Extremely simple strategies with > 100% annual return

[jcl]

By filtering price data with highpass or lowpass filters, it's quite easy to write strategies in just a few lines of code that return about 100% per year with FOREX and CFDs. The simplest strategy I found looks like this (in lite-C)

var *Price = series(price());
var *Trend = series(LowPass(Price,1000));
Stop = ATR(100);

if(valley(Trend)) {
  sellShort();
  buyLong();
} else if(peak(Trend)) {
  sellLong();
  buyShort();
}

This strategy with only 9 lines of code returns about 75% with EUR/USD in a 4 years backtest. However it is not really tradeable, as the Sharpe Ratio is only 0.72.

But when combining several such simple filter strategies to a compound strategy that trades simultaneously with many filters and uncorrelated assets, I got up to 300% annual profit and Sharpe ratios up to 3 in a walk forward test. I'm asking myself if even higher profits aren't possible with more sophisticated strategies?

 

[abattia]

Quote from jcl:
... var *Trend = series(LowPass(Price,1000)) ...

"1000" is a frequency? What units? Or doesn't it matter?

 

[jcl]

"1000" is the filter frequency in bars:

http://zorro-trader.com/en/filter.htm

and the "100" is the number of bars for the ATR function.

Other values might give a better result, but I haven't tried them as the compound strategy optimizes those parameters anyway. The above script uses 1-hour bars.

 

[abattia]

Thanks.

Quote from jcl:
By filtering price data ...

As you'll be aware, "price data" (i.e. ticks) tell you nothing about queue position, and so nothing about realistic likelihood of execution at the prices in question. Is queue position relevant for the timescales you're considering?

 

[ssrrkk]

Quote from jcl:

By filtering price data with highpass or lowpass filters, it's quite easy to write strategies in just a few lines of code that return about 100% per year with FOREX and CFDs. The simplest strategy I found looks like this (in lite-C)

var *Price = series(price());
var *Trend = series(LowPass(Price,1000));
Stop = ATR(100);

if(valley(Trend)) {
  sellShort();
  buyLong();
} else if(peak(Trend)) {
  sellLong();
  buyShort();
}

This strategy with only 9 lines of code returns about 75% with EUR/USD in a 4 years backtest. However it is not really tradeable, as the Sharpe Ratio is only 0.72.

But when combining several such simple filter strategies to a compound strategy that trades simultaneously with many filters and uncorrelated assets, I got up to 300% annual profit and Sharpe ratios up to 3 in a walk forward test. I'm asking myself if even higher profits aren't possible with more sophisticated strategies?

The problem is not the sharpe ratio. The problem is this will not work with commission and slippage. It will consistently lose. I can almost guarantee it.

 

[ssrrkk]

Quote from ssrrkk:

The problem is not the sharpe ratio. The problem is this will not work with commission and slippage. It will consistently lose. I can almost guarantee it.

I am pretty sure every back tester has found lots of strategies (10s or 100s) that make money consistently without slippage and commission (myself included). The minute you add those costs, you will see every single one of them will become money losers. Then you will go on an extended journey to find the one that survives slippage and commission which for some takes multiple years.

 

[jcl]

Quote from abattia:

Thanks.

As you'll be aware, "price data" (i.e. ticks) tell you nothing about queue position, and so nothing about realistic likelihood of execution at the prices in question. Is queue position relevant for the timescales you're considering?

Yes. This is unrelated to the strategy script, but is a very valid argument for testing any strategy. The testing platform must calculate the typical slippage dependent on the tick volume, trade size, and the asset, and must consider it when testing strategies. This has been done in the Zorro platform that I'm using. It stores a slippage matrix. I hope all trade platforms do this for strategy testing.

 

[Scataphagos]

Quote from ssrrkk:

I am pretty sure every back tester has found lots of strategies (10s or 100s) that make money consistently without slippage and commission (myself included). The minute you add those costs, you will see every single one of them will become money losers. Then you will go on an extended journey to find the one that survives slippage and commission which for some takes multiple years.

I don't know how anybody can succeed as a scalper. Imagine there are some, but too manic for my blood.

To minimize the impact of slippage and commish, you need some variety of swing strategy.

 

[jcl]

Quote from ssrrkk:

I am pretty sure every back tester has found lots of strategies (10s or 100s) that make money consistently without slippage and commission (myself included). The minute you add those costs, you will see every single one of them will become money losers. Then you will go on an extended journey to find the one that survives slippage and commission which for some takes multiple years.

There is no commission when you trade Forex and CFDs. Trade costs are caused by slippage and spread.

I don't think that back testers tend to forget the commission, as this is quite trivial. The usual mistakes by back testers are more related to selecting the price data for optimizing and testing. At least, that were the mistakes that I made at first.

 

[austinp]

Quote from Scataphagos:

I don't know how anybody can succeed as a scalper. Imagine there are some, but too manic for my blood.

To minimize the impact of slippage and commish, you need some variety of swing strategy.

they don't... almost all of the big-time scalpers from yesteryear have either quietly switched to longer-term holds, or went broke and quit altogether.

the masses who are always late to any paradigm shift are just starting to catch on now