ForumSections

Trading Computer Hardware Benchmarking - Standard Test

New comment

Hi All,


It seems no one who benchmarks computer hardware components is interested in anything other than playing games.

This is not useful to us here. Since I have identified a gap in the 'benchmarking practise' I have decided to define a Standard Benchmark Test that we, the Trading Community, will find far more beneficial.

Note: we are testing PC performance, not building a trading system, so some settings may be different from what you might normally choose.

The Standard Back Test Benchmark

1. Settings

  • Client: MetaTrader4 with Birt's Tick Data Suite plugin
  • EA: MACD Sample.ex4 (default out-of-the-box EA included with MT4)
  • Symbol (Currency): AUDUSD (Australian Dollar vs. US Dollar)
  • Model: Every tick
  • Date (Test Window): 2014.01.01 - 2016.12.31
  • Use Tick Data: True
  • Period: H1
  • Optimization: False
  • Tick Data Settings:
    • Basic > Source: Dukascopy
    • Basic > Use Variable Spread: False
    • Basic > Enable Slippage: False
    • Advanced > Stop loss: True
    • Advanced > Stop orders: True
    • Advanced > Take profit: True
    • Advanced > Limit orders: True
    • Misc > Use FXT files to speed up optimizations: True
    • Misc > Overwrite the FXT with the new content: True
    • All other settings: default
  • Expert Properties:
    • Testing:
      • Initial deposit: 10000 USD
      • Positions: Long & Short
      • Optimized parameter: Balance
      • Genetic algorithm: False (remember, we want to hammer the PC for performance!)
    • Inputs:
      • TakeProfit: unticked - 50
      • Lots: unticked - 0.1
      • TrailingStop: unticked - 30
      • MACDOpenLevel: unticked - 3.0
      • MACDCloseLevel: unticked - 2.0
      • MATrendingPeriod: unticked - 26
    • Optimization:
      • Balance minimum: unticked - 200
      • Profit maximum: unticked - 10000
      • Minimal margin level %: unticked - 30
      • Maximal drawdown: unticked - 70
      • Consecutive loss: unticked - 5000
      • Consecutive loss trades: unticked - 10
      • Consecutive win: unticked - 10000
      • Consecutive win trades: unticked - 30

2. Click START

3. Measure and post your Results in reply to this thread:

  • Last entry should read:
    • YYYY.MM.DD HH:mm:ss.sss AUDUSD,H1: 67300345 tick events (18791 bars, 67300445 bar states) processed in H:mm:ss.sss (total time H:mm:ss.sss)
      • YYYY.MM.DD HH:mm:ss.sss is the Test timestamp - this is not important.
  • The benchmark results must include:
    • total time H:mm:ss.sss - taken from the Journal entry above
    • System HW Specification (minimum):
      • CPU Make/Model (exact)
      • CPU Base Clock Fequency (GHz)
      • CPU Actual Frequency used during test (GHz)
      • RAM Size (GB)
      • RAM Type (DDR1/2/3/4/x
      • RAM Speed (1066/2133/2666/3200 MHz etc.)
      • Hard Drive Model and Type (where data history is stored)

The Standard Optimization Benchmark

1. Settings as above, except the following:

  • Optimization: True
  • Expert Properties:
    • Inputs:
      • TakeProfit: ticked - 50, Start 50, Step 20, Stop 100
      • Lots: ticked - 0.1, Start 0.1, Step 0.3, Stop 1.0
      • TrailingStop: ticked - 30, Start 10, Step 10, Stop 40
      • MACDOpenLevel: ticked - 3.0, Start 3.0, Step 1.0, Stop 6.0
      • MACDCloseLevel: ticked - 2.0, Start 2.0, Step 1.0, Stop 4.0
      • MATrendingPeriod: ticked - 26, Start 13, Step 13, Stop 39

The above Optimization settings should produce a suite of 2304 different combinations it will test during the Optimization run.

2. Reply to this Publish your results, as per above.

There is no reason for me to do a benchmark test.
Juergen Frank:
There is no reason for me to do a benchmark test.

This is just for those who are looking to build a back testing and optimization machine, given MT4's serious single threaded limitations but widespread user-base.

You are under no obligation to benchmark your system. I'm not even sure why you replied...?

Steve:

This is just for those who are looking to build a back testing and optimization machine, given MT4's serious single threaded limitations but widespread user-base.

You are under no obligation to benchmark your system. I'm not even sure why you replied...?

That's my opinion and as long as this is an open post I can write as often as I want.

Juergen Frank:

That's my opinion and as long as this is an open post I can write as often as I want.

Indeed you can and your opinion is valid. Not sure what you gain by writing useless replies, though.
New comment