Strategy Backtesting and Optimisation in Forex

 Between you and NT support, took all suggestions and tried.

It's true that Renko does not work in back tests.  But neither does
Better Renko (tried that).  Sim real time, market replay, back tests -
all give radically different results.  NT platform support says that NO
forex testing is accurate.  See answer below.  He suggests that Replay
works. but it doesn't.  Results are bogus compared to real time.

So seems to me that testing forex strategies, except in real time, is a
total waste of time.  But real time tests at a rate exponentially slower
than backtest or market replay, so I see no future in forex.

Is there an asset class that actually tests accurate with NT7, in your
experience?

"Hi Steve, The issue is that with futures and equities Last data is sent
and this is used for the chart and historical data.

With forex, there is no last price (because there is no centralized
exchange for forex while there are centralized exchanges for equities
and futures). The last price is the last trade on the exchange and will
be made using either bid and ask (so requires both) depending on the
direction of the last trade.

Since there is no last price with forex, NinjaTrader defaults to
using the Bid price on the chart and for historical data. You can
manually choose Ask data for the chart or for the historical data and
then the script and chart would use Ask data and not Bid data. If you
select Last (which is selected by default) then only Bid data loads
(again by default).

You can switch it to the Ask if you want, but then the Bid data would
not be included and it would be the same issue.

(In other words, the ask data is available but NinjaTrader can't use it
for the chart and historical data the way it uses Last data.

The reason last is an option in that drop-down is because this is used
for equity and futures. Meaning that we cannot remove the Last option
since this is still used for other instrument types.

Regarding the Market Replay, this is recording of the real market (The
datafeeds may be slightly different but all the prices are real).

I recommend doing further comparison by trading a particular day,
then downloading the market replay for that day and trading the exact
same session.

Please let me know if I may further assist."

*************************************************************************************************************************************************************************************Better renko does not lie as badly as the normal renko - due to the wicks and tails - always check for that with all bars especially the illusional ones that suck many people in such as mean renko and hybrids that alter the open of the bar and smooth it all to create a trading illusion/delusion on the chart

Yes Forex is a tricky one to backtest - also due to spreads... also to the way NT reports P&L in backtest foe some cross pairs such as GBPNZD GBPJPY doesnt seem to correlate to the realtime P&L...

We only test in realtime -we use backtest only to show what is completely unviable - and what might be plausible.
The out of data sample and walkforwards in realtime is the only we test.

So if you get really great results in a backtest can you divide the success by a margin of errorand still be viable for example - adding 1 to 2 ticks slippage is a good idea.

understanding the testing granularity tick by tick for example in backtesting etc.
try avoiding exotic/ parametersets - (fine for realtime) then making iot more complez - working to a baseline to compare.

replay can be of some use for that scenario - but only live trading is realistic slippage, latency, fill errors, connectivity etc

All backtesting is innacurate, some data more than others..
the nearest to live is realtime sim trading but even that is not real

We are not able verify the accuracy of the NT7 trade platform or data or provide trading or investment advice -
You can procure data from a high quality provider such as CQG - then comparing backtests to realtime tests
You should note and understand back adjusted data etc

Tip: If you are using MBTrading for Forex trading then use their historical data.
CQG0 use CQG
FXCM use FXCM
the same applies for any instrument.

stocks, Futures are more accurate due to spread and data quality

For futures - beware of:
Continous contracts - data joining - expiry dates and rollover - should you be flatt n periods before expiry and rollover and how about the rollover handling etc....

So yes its a massive huge undertaking to do it correctly - so one might even say backtesting can take several weeks to months to get right - and then you still got to test in realtime... so i recommend 3 months to 6 months as a good testing window for that- with 3 years data atleast - to try to cover different market conditions etc.

Yes Backtesting - Generally It is a waste of time unless you know how to collate the data and make use of it and understand the pitfalls.... but its good for immediately see what is totally out of the question and what might work - but you really do need a decent dataset to do this on.
- good for comparing dataseries versus another etc...
-always best to backtest/optimise on a basket
-utlimately it depends on your level of skill and interpretation only found via hands on experience of reading the results and the context of the tests...
-- backtest is a very useful form of intelligence - or realtime sim - that can be used as a guide for when to trade with the live system

recently i conducted a series of tests on 1 system for a client it took me 4 weeks to run the tests and collate the data and then present in a format which was useful - now we can try to select parameter ranges that robust and go forwards with setup that is sensible to the money manager -and test in live small accounts to limit costs.

so the exercise has a cost and budget and it is known in advance the test might return facts that say the system is not viable - or that might say its viable on a basket of non correlated instruments when markets are trending... there is a budget for all of that...

In FX we have the ability to use micro.mini lots for that testing
in Futures you have that - but slippage is not good for some styles of trading

Tip:You should select something with good intraday and weekly ranges
Popular futures contracts are e.g. CL,TF,GC,NQ,ES,ZB,ZW, etc

Bare in mind
generally swing trading is much better - less brain power, less stress, easier to live a normal life, huge rewards, slippage and latency is irrelevant...
day trading - feeds brokers - and is far less viable - people are attracted to the bright lights and slot machine style payouts and activity -

So then if you match up swing trading entry or buying/selling zones with intraday entry and let the system run... easy to do in FX - as no overnight margins or gaps like stocks and futures - then you might find something that is solid...

hope that helps

best

Tom

MicroTrends Support
support@microtrends.co
http://www.microtrends.co/support/

Tip! You can visit the help desk forum to see previous answered frequently asked questions:
http://microtrends.zendesk.com/forums