The sets should be used on D1 timeframe. For example, optimization using wfo-demo-cluster.set from above post on EURUSD D1, for date ranges from 01.01.2017 to November 2019 produces the following cluster report:

As you see, the best combination of window/step sizes is 120/10, that is 120 days window for in-sample optimization and 10% forward step (12 days) out of sample. To jump quickly to the rolling walk-forward test for this combination just click on corresponding cell in any of the cluster tables, and get something like this (depends from your broker quotes and other tester settings):

This is the complete forward performance of the EA on the specified dates range (2 years), not a backtest.

I tried the EA today and it didn't produce a cluster report but only the overall summary.  

Feel free to send me more details (what you did exactly, what you see in logs etc).

Thanks for your rapid reply.

Here are what I had done and what I got in return.

I attach the optimization screen, inputs screen and the html result screenshots.

I haven't got the walk forward matrix, and that's what I need indeed.

Can you say where I go wrong?

walk forward matrix= cluster report

But one last question: As I know, wfo_expression  is 'custom indicator formula'. Does it mean that a coder can add his own criteria in there? 

For example I want to use Sortino Ratio (Annualized Profit/Downside Standard Deviation). Can it be done?

Thank you in advance.

The formula supports only a limited set of predefined variables (statistics) you can use for calculations - see full description in the blogpost. This is a simple way, which does not require coding - an end-user can elaborate and enter his/her formula on their own. Unfortunately, the default statistics do not contain enough information for calculation Sortino Ratio. On the other hand, the library provides much more powerful way of custom calculations using a callback function - it can be assigned in EA's source code via wfo_setCustomPerformanceMeter. This is indeed a task for a coder (or an advanced user). There you can implement Sortino Ratio.

The sets should be used on D1 timeframe. For example, optimization using wfo-demo-cluster.set from above post on EURUSD D1, for date ranges from 01.01.2017 to November 2019 produces the following cluster report:

As you see, the best combination of window/step sizes is 120/10, that is 120 days window for in-sample optimization and 10% forward step (12 days) out of sample. To jump quickly to the rolling walk-forward test for this combination just click on corresponding cell in any of the cluster tables, and get something like this (depends from your broker quotes and other tester settings):

This is the complete forward performance of the EA on the specified dates range (2 years), not a backtest.