What do you think of this strategy?

[chaykapwr]

Quote from intradaybill:

It is quite possible that you have implemented a look ahead of some kind in your backtest or you have fallen a victim of a backtesting flaw in the software you are using. These are typical in all platforms.

Forward testing has provided almost identical results.

I also consided look ahead, however, what I did was take random samples of data from different markets from different times. This reduced/eliminated any bias I might have had. If there was some look ahead, then the results should have been scattered. I.E. the chance of me getting the same results +- 5% for the win ratio, on different trading platforms and different data providers on different markets and different times is practically 0 unless the true mean is within that range. It is more than a coincidence

 

[chaykapwr]

Quote from abattia:

Looks great. Well done ...

With limit orders, slippage manifests itself in not getting filled when you want to get filled (because you are too far down the order queue to get filled) [Others may not call this "slippage", but let's not all get into an argument about what it's called ...]. If in real trading you miss some of the winners that you would have got in forward testing, sometimes the strat's stats can look quite different...

Perhaps you have already eliminated this as an area of possible uncertainty, but if you are using NinjaTrader, you can get (an overly pessimistic) feel for this by using Fill Type = "Default" setting, i.e. assume limit orders only fill is price moves at least 1 tick beyond your limit price. If you are not using NT, I would assume other packages have something similar.

You were forward testing on real-time market data? i.e. ticks?

forward testing is on real time data.

Fills only counted when it moved passed it minimum of 1 tick

 

[chaykapwr]

Here are the results for 10/1-1/13

Trades-144
Win Ratio-67%
Net Profit (ES Points)-314
Average Win/Average Loser-2.71
Profit Factor-5.73

 

[ssrrkk]

Quote from chaykapwr:

Here are the results for 10/1-1/13

Trades-144
Win Ratio-67%
Net Profit (ES Points)-314
Average Win/Average Loser-2.71
Profit Factor-5.73

Looks really good. Is the above from back testing or forward testing? When you say forward testing, are you running live on a simulated or small real account?

If so, can you show the stats on both the back and forward test? Are they similar? They should be similar but they cannot be identical because your executions will be different. You can calculate the actual slippage amount because you can see the executions in the back test and compare with the forward test and take the difference in the price it got and then average over all trades.

 

[JohnTrader]

It looks like this system has a "peeking" issue.

 

[chaykapwr]

Quote from JohnTrader:

It looks like this system has a "peeking" issue.

By peeking issue do you mean that because i backtested, and knowing where price ended up going that there was a bias ?

The stats given since oct were forward tested. If you dont mind explaining how that does not resolve this "peeking" issue?

 

[chaykapwr]

i want to further clarify that this strategy is strictly PA with no indicators what so ever.

 

[bwolinsky]

Quote from chaykapwr:

By peeking issue do you mean that because i backtested, and knowing where price ended up going that there was a bias ?

The stats given since oct were forward tested. If you dont mind explaining how that does not resolve this "peeking" issue?

No, you're strategy is peeking in the sense that you must be looking at future data to make current decisions.

This is considered "peeking." It's not the backtest itself but something in your code is wrong.

Also, it is not at all reasonable to not include slippage, nor is it reasonable to assume limit orders that go 1 tick past will get filled. Without slippage you end up with an overtraded result that doesn't work in reality.

The profit factor is too high.

 

[bwolinsky]

Quote from intradaybill:

It is quite possible that you have implemented a look ahead of some kind in your backtest or you have fallen a victim of a backtesting flaw in the software you are using. These are typical in all platforms.

That's probably a backtest in excel, vba, or custom chart.

Yes, it is obvious that that's peeking. Re-check your settings and making sure you're not looking at a future bar to make your decisions. No jaggedness and the fact it's a custom chart in excel implies this doesn't work in reality because of peeking.

The equity curve is not realistic, and was not produced in any standard platform and I know because I've seen them all, and yours is a custom chart probably out of excel.

If it's in excel, you can admit it, but it's not really that it is an excel chart but that I know equity curves that smooth aren't realistic, no matter what newbie argument you'd like to give about it.

 

[chaykapwr]

Quote from bwolinsky:

No, you're strategy is peeking in the sense that you must be looking at future data to make current decisions.

This is considered "peeking." It's not the backtest itself but something in your code is wrong.

Also, it is not at all reasonable to not include slippage, nor is it reasonable to assume limit orders that go 1 tick past will get filled. Without slippage you end up with an overtraded result that doesn't work in reality.

The profit factor is too high.

My "zones" are defined sometimes 2 hours in advanced, there is no peeking. I.e. i do not wait for a moving average to close at a certain level, meaning my decisions are not based on the second a bar closes.

In all of my time trading in the ES, slippage has never been a problem. As stated earlier, even if i did account for slippage of 2 ticks (which I have never have happen to me in ES), it would not affect the system much because as stated before my targets are fixed. I would just move my target frome entry two ticks. In other words, lets say my first target is 4 points, well generally it moves in my favor at least 5 points. So the slippage would not be a problem.