There 3 markets are build in almost the same way, but with subtle differences.

We will start with the Nasdaq. We are NOT using GSB AUtomation in this example. GSB Automation is much simpler, especially when you want to do this multiple times.

This document is talking more about the practicalities of build trading systems, than the theory behind it.

There are 6 steps.

1) Allow GSB to choose the top 10 indicators on a market.

2) Build 50,000 systems, and allow GSB to choose the top 300, and the best of these 300 systems into Favorites B

Favorites B systems are all the systems that have good out of sample results.

All data before Feb 28 2016 is in sample, except that the indicator selection has only seen 50% of this data.

3) Put the favorite B  systems in family groups.

4) Look at the family group where most of the entire group has good out of sample results. You may get none, or quite a number.

You want at least 3 members in a family group.

5) Walk forward all members in this group. Choose systems that have got good walk forward results for live trading.

If you are using GSB Automation, run this 4 times and choose the best groups of systems from either the best or all runs.

6) Put the system into Tradestation / Multicharts / Ninja Trader.

If there is a reasonable length of time out of sample after the late date in GSB data file, check out of sample results are good.

One month for example, as further out of sample results is too short to be significant.

If using GSB Automation, steps 1 to 3 are done automatically, and can be repeated 4 times with the settings below.

GSB Automation is available on 14 day trial for any purchased users. Its not available for trial users.

Note you need to load turn the 150 or so indicators on and save the optsetting file before you run automation.

Purchased users can also use AnyIndicatorCross method, rather than cross indicator method. This works better on Nasdaq,

though there are some extra tests that are good to do.

1) Allow GSB to choose the top 10 indicators on a market.

Open the GSB Manager. Next load the Nasdaq opt settings shown.

Click M1, then play. If you cant see macros, then you did not load the NQ with advanced mode on.

This will load 87 indicators for trial users, and about 150 indicators if you are a purchased GSB user with beta features enabled

Check that you have Custom Indicator Parameter set. See above in red.

If this is the case, skip this next step below

Loading Custom Indicator parameter set.

Click m6. If there is a macro loaded already on m6, then right click m6. This will clear any previously linked macros.

Click on the most recent custom macro file below

Click m6, then play. This will load the custom inidicators.

Save the optsettings so the custom parameters are remembered.

Click m2, then play.

This will build 50,000 systems, and work out the best indicators for the market you are on.

The top 10 indicators will be chosen

50,000 systems is the optimal. If you want to do a quicker test, choose a smaller number. Note also that this test will stop after 2 hours 30 minutes

if 50,000 systems are not made.

Now you need to run as many workers as your machine allows.

GSB purchasers can just load GSB Resource manager. This takes care of the matter accord to your CPU and ram availability.

Trial users will have to start worker(s). Each worker will use extremely roughly 10 GB of ram. You also need to allow extremely roughly 3 to 6  GB of ram for the manager.

If you don't have this sort of capacity, hopefully you will get some free workers from users who contribute to the GSB cloud.

Shown here is what happens when you run M2 GreenMin10

You can see below that with 20 to 40 workers, 50,000 systems tool 96 minutes to build. These were all free cloud workers, but this is an unusually high number of workers that were on the free cloud. Free cloud power tends to be really good, or really poor - with little in-between. This depends if I'm personally using all my servers. You can hire a server with about 256 GB of ram for $10 a day.

The macro completes by choosing the top 10 indicators. Every time this test is run, results will vary significantly. This is a good thing as it gives more diverse output and also means GSB users will not all be building the same systems.

The macro will also have save these new settings as NQ30-AdvancedModeOn_GReen.gsboptset. Note that _Green was appended to the existing file name.

Assuming this file was saved, you can now close the manager, and re-open another manager. The new manager should open the last optsettings file that was saved.

2) Build 50,000 systems, and allow GSB to choose the top 300, and the best of these 300 systems into Favorites B

Favorites B systems are all the systems that have good out of sample results.

Note below the new manager has the 10 indicators selected, and the file opened was NQ30-AdvancedModeOn_GReen

click M3, then play. This will build 50,000 systems.

If you don't have this file NQ_is2016.2_2017.2_2018.2_2919.2_2021 linked to M3, then attache M3 to this file.

Building systems is about twice as fast as building indicators.

Note the best results (average of 4 identical tests) using all the beta tester indicators, and the entry mode AnyInidcatorCrossed is below

The outputs are significantly better in all out of sample periods than the example above. This is the output of the VB spreadsheet that comes with GSB Automation.

What systems are in favorites B?

Look at the macro 2 pictures above here.

Favorite b systems have good metrics from the first out of sample period (2016.3) to 2019.2.28

So we now want to look further out of sample and choose families that most or all members have good out of sample results.

In the example above we have 118 systems in Favorites B. The bigger this number the better.

We have 36 familes, and the top family has 38 members. What that means is the top system has come up 38 times in favorites B, but the parameters are different.

What we want is the results of all 38 variations in the out of sample period 2019.3.1 to 2021.2.28 to have good out of sample results.

Make sure you expand the families as shown above.

Note also how dates are set to ALL.

We now want to change the dates of all 118 systems to no trade. This gives us out of sample dates of 2019.2.28 to 2021.2.28

You can see on the graph, the out of sample results.

From the statistics done previously above, we can see the average stats of the 118 systems is in the red box below.

We see on family1 the results of all 38 systems are nice and consistant. However the average fitness of the 38 systems (2.45) is below the average of the 138.

So for this reason we choose another family.

Family 2 has an average fitness of 4.87, and all members are consistently profitable.

We will choose this family

Family 3 is a good example of a bad family. Sure some systems are profitable, but 3 of them are really poor. This implies it has great sensitive to parameters. (A bad thing)

.... to be continued.

6) Put the system into Tradestation / Multicharts / Ninja Trader.

It is critical to use tick settings for volume. Failure to do this will lead to severe miss-match between GSB and TS/MC/NT for any indicators that use volume. IE accumdist.

Exchange time is being used because the data in GSB must match TS. If your computer is not on central USA time, then the data would not match. This is a work around for this.