What is the difference between backtesting on a demo and live account?
I’m using EA’s in strategy tester. Is there a difference in profit between demo and live when using an EA? For example, is the execution usually better in demo?
Also, in my Live account I select ‘every tick’ in strategy tester so that I can slowly backtest previous charts. Is there a difference in profit than when I actually run the EA live? For example, are there no commission costs included when backtesting previous charts?
First and foremost, how’s your modeling quality on strategy test. If you are not testing at 99.9% quality then disregard all results as invalid. They simly can not be relied upon. There is a lot of work (kinda) to achieve 99.9% but it can be done and best of all it can be done for free.
Difference between live, demo and strategy tester are critical to understand when evaluating a bot. On a demo account and in tester you have instant exuction of your trade. And a little known fact, some brokers have different data feeds and pricing for demo accounts luring you in with tighter spreads than they’re live accounts. Which can also vary depending if your on they’re “A” or “B” books. So your results will be skewed in your favour and not repeatable in a live environment.
So when trading live, variables that will affect your results include latency. The time between when you hit your buy button and when its recieved at you brokers terminal. Then exuction speed. The time it takes for your broker to find a “counter trade” . Actually this is a bit of marketing bull. Your broker will always be your counter party as your trading OTC derivatives offered by them and them alone. No-one else is involved in your specutive trade. How they manage your risk is generally written in fine print serval pages into they PDS. It is not up to your broker to find the best fill price for your order
So for us poor layman there can be delays of significant time between when we hit the trade (manually or through a bot) and the trade being exucted. This is where the phrase slippage comes from. Your broker is always managing their risk. Being able to hedge your trade is more important to them than filling your order. Thats way generally you’ll get negative slippage. Not too much of an issue if your chasing triple digits pips, but if only chasing single digits it can break you. Then as you’ve identified there’s commissions costs.
At the end of the day it’s a process and that process is to evaluate potential. If using strategy test the bot shows potential you move to a demo account. If similar (but generally less) results are achieved you move to a small live account. If your broker sends you an email notifing you they are are changing servers you know they’ve change you on their books and your bot is working well. Start increasing your capital. 6 months down the track and things are still going strong add more capital.
ive noticed that backtests correlate more with trade history when you run timeframes on larger strategies with larger targets.
low timeframe scalp strategies usually dont correlate as much.
another thing you have to consider is that there are sometimes delays in execution with live accounts whereas demo accounts are instant which is another reason why high TF and small target scalps dont work as well.
Plork is right, the higher time frame and larger profit margins you shoot for the less effect that slippage and spread will have on you. Simply because the math works in your favor, if your EA takes on average 100 pip profits, then a 1 pip slippage is only 1% of what you are shoot for, where as if you take 5 pip profits, that same slippage is 20%. Because of that, I would say the much bigger concern with proper back testing isn’t slippage or non 100% perfect tick data, but rather curve fitting. Also for the record, slippage can work both for and against you, so it kind of works out to be a moot issue anyways. There is a lot of material online from people who have been writing EA’s for years, that will help refine your back testing process to help you achieve better live results.
my friend used same EA on demo/live of same broker. Results are different.
Not much, but still interesting. I found that quotes are different a little… this
is the reason, probably
Strategy tester seems to be useless because it sometime gets different results while running on different timeframes even when there is hard coding of the timeframe. Instead of using Time[0] I use iTime(…) And still get strange results. Please can someone explain why
A demo account will only give you a rough idea about how your strategy or system would work out. Live account is the real thing. You can test it out before going live, good luck.
It’s obvious that the results that you get on demo trading are not what you will get when you go live. Demo trading is just to make you well-versed with the market so that you are prepared for some typical situations. But that doesn’t mean that you will be prepared for every situation that comes.