I think that you should check the individual numbers. I’ve coded it in a spreadsheet and manually checked each answer. I cannot see how you can return anything other than 923 from code. The increase of 99 STEEM with each additional unique character is logical too.

All the same = 0
1 additional variant = 99 more
2 additional variants = another 99 more.

For example, the winning number 111.

0 entries will win 1 STEEM.

The winning number 121:

All of the numbers starting with a 2 will win 1 STEEM.

Which is 99 more.

We're agreed to this point.

The winning number 123:

All of the numbers starting with a 2 or 3 will win 1 STEEM.

Another 99 more (there's no overlap with the "starting with 2/3 subsets").

So it's got to be 923. Which I think @event-horizon got first.

Where does the 1040 come from? This seems impossible without multiple rollovers.

EDIT: Based upon the probability of there being 0, 1 or 2 matching numbers in the answer, a user buying every single ticket every single week can expect an average return of 753.71 STEEM returned (assuming no rollovers).