Some criterions seem to have an overall trend for the positive and negative trams. A 24-number criterion I like is: POSITIONAL LINES.

This goes like this:
654321 <-Spun order. 1 is most recent line spun. 6 is latest one.
235416 <- Actual layout lines
****** <- POSITION.

Example. We have this:
654321
235416

Now line #1 comes up. It is at #2.
6543(2)1
2354(1)6
***** 2 <-We place 2, it “pulls” the other.

We update, now line 1 comes to be the last one line spun:
654321
235461
*****2
Say now line #3 comes up:
6(5)4321 <- it is at “slot” #5.
2(3)5461
****2 5 <- We add its position.

Now we update line #3 to position 1, since it is now the latest spun:
654321
254613 <- updated.
****25

Ok, say line #3 repeats.
65432(1) <- It is a slot #1.
25461(3)
***25 1 <- We add it.

And no update required, since line 3 is already at position #1.

Now, line #4 comes.
65(4)321 <- It is a slot #4.
25(4)613
**251 4<- We add it.

And update line #4 to slot 1:
654321
256134 <- Updated.
**2514

OK I think you get the gist of it.

I make two Lw registries, one for 4 rightmost positions:
654321
256134
**{2514} <- Four at the right.

And another registry for the 4 positions at the left:
654321
256134
{**25}14 <- These four at the left.

The 4 at the right get a “w” when any of them hits like this:
654321
256134
**2514 <-Cross reference the positions.
6(5)(4)3(2)(1)
2(5)(6)1(3)(4) <-Actual lines making a “w”: 5,6,3,4.
**2514

There are usually very good cycles since the lines show a tendency to an internal rhythm of their positions. Originally I only tracked rightmost 4 positions, since they stick like magic some times, then at the bad cycles, the 4 at the left came up with a concentration, so I made up the second Lw cycle.

Of course, make a test in millions of spins and you can expect to have the exact result flat betting as everything, but this is one of the things I’m using because of the streaks.

See you around,
Victor.