How Can I Backtest Intraday Data?

drtro

Active member
Messages
216
Likes
21
I can only go back so far on the 1 minute time-frame in ThinkOrSwim. How can I test back a year, or five?

Here's my 10 day 1 minute time-frame result.

2nmrk.png
 
Well that's no good. But it's trading 1 contract of /CL.

Max trade P/L: 1,750.00

Total P/L 84,690.00

Total 1013 order(s)

That's all the information TOS gives me. Wish I could see total unrealized loss to define risk, average win, average loss, etc.
 
/CL 1 Contract 10 Day 1-Minute Time-Frame
1013 Total Trades 995 Winners 18 Losses (98.2%)
+$86 Average Winner -$64 Average Loser (1.34 R:R)
Total P/L $84,690

Is all backtesting this good, or did I break the system?
 
It is rarely that simple, have you factored in spread or commission costs plus slippage ?
 
/CL 1 Contract 10 Day 1-Minute Time-Frame
1013 Total Trades 995 Winners 18 Losses (98.2%)
+$86 Average Winner -$64 Average Loser (1.34 R:R)
Total P/L $84,690

Is all backtesting this good, or did I break the system?

Hi drtro

Just out of interest - what instrument are you doing it on ?

Over 100 trades per 24 hrs means approx 4 to 5 trades an hour - fine in the busy 10 hrs of the day but very difficult in the other 14 hrs unless your targets are just a 5 to 7 pip/ point move with your stop at 75% of your target size.

You will not get that result in real forward time - not just because of the spread and slippage etc even if you have a great robot - simply because the market is the market and theory and live conditions - never quite line up ;-) - well they might if you have the resources of super computers and a very large R & D budget etc etc.

Also - pick out another 5 x 10 day periods over next 5 -6 months to see how they compare - that would be interesting

GL

Regards

F
 
Hi drtro

Just out of interest - what instrument are you doing it on ?

Over 100 trades per 24 hrs means approx 4 to 5 trades an hour - fine in the busy 10 hrs of the day but very difficult in the other 14 hrs unless your targets are just a 5 to 7 pip/ point move with your stop at 75% of your target size.

You will not get that result in real forward time - not just because of the spread and slippage etc even if you have a great robot - simply because the market is the market and theory and live conditions - never quite line up ;-) - well they might if you have the resources of super computers and a very large R & D budget etc etc.


Also - pick out another 5 x 10 day periods over next 5 -6 months to see how they compare - that would be interesting

GL

Regards

F

This is a future contract, not FX. Don't get what you mean by pip? And of my years in trading futures, I've never seen /CL have a spread larger than 2 ticks, even when I was trading at 2AM.

I'll run some tests on other instruments. Will try a FX pair and few stocks.
 
The following is 10 Year Daily of the EUR/USD 100,000 units.
2yv53q0.png


And then here's 20 Day 5 Minute of the EUR/USD 100,000 units.
10yi9o1.png


20 Day 5 Minute of BABA 100 shares.
igf71s.png
 
Had my friend run it through NT with different spillage and exit timing settings, roughly the same results. Can trades be automated in TOS, or is this just a backtesting tool? Would like to know if I can trade paper money but I read online that it doesn't fill orders, it's just an indicator. Any thoughts?
 
I'm afraid the range bars may be flawed due to the lack of phantom bars. But I'm looking forward to testing the 1m and 5m timeframe.

Still don't understand the big deal about slippage and fees though.

10 days on the 1 minute time frame yielded $84,690.

$2.25 TD Ameritrade Commission + $0.02 NFA Fee + $1.45 GLOBEX Exchange Fee = $3.72/contract.

1,013 orders + 3.72 = 3768.36

84,690 - 3768.36 = 80921.64

$80,921.64 NET PnL.

Let's imply orders are executed within 5 ticks of the signal.

5 * 1,013 = 5,065.

tick = $10

5,065 * 10 = 50,650

$80,921.64 - 50,650 = $30,271.64 justified on 5 tick slippage on every trade.

Let me try to find something else that makes $30K every 2 weeks.
 
Top