Questioning backtest results

I'm curious how the OP ended up?

I just realized, after reading this thread, that the strategy I'm working on has entries and exits on the same bar. Which explains why the results are not as good when I go to smaller bar size. ;)

Thanks to all for the information, and I hope the OP will update with his results.
 
I have optimized a few signals now based on a 3-year backtest, and using multi-time frame elements of 5 minute entries and 1 minute order management.

I am continuing that effort, fine tuning it, customizing it for different future indexes, etc.

The original post I made ended up being worthless, because so many of the entries/exits were on the same bar. I have learned my lesson, with the help of ET, and moved on.

My strategies are mainly built on MACD and SMA's with a lot of fine tuning on exactly when to enter and exit.

GL!

Mike
 
Quote from ctrlbrk:

I have optimized a few signals now based on a 3-year backtest, and using multi-time frame elements of 5 minute entries and 1 minute order management.

I am continuing that effort, fine tuning it, customizing it for different future indexes, etc.

The original post I made ended up being worthless, because so many of the entries/exits were on the same bar. I have learned my lesson, with the help of ET, and moved on.

My strategies are mainly built on MACD and SMA's with a lot of fine tuning on exactly when to enter and exit.

GL!

Mike
More...

Thanks for reporting back, I appreciate it. I had to scrap my strategy. It was basically buying the low and selling 4 ticks later. Well almost. If only it were that easy! :)

I have yet to come up with something profitable but I keep working on it. I get the feeling that so many people have designed systems for ES that there may not be many edges left.
 
Quote from ctrlbrk:


The original post I made ended up being worthless, because so many of the entries/exits were on the same bar. I have learned my lesson, with the help of ET, and moved on.
More...

You always need to check your backtests against the charts to be certain that your backtested entries and exists are valid, and would have indeed filled at those prices.

And as you have found, if any backtested results seem unusually strong, they are suspect and need to be checked very carefully.

Everyone new to backtesting has many moments of exhilaration dreaming of quick riches, until they double check the charts and the reality sets in that their code was cheating the devil, or looking ahead in time.

Look out for backtests that log trades at the open or close of a bar when you intended the opposite. This can happen when using longer bars such as 5 or 10 mins. Also watch for buying lows or highs within longer bars which may have printed but didn't exist, or were spikes. Daily bars are famous for this. The shorter the timeframe you can test, the better.

Again, always go back to the charts. Could this trade have been made at these price levels?
 
Interested in whether you've tested your system live yet, Mike? Unfortunately, no system can be said to work until you have done the acid test -- trading live using real $$.

I am trading ES live using NT, after many months of toil using backtesting and live trading using the NT sim account. I plan to run live at a low scale (1 contract) for at least 6 months before reappraising the system and scaling up.

Backtesting is invaluable, but has it limits (too many to go into here). I would recommend backtesting on at least 10 years of historical data - anything less and you may find that your system works well for a few months and then suddenly starts to stink as the market dynamics change (e.g. volatility is high right now, but won't last forever...or maybe not!). Also, be VERY suspicious of any system that generates a high proportion of profitable trades: "if it sounds too good to be true, it probably is...".

Good luck.
 
I abandoned the initial strategy after my discovery.

I have live traded the revised strategy mentioned in my last post, the 3-year backtested strategy.

I am close to break even. The biggest issue with the strategy is the exits. If I sort Ninja by MFE (max favorable excursion) I can clearly see a lot of money is left on the table.

So right now I am trying to refine the exits.

Mike
 
I decided to update this thread with what I have basically spent ever waking moment on since my first post. Not joking. :)

This strategy uses a basic SMA cross as it's primary signal. From there it uses a couple of indicators to determine the market trend, and best entry and best exit.

It has no profit target, the exits are generated based on what is happening in trade. I do have a max cap of 120 minutes in which it will exit the trade. It also will exit a trade 10 minutes before the end of the day.

The strategy waits until one hour after market open before making any trades. I found the first hour to be too volatile.

The below screen images are for a 1 year period (beginning Nov 2007 to current day).

I look forward to comments and feedback.

<img src=http://www.mgbi.net/ninja/sma_long1.jpg >

<img src=http://www.mgbi.net/ninja/sma_long2.jpg width=800>

<img src=http://www.mgbi.net/ninja/sma_long3.jpg width=800>

<img src=http://www.mgbi.net/ninja/sma_long4.jpg width=800>

<img src=http://www.mgbi.net/ninja/sma_long5.jpg width=800>
 
Looks pretty good, but you need to add more for the expense of commissions and slippage. I suggest at least $10 per side per contract. $20 per trade, as a minumum.

This will bring your win/loss closer to 1/1, but with 66% profitable you will be in business.
 
Thanks for the feedback. Commission is correct at $ 4.80 per round trip.

If I recalculate using 4 ticks of slippage per round trip which should be more than ample worst-case, the numbers are now $ 12,528 net, draw down $ 1,798.00, percent profitable 63.57%.
 
You must log in or register to reply here.
Back
Top