Best Performing Strategies, share your thoughts...
- Thread starter xbaha
- Start date
Xbaha,
Current markets do have a great deal to do with the execution of a strategy. Price bars may have characteristics in the current market that match the traderâs strategy. But it is the change in these price bars characteristics that may stop the strategy dead in its tracks in the future. Or as many traders say they lost their âedgeâ.
I am not going to go into this in any detail on how to deal with this or I will start giving edges away that took me years to eke out. But I will give you one example that is from the text books to explain what I mean.
How a trader works with price volatility in strategy can make or break the strategy. For example your strategy says buy one tick above yesterdays high on the price bar when a set up signal occurs. Then a stop goes one tick below the low of the price bar. This strategy back tests well for the first 6 months of a year ago and then excellent for the last 6 months going forward. All seems to work great in the first 3 months of live trading. Then all of a sudden a political change hits the market and price gets volatile. Soon the losses pile up from this strategy that was doing great and it has to be shut down. So, what went wrong, it tested so well?
The answer was the trader never looked at what can happen in higher volatility. The strategy tested with the average price bar from the high to low at 2 points or an average 2 point loser. But when the volatility increased the average price bar went to 3 point losers while winners did not change as much. This increase of 1 point in losing trades turned a winning strategy into a loser.
What this example points up is price is dynamic not static. What it does not show is that price dynamics can change:
- Within the market and trading instrument.
- Between time intervals for the same trading instrument
- Between stocks or commodities in their respective markets.
What the trader must do is identify how to adjust their strategy dynamically so that the strategy can adjust to as many changes in price dynamics as possible so that these can be picked up in live trading. This is a difficult programming task and an even harder one to handle in trade management (because the expectancy can vary depending on trading conditions).
As programmers we tend to code a set up in static segments with specific rules the way we write them in a trading plan. We forget that as traders we personally can analyze a set up in great detail in our brains before taking or rejecting it. But coded in a strategy it is normally cut and dry using a small set of the rules our brains use.
Then once we code this dynamic strategy we find this strategy rule set doesnât work with the variations or other sets of price bars. Discovering the variations in all price data and coding it is what makes a strategy general purpose. However, there are very few if any general purpose strategies available today that work effectively and consistently with different types of price data through all kinds of volatility changes. Most of these dynamic strategies are coded by big institutions, with AI mined rules sets, on high speed mainframe computers with direct institutional market access.
Current markets do have a great deal to do with the execution of a strategy. Price bars may have characteristics in the current market that match the traderâs strategy. But it is the change in these price bars characteristics that may stop the strategy dead in its tracks in the future. Or as many traders say they lost their âedgeâ.
I am not going to go into this in any detail on how to deal with this or I will start giving edges away that took me years to eke out. But I will give you one example that is from the text books to explain what I mean.
How a trader works with price volatility in strategy can make or break the strategy. For example your strategy says buy one tick above yesterdays high on the price bar when a set up signal occurs. Then a stop goes one tick below the low of the price bar. This strategy back tests well for the first 6 months of a year ago and then excellent for the last 6 months going forward. All seems to work great in the first 3 months of live trading. Then all of a sudden a political change hits the market and price gets volatile. Soon the losses pile up from this strategy that was doing great and it has to be shut down. So, what went wrong, it tested so well?
The answer was the trader never looked at what can happen in higher volatility. The strategy tested with the average price bar from the high to low at 2 points or an average 2 point loser. But when the volatility increased the average price bar went to 3 point losers while winners did not change as much. This increase of 1 point in losing trades turned a winning strategy into a loser.
What this example points up is price is dynamic not static. What it does not show is that price dynamics can change:
- Within the market and trading instrument.
- Between time intervals for the same trading instrument
- Between stocks or commodities in their respective markets.
What the trader must do is identify how to adjust their strategy dynamically so that the strategy can adjust to as many changes in price dynamics as possible so that these can be picked up in live trading. This is a difficult programming task and an even harder one to handle in trade management (because the expectancy can vary depending on trading conditions).
As programmers we tend to code a set up in static segments with specific rules the way we write them in a trading plan. We forget that as traders we personally can analyze a set up in great detail in our brains before taking or rejecting it. But coded in a strategy it is normally cut and dry using a small set of the rules our brains use.
Then once we code this dynamic strategy we find this strategy rule set doesnât work with the variations or other sets of price bars. Discovering the variations in all price data and coding it is what makes a strategy general purpose. However, there are very few if any general purpose strategies available today that work effectively and consistently with different types of price data through all kinds of volatility changes. Most of these dynamic strategies are coded by big institutions, with AI mined rules sets, on high speed mainframe computers with direct institutional market access.
so this means that strategy must be written and optimized specifically to the current market its traded on, right?
Xbaha,
I think "human physiology" is the most important. I have a system that trades all markets:
If physiology is good
then buy is the market will go up
or sell if the market will go down
set(target) at maximum profit
(stop is not required since physiology has 100% win rate)
Now, to the cashier.
my first comment above doesn't really make any sense.. what i meant to say is that your strategy might be good and profitable during market market hours(NY) but not overnight or after hours for an index future ES. in FX or FX futures, a strategy might have an edge during a local hour like JPY trading only during Asia hours for whatever reason.
I'm not sure if this is what you mean, but it is common enough to create a random* time series and test against that. Basically this can, among other things, determine how much random element is part of the time series. Also look into cross validation. Cross validation helps avoid over-fitting. And yes, it would be expected for e-minis to perform differently than fx.
http://lmgtfy.com/?q=strategy+testing+random+time+series
*Note, there exist multiple definitions of "random".