A lot of it is a return to stuff from older editions that I was glad to see go away, but there's one new mechanic that I'm not thrilled by: the ability check system. So, if your ability score matches or exceeds the DC, you succeed automatically. But if it doesn't, you roll d20+ability mod vs that DC. Which means that with, say, an ability score of 16, you automatically succeed all they way up to DC 16, but at DC 17, success chance drops all the way down to 40%. But then DC 18 and up, it just drops 5% each time. That sounds incredibly wonky.
This of course assumes a 2 to 1 score to mod conversion, as you pointed out. But truth be told... having now read these "leaked" playtests... it sounds a lot more likely that these modifiers are probably
1 to 1 up from 10 (which is a theory someone else put forth here on the boards several months ago, which I now crib.)
Evidence:
If your ability score matches or is higher than a DC, you succeed automatically. If we assume 10 is a +0 and everything goes up 1 to 1 (so that a 16 has a +6 mod for example)... then your ability score equals Taking 10 on something. So any check that you can Take 10 on (aka Passive Perception, or Insight et. al.)
should succeed automatically by definition. If you can't Take 10 on it... it means you HAVE to roll to attempt it. But like you said, if mods went up 1 point every 2 ability points, then the math goes wonky. If they go up 1 to 1, then it just becomes like any other mod check. Forex a DC 18 Dex check for a PC with a 13 Dex... that's a 1d20+3 to hit DC 18. Makes all the logical sense in the world.
Rolling for ability scores and the race/class +1 score bonuses. The problem with rolling scores and those +1 bonuses was the 'odd number' problem in 3E-- where a character had a better chance of being gimped compared to others depending on how many odd numbered ability scores showed up, and some racial adjustments not actually being bonuses because it just raised a score from the even number to the odd number (with no bonus to mod). By doing point-buy... you could get as many even numbers as you wanted (along with their beneficial mod bonus), and 4E's +2 racial modifiers meant you always got a mod bonus from your race.
However, if you raise mods 1 to 1... then the odd-number problem goes away, because EVERY score you roll is valid from a modifier point of view-- odd or even. That removes some of the stigma of rolling for ability scores. It also means that a +1 from your race or class also impacts your modifier. Those minor bonuses are all worthwhile.
Doing 1 to 1 also gets us closer to the idea that many people had wondered about for a long time about whether we should have just gotten rid of the 3-18 ability score format sacred cow and just did the modifiers themselves (which Blue Rose did I believe) seeing as how the 3-18 scores really didn't serve a purpose. A 1 to 1 modifier (thereby creating automatic Take 10 numbers) does make the 3-18 sacred cow actually viable and worthwhile.
*****
I'm not saying this theory is definitely correct... but it just seems to make more sense from the evidence we have thus far seen.