Our algo's are rendered useless depending on the broker we use!

Kate
24 Oct 2013, 12:03

In my experience results are almost similar. But I always rely on MarketSeries and do not use exact ticks, prices or tick volume. Commissions are different of course.

cAlgoFx
24 Oct 2013, 12:08

It's advisable to run tests with different brokers especially if your robot executes on each tick.

Spotware
24 Oct 2013, 12:22

RE:

supafly said:

Has anyone here tried to simulare their algos on the platforms listed below and get similar results? How much can prices vary? Shouldn't prices be roughly the same? How do I know that the live data feed from the broker is not tampered with?

Just a note to all the forum readers, I have been informed from a reliable source that some brokers smooth the prices on spikes.

1.spotwares calgo
2.fxpro's calgo
3.liquidmarkets calgo

Brokers can have different Liquidity Providers and therefore there will be differences in execution. In fact, the same Liquidity Providers can stream different feeds to different clients.

supafly
24 Oct 2013, 16:35

I would advise you guys to download the platforms, run your algos on them and see the results for yourself. I am using MarketSeries and my system is NOT a scalping system that can be affected by broker commisions, floating spreads and even a price difference of 5 pips.

supafly
24 Oct 2013, 18:02

One other thing which i have noticed that is happening recently is that I might run my algo on 2 computers connected to the same network, they both use fxpro calgo but the results somewhat vary. The settings are exactly the same on both calgos, where all the parameters are preconfigured.

supafly
25 Oct 2013, 17:25

Correction for the last message. The reason why the results vary is that for some bizarre reason, most of the trades executed on both computers using fxpro calgo are the same but some are different. Some trades are skipped on one calgo while they are executed on the other calgo and vice versa.

What solution can spotware provide for this issue?

Spotware
28 Oct 2013, 17:56

First of all, one possible reason of this behavior is incorrect backtesting data. You can remove your backtesting cache and try to test robots one more time. Please do not remove cache completely, you can send it to us for further investigation.

We also suggest you to investigate actual robot behavior with backtesting chart. What exactly is wrong there? Some positions were not open? Opened at different price? Closed at different prices? Are indicator values the same or not? We can help you to investigate this if you send us your code (may be partially), discretion of how your strategy must behave, screenshots of charts, logs or something else that can be useful.

supafly
28 Oct 2013, 23:07

I have removed the cache from both the computers (I only use 1m EURUSD) and the new data was re-downloaded. The results from the backtests were exactly the same as before and they varied between the two computers. But I decided to take the EURUSD_Minute.ctb from one computer's cache and copy it to the other and when I ran the backtest on that computer, the results were 100% match for both computers. How can this in-consistency be fixed?

Spotware
29 Oct 2013, 11:02

Could you send us backtesting cache files from both computers to engage@spotware.com - we will check what exactly happen.

supafly
29 Oct 2013, 16:25

The files are 20mbs each and are too large to be sent by email. I’m sure that the calgo product manager is more than capable of simulating one algorithm on two separate computers and then analyzing the cache files of those two computers.

Spotware
29 Oct 2013, 17:23

Another question: As we understood you use the same broker's cAlgo on both computers. Do you use demo or live account both? Or one is live and one is demo?

algoforce
01 Nov 2013, 03:21

Data varies

supafly said:

Correction for the last message. The reason why the results vary is that for some bizarre reason, most of the trades executed on both computers using fxpro calgo are the same but some are different. Some trades are skipped on one calgo while they are executed on the other calgo and vice versa.

What solution can spotware provide for this issue?

Since you are using Market.Series, the data within these series will vary between brokers.  Strategies that trade onBar or rely heavily on OHLC often result in behavior as you mention. 

The easiest comparison between brokers is to link your demo accounts to myfxbook, start fresh on all systems(custom start date), and let the statistics build up. You can easily spot out a Missing signal or better yet compare the systems statistically.

Do you get a similar Backtest as Forward Test on at Same Broker/data feed? If not, I would start comparing +/- pips on entry and exit through some coding and tally up your slippage on your live test