Topic: Backtesting and Evaluation of live trading strategies
This topic about our expirience with automated trading.
How to develop successfull stratagy from scratch.
Discussion has been moved from: http://forexsb.com/forum/post/6128/#p6128
A new strategy is a strategy we are built/generated/created/found which we think is ready to go live/demo.
I believe there are a few stages to develop a successful strategy, let's define them.
1. To create/generate a strategy
2. To optimize it and see how it performs on different data/time range/time frame/pairs
as if strategy works on multiple TF there are more chance it stable in future results.
3. backtest on OOS with ratio 70/30 (just an example)
4. Evaluate it using individual parameter which Footon has just listed.
5. Set on Demo for some time to confirm that OOS and Demo almost equal
6. Move to real on low risk
7. Increase risk
8. Defined parameters are dropped below certain level, remove/re optimise etc, we need to define further steps.
I'm talking #4 and up.
~~~~~~~~~~~~~~~~~~~~~~
we define a rules for evaluating:
1. What parameters to take into account to evaluate a new strategy
2. What parameters to take into account to re-optimise strategy
2. What parameters to take into account to remove it from live trading
Please dump here all your thoughts, we will brainstorm this topic and I take time to summarize as we go on the first post.
So, let's do it.
~~~~~~~~~~~~~~~~~~~~~~
I personally don't use one strategy to trade, I strongly believe in diversification, as probability of DD with different strategies and different Currencies decrease dramatically as you add more strategies and low risk per one strategy.
Let's say you have Break Out system and scalping one, at any given time one has advantage over another, thus if each of them trades profitable by its own, and have a good curve, then when you combine then you might see that combined DD is even lower than the lowest of each of them. (Strategy#1 DD=12, Strategy#2 DD=7, ==>> combined can be between 6-9).
As more strategies we add with different trading logic the better (in most cases). it includes different time of trading, different TF etc, so margin is used only X% .
Technically speaking it is possible to reach high level of Sharp Ratio with diversified strategies.
Looking forward to your input.
~~~~~~~~~~~~~~~~~~~~~~
footon wrote:
hope you don't mind my intrusion into this topic, but here are my thoughts
1. What parameters to take into account to evaluate a new strategy
* the period of profitability
* win/loss ratio
* maximal drawdown
* number of trades in relation to trading period
* sharpe ratio
* linearity of equity curve
those parameters I put above apply more or less to all 3 questions Fxwinner stated, do you agree? and the period is actually a key point - for instance, how long should be the time span to evaluate the profitability (or any other parameter) of a strategy?
can we define exactly what we mean by "a new strategy"? is it newly built or generated strat? or just the starting material or building blocks, if you will, like the basic compilation of indicators which then are set up more accurately using optimization?
hope you understand what I mean, and that it doesn't sound too ridiculous.