|
Amsterdam, The Netherlands
Market Wizard
Experience: Intermediate
Platform: NinjaTrader / IB /Rithmic
Broker: EdgeClear / InteractiveBrokers / Rithmic
Trading: DAX / (M)NQ / (M)ES
Frequency: Daily
Duration: Hours
Posts: 631 since Jul 2019
Thanks Given: 2,822
Thanks Received: 787
|
I have been working on a system for FDAX and I would like to ask for some feedback. Totally inspired by GFIs1’s successful journal, so it is based on the 9:00 – 9:30 bar (=initial balance=IB).
I created Excel files with 11 years of FDAX RTH data (from Kinetic). One file for each day of the week and each weekday can have the following situations:
a> IB long-short trade
b> IB long-long trade
c> IB short-short trade
d> IB short-long trade
So, 20 situations/files in total.
Each of these files contain data for:
Trade entry time every 30 minutes from 9:30 to 12:00 (6 options)
Trade exit time every 30 minutes from 10:00 to 15:30 (7-12 options, depending on the entry time)
Stoploss 5 to 60pts with 5pts interval (12 options)
These files have some other data, like:
2 days ago Open/Close red or green (2 options)
Prior day Hi/Lo pts
Prior day Op/Cl pts
Gap up, down or none
Gap closed during IB
plus data for the IB Hi/Lo, IB Op/Cl and IB Volume
And then the number of trades that would have happened with their result.
I excluded some German and US holidays, roll-over days and German IFO days. I use (Excel) pivot tables to analyse the data.
In 11 years, there are roughly 2800 trading days (after excluding holidays/IFO days), if you assume 50-50 on short and long IB’s, there are about 280 short trades per weekday and 280 long trades. Is the population size big enough to count on these statistics? And if yes, to rely on the results, how many trades come out of the analysis as a minimum to take the trade?
This example (from the file Tuesdays - long IB - long trades) might clarifies what I mean:
Without other filters, the best result would be 1867 pts with 243 trades (9:30-14:00 SL60). Not good.
Adding a filter on PrDay range Op/Cl: Green gives 1891 pts with 126 trades (9:30-15:30 SL55), 78 wins/48 losses. Sounds better, right?
Adding a filter on 2PrDay range Op/Cl: Red gives 1271 pts with 69 trades (9:30-15:30 SL55). 42 wins/27 losses. More points per trade. Not sure if this is better than the previous example with the 126 trades?!
Adding a filter on PrDay range Hi/Lo cat: 161-200, the best result is 654 pts with 11 trades (9:30-15:30 SL50), 8 wins/3 losses.
I would think that the last one doesn’t have enough samples, with only 11 trades in 11 year. What do you think, and what minimum would you use (based on the 280 population size)?
Thanks for your input,
Deetee
PS. Based on input from @askerix I made the Hi/Lo ranges (and other ranges) more dynamic (ATR and average volume based), but I already had this example and haven’t updated all the files yet, because I hope to get some more input here first.
|
