Probst explained the thinking for the twist to Entertainment Weekly while on location back in March. “First we decided on three tribes,” said Probst at base camp on the island of Caramoan. “That was the first thing. And that was just an effort to change up the way the numbers were playing. People were locking into these alliances of five. So we thought, okay, let’s try a three-tribe scenario and see what that does. Once we had the three tribes, we talked about are we going to do returnees? I love returnees. I love second chances. I love the losers bracket in sports. I love all of it. So I love these guys coming back.
Hmmm, so Jeff, we eliminate 6 of the 18 for a merge at 12. Assuming you balance the tribes well in terms of their odds of winning the 6 ICs (the game's ability to break a bonded tribe of 5 would diminish to the extent that the tribes are unequal) ... the expected value for each tribe would be to lose 2 ICs and merge with 4. But this of course ignores the basic math of variance and the reality of weakening. Let's set aside the issue of weakening (that a tribe that loses an IC is more likely to lose subsequent ICs) and just look at the math of variance. In a completely random game (equally matched tribes), the vast majority of cases will result in one tribe losing 0 or 1 ICs and thus merge with at least 5 (who are likely bonded). Probst has designed a game that does exactly opposite of what he intends. S26 in a 3x6, merge at 12, will almost insure that a tribe of at least 5 (presumably bonded) will exist at the merge.