Looking at trade duration on the snapshot. Something is triggering at 100 minutes (00:01:39.000999). Is this a calculation error or by design?
1. Slippage of $2/side is way, way low...try $12/side. 2. Try the system on several years of 5 minute data and then report back on the success. A few days is worthless.
Probably a good idea lol.. I have no idea why the time calculations look like that sometimes, but is 100 minutes really that odd? I haven't checked into it. I would do this except the data with OEC is very limited... Of course the longer you can back test the better. Ticking that "intra bar back test" limits the back test to several days... I found I get much better results during the days in which all the volume is in that contract month. I will increase slippage to see how that affects results.
hehe, I wasn't trying to be mean or anything, just seems like you have a good thing going and crowd participation is fun. Would be easier to follow you and keep track of you on an RSS feed without cluttering my ET inbox. How about it?
I had some systems that looked a lot like this in backtesting, but I wasn't properly factoring in slippage and fees (mostly slippage). I took it live, and the first month, I made 100R, which I knew was so far out of likelihood that it made me nervous. The next couple months it came back to earth and after 4 months was a net loser. I stopped trading it because I couldn't understand the results (didn't match when I expected AT ALL). That was 2 years ago. I encourage you to start trading with a manageable amount and see how closely reality matches your projections. Keep your risk amounts pretty low and it will be worth the education even if you lose some money. Good luck, Damien
Very good advice, I still have much more testing and development left in my systems. I am currently running a couple of my strategies on a live demo account across the ES, NQ, YM, Crude Oil, Gold, EuroFX future and Dollar index. The nice thing is I can see my last order entry, as well as where the buy/sell signal originated. From what I have experienced, the fills are pretty similar to a real account. Can anyone attest to this as well? When you put "100R" does that mean 100 round turns or 100 thousand dollars lol?
I agree. Most of these backtests appear to have several serious issues. 1. For the ES I've found 6.25 per side is good for slippage plus 2.40 for commissions. So your looking at ~$18/per round turn minimum as far as costs go. This number will be higher for less liquid contracts. 2. I see you're getting the low and high of certain bars in the backtest. If true, you're using limit orders where you need to simulate a full tick trade through to get the fill. Unless you're the largest size in the queue, you will not be filled on those trades. Best thing to do is simulate all orders with the "buy at market next bar at close" statement. No limit orders unless you know how to test them properly. 3. 5-10 years is the minimum validation time for any intraday strategy. Looks like you're testing months back. This means nothing. 4. For the ES, a minimum profit per trade is $50. Anything under this and events beyond your control will make the strategy marginally profitable at best, i.e. bad fills, connection issues, user error, etc. Sorry if I'm coming across as harsh, but what you've shown is very far from tradeable. Mike
I realize the serious issues with these back tests. I am going to trade one of my systems - NightOwl (named so because it trades overnight) - for a month and see how it does in real time and see how it does. The program uses limit orders so I really don't understand that point about slippage. The order always goes in at a predetermined value +/- 0.25 point limit.