From paper- to real-trading

[fliesen]

Hello
After many simulations, now I'd like to use my scripts in realtime. However, I was told that trading in real markets involves a lot of differences with virtual backtesting, and this is discouraging me.
How did you forum readers overpass this delicate step from theory to reality ?
Thanks for any help.

[P-Sionic]

I'm ahead of the same step. presently I'm running NT in simulation mode with FXCM data feed trying to tweak all variables and boundaries (e.g. ordersize, trading hours, etc.). After taht I'll open a live account via FXCM and wil try to hook it up to the FXCM demo account. Only when that seems to perform reasonable, I'll link it up to the FXCM live account.

[fliesen]

Thanks for your answer. We probably are at the same level (well, I already tested my codes and have a live account since long, previously I opened one with FXCM too but - just for your information - I found they have a pretty limited number of instruments).
In any case, I fear that hooking NT to live trading by yourself will be more complicated due to the reason I mentioned in my first post, but if you want to share your next steps, I'll appreciate and contribute on my turn with pleasure.

I remain interested to hear someone who already went through this stage (there should be many).

[koganam]

Hello
After many simulations, now I'd like to use my scripts in realtime. However, I was told that trading in real markets involves a lot of differences with virtual backtesting, and this is discouraging me.
How did you forum readers overpass this delicate step from theory to reality ?
Thanks for any help.

What exactly do you mean by "virtual backtesting"?

[balltrader]

Perhaps if you list your concerns, they can be address by experienced automators.

Example, worried about opening a position and getting stopped out too soon, esp if forex on FXCM. Or having position run against me too far before getting stopped out.

Does your code allow you to force close or force open positions as you watch it in real time, i.e. a panic button, if your automated code is not working as well as the backtesting did?

what differences have you heard about with backtesting?

do you have enough capital to risk an automated strategy or will a mistake wipe you out too much? what is the liquidity?

does your script print heavy debug statements to a file for post mortem review of real time variable values? so you can see what is happening compared to your backtesting?

[sledge]

https://ninjatrader.com/support/helpGuides/nt7/?discrepancies_real_time_vs_bac.htm

[fliesen]

koganam: by virtual backtesting I meant backtesting (tests on a demo account)
balltrader and sledge: the remarks pointed out by you, could all take place in a list of my concerns, and there can be surely more. This is exactly what discourages me: my scripts give me good results in all backtests, yet how reliable are they? If they disregard so many aspects of the real market, live trading is anyway going to be a hazard and I wonder how I may reduce the risks for my capital, whatever it is.

[koganam]

koganam: by virtual backtesting I meant backtesting (tests on a demo account)
balltrader and sledge: the remarks pointed out by you, could all take place in a list of my concerns, and there can be surely more. This is exactly what discourages me: my scripts give me good results in all backtests, yet how reliable are they? If they disregard so many aspects of the real market, live trading is anyway going to be a hazard and I wonder how I may reduce the risks for my capital, whatever it is.

There is only one real useful purpose for backtesting.

ref: http://ninjatrader.com/support/forum...75&postcount=2

After a system seems to be viable in backtest comes the real testing work.

1. It starts with a walk-forward test, to see how stable the system is, or if it requires persistent optimization, and how frequently.
2. Then using Market Replay to be sure that your backtest results will hold up in a period different from the one used to optimize the system.
3. Next comes running in Sim in a live market with live data.

Only if consistently profitable over a reasonable period, that encompasses different market conditions, should one then consider going live with minimum size. If that is profitable, it may be time to scale up using sensible risk controls, proper sizing and money management.

All of which is to say that if all that you have done is a backtest, typically that is nowhere near being ready to go live.

[fliesen]

Thanks. This is exactly what I was starting to fear: months of scripts and backtests with limited effects as to practical purposes.
Do you also suggest to me to ask for some technical advisor's help ?

[koganam]

Thanks. This is exactly what I was starting to fear: months of scripts and backtests with limited effects as to practical purposes.
Do you also suggest to me to ask for some technical advisor's help ?

I am not sure what you mean by "Technical Advisor", so it is hard to say, but if you think that you need help then you should probably seek it. You just have to figure out that with which you need help.

I am not sure that I would say your backtest was a waste, especially not if it looked promising. All that that means is that the system is now worth more vigorous examination in a manner closer to reality.