Tackling Profit Per Trade Issues

Today I wish to spend a little time discussing profit per trade.  This is an important figure for any trading system and is sometimes overlooked when chasing great equity curves.

When it comes to trading system development, everyone has there own idea of what characteristics are important.  For some it is simply a smooth exponential equity curve during backtest.  For others it may be consistent annual profit.  For others, minimum drawdown, etc...

But whatever characteristics you are looking for in a trading system, don't turn a blind eye to profit per trade.  This is one of the most important backtest statistics.  Some people develop trading systems with extremely good equity curves but very low profit per trade.  I have seen some systems with less than 2% profit per trade.

For me, I won't consider trading any system that shows a profit of less than 10% profit per trade.  The reason is that slippage is often underestimated and the system will invariably perform worse going forward.  So 10% may ultimately be 5%.  If you start with 5% profit per trade then you will likely end up with an unprofitable system.  The objective for any system developer is to design a tradeable and profitable system.  There is no point in designing a beautiful backtest that can't be achieved in real life.

There may be some things you can do to improve the profit per trade if you have a system you like but the profit per trade is too low.   Here are some of the system issues and how to deal with them.

(The following discussion assumes use of Portfolio123.)

First of all, have a look at your stock universe definition and how it compares to your entry / exit rules.  Many developers put minimum liquidity definitions in their stock universe.  Such filters as:

    Close(0) >= 3                          {minimum stock price}
    AvgDailyTot(20) > 500000   {minimum average dollar-volume}
    MktCap > 100                           {minimum market capitalization}

These liquidity filters often work against the ranking system factors as the ranking system is often designed to reward lower market capitalization or spikes in trading volume.

So what happens when you buy a stock trading at $3 and the price immediately drops down to $2.95?  Because of the universe definition it is sold at the next rebalancing period simply because it dropped out of the stock universe.  It may have nothing to do with the stock becoming a poor hold.  It just had a minor drop in price.

The same logic applies for volume-based filters and market capitalization.  The trick is to introduce buy rules to prevent this from happening.  With the universe definition above, consider adding the following buy rules:

   Close(0) >= 3.50
   AvgDailyTot(20) > 600000
   MktCap > 125

Now the system will not buy stocks under $3.50 but once bought, the stock does not drop out of the stock universe until it drops below $3.00.  Similar logic for AvgDailyTot and MktCap.  When you introduce such buy rules you should notice an increase in profit per trade.  Keep in mind that there is often a trade-off - your system may end up with a reduction in overall profit.  You have to decide what is better IN REAL LIFE, not a backtest.

Another technique (I won't call it a mistake) is for rank buy/sell rules to be set too tight.  For example, the buy rule might be Rank > 99 and the sell rule might be Rank < 99.  Although these rules mean that the system only holds the best of the best stocks, it also means that the profit per trade is likely to be low.  Consider lowereing the rank for the sell rule.  I find what often works the best is a sell rule that is 4-5 times lower than the buy rule.  Some examples:

Buy Rule	Sell Rule
Rank > 99	Rank < 95
Rank > 98	Rank < 90
Rank > 95	Rank < 80
Rank > 90	Rank < 60

Now you will have to play around with the buy/sell ranks to get the best results for any given system.

In summary:

• Low profit per trade is not desirable
• Use higher buy rules than the filters for your stock universe
• Give some slack between rank buy and sell

Good luck with your trading!
Steve

Comments

[gfagerlin]

Good points.

I might add to your comment "don't turn a blind eye to profit per trade" that one also look at "Avg Days Held" while you are there. A 10% profit per trade over 60 days held is going to be about half as good as 10% over 30 days. I use Denny's favorite Gain/Stock/Day (Avg Return / Avg Days Held) to "normalize" things somewhat. Values around .3% have been good to me - bigger is better.

Regarding your 10% personal preference, since you are prone to gifts, have you consider sharing your rejects?

Glenn

[StockMarketStudent]

Glenn - excellent point about avg days held. One has to keep everything in balance. I'll think about what I can share!
Steve

[Free Mcx Tips trial]

Wow! This can be a single of the very best blogs I’ve at any time arrive throughout on this subject. Basically Magnificent.