS - but this can not be relied
on for long term performance. Had the market turned
quickly, the lower latency VPS will have been better able to
respond quicker than the higher latency VPS.
Scalping EA’s will realize the largest advantage in low latency
trading.
What’s low latency to one
VPS is not necessarily low latency to the other VPS
It is important to remember that latency is
relative to the distance between the broker’s server and the
VPS. The same east coast VPS can not be expected to perform
nearly as well with a west coast broker, when compared to the
same EA running in a west coast VPS.
How to use latency to your advantage
Most traders already have an account with a
broker before they setup a new VPS. Since the location of the
broker is known, the trader only needs to setup in a VPS that is
located physically closest to the broker. The VPS that is
physically located closer to the broker will have the least
latency to that same broker. After the VPS is setup, the trader
can transparently enjoy all the benefits of the low latency
trading platform and not worry about it again. It is important
to note that some brokers may give away a free VPS with a
trading account, but these VPS’s do not necessarily have the
least latency to the broker.
Commercial Network Services publishes
latency to major brokers from CNS data centers in the
Knowledge Base, allowing subscribers to setup or move their
Trader’s Desktop to the data center closest to their broker.
This document also contains information how to test the latency
to your broker from any VPS or PC.
What CNS is doing to help subscribers earn
maximum profits through the lowest possible latency
2009 is the year of growth for Commercial Network
Services. With data centers in New York City and San Diego, CNS
subscribers enjoy the lowest latency to all major brokers in the
USA.
We are just a few short weeks away from opening our newest data
center in the UK, to offer our subscribers the lowest latency to
Alpari-UK and other European brokers. (we will begin accepting
reservations soon) We are not stopping there – more data centers
are planned for 2009.
Some CNS subscribers have
trading VPS’s in both west and east coast data centers. It
is easy to setup a secure and private network to transfer files
between them. A future version of CNS Autoboot will leverage two
or more VPS’s into “clusters”, to run fully redundant trading
platforms across multiple CNS VPS’s and data centers.
If you have comments or questions
regarding the content of this article, please contact
CNS
Support and we will be happy to help you.

...a follow-up:
Wackena sent us a high frequency EA (Spikes_Martingale_From_Hell)
to setup in both data centers to get a better demonstration of
how latency can affect trading.
It's only been since market open on Sunday (less than 24 hours)
and here are his comments:
<snip>
Attached are excel files for NYC, SDCA and my local PC.
The yellow and blue sections indicated trade clusters. The red
section is the difference from NYC, SDCA and local PC.
On 2009.5.11 06:00, both ve's had BUY signal and my local PC had
a SELL signal.
The Buys that closed at 2009.05.11 18:43, the NYC reacted 1 pip
faster than SCDA and my local PC.
It appears in this short term review that VEs are better than
local PCs. And that NYC reacted quicker than SDCA and local PC
on two order close calls to server.
Wackena
<end>
I posted both statements here:
http://earesults.commercialnetworkservices.com/Spikes/SDCA/
http://earesults.commercialnetworkservices.com/Spikes/NYC/
I also zipped up the mentioned XL's and posted them here:
http://helpdesk.commercialnetworkservices.net/index.php?_m=downloads&_a=viewdownload&downloaditemid=97