I ended up making the first (but not the second more radical) change; stats are now rolled on a d8-4 and the range (prior to racial adjustment, etc.) is -3 to +4. This is opposed to the traditional D&D method of using 3d8 to roll stats and then calculate from the stat what the modifier is, which is the only numerical value that usually matters. This method skips the intermediate step and goes straight to the modifier as the stat. Because it's a single die roll rather than three, the distribution curve is flat rather than bell-shaped. I do, however, allow for four rolls and the drop of the lowest, and in theory, I even allow for rerolls if you get a really crocked roll. Unlike some boot camp GMs, I've never seen the attraction of making players stick with what they get entirely; rolling stats in order and forcing them to play a crocked character just isn't very fun for most players, in spite of lots of verbiage spilled on its attraction. I did a few samples just to see how it turned out, and I got a lot of good rolls with rolling 4 of them and ignoring the lowest, but I also got some really bad ones, and mostly an array that included one or maybe two good scores and then average scores for the rest. This method does tend to give better than "average" scores in practice (and mathematically, of course, that is what you'd expect). I was pretty happy with my sample of rolls; I prefer a system that is more likely to give you slightly top-heavy results.
There's a few other implications of this. A +1 to a stat is now different than it used to be (for the better, in my opinion. A +1 to a stat that didn't necessarily add anything to the modifier was always something that I didn't particularly like.) Whereas getting to 0 in a stat was where it had significant implications (death, for instance, or immobility) -5 is now that threshold. This means that the stat increases at 3rd, 6th and 9th level will actually have a guaranteed mechanical implication each time, unlike the way it was before, and the racial adjustments also have a guaranteed impact on play.
I also scrubbed the monster list to make sure that nobody is at -5 or lower; -4 is essentially unintelligent for a MND score, for instance—most animals and "mindless" creatures like zombies or skeletons. Some animals that have a reputation for being relatively intelligent are at -3. Really stupid people would be in the same range, whereas -2 is "not very bright" and -1 is "below average, but not particularly noteworthily so.) 0 in this scheme represents "average" for any stat. Of course, the math guys out there will note that for PCs, if you roll a d8-4 the average isn't actually 0, but 0.5; you're as likely to average 1 as 0. Of course, that's rolls for PCs. Different use of the word average. 0 is the default average for NPCs, if that ever needs to be known. I hope I didn't miss anything in my monster list scrubbing, but if I find that I did, I'll fix it without updating the version, I think. I also scrubbed the stats for common sense; a few monsters had MND scores in particular that were either too high or too low given how the monster is envisioned. I mostly focused on the MND stat, but not exclusively; a few DEX and STR scores were adjusted as well.
I thought I might scrub the monster list to remove some of the critters and put them in an appendix, but I decided not to after all. Speaking of which, I did also notice that the wose as an alternate character race with a +4 to STR and -2 to MND didn't make a lot of sense anymore, now that I altered the way stats worked, so I changed it to +3 and -1. That still gives a net +2, which is how it works using the RTPs for racial stats; they also break even on their racial skill bonuses. It is possible, I realize, that a wose character could be at MND -4 if he rolls really badly. Such a character would literally be of subhuman intelligence, and probably only barely capable of using language.
Of course, that's the appendix, but I know how it works. Since when has any player ever really ignored the material in the appendix or really treated it as optional? If it's there, it's expected to be used. I also thought about scrubbing the Appendix I monster tables since I added several more monsters, but I won't do that either; GMs can, if they're going to use those at all, modify them to taste as needed.
Oh, and I modified the included character sheet a small bit, since I no longer need to note the stat score and modifier as separate numbers.
This new v1.4 is, I believe, pretty final—although no doubt in a few months, I'll think of some other change to make and do something else with it. In fact, it's so final, that I'm thinking of starting a new series of posts where I literally post the rules here, in chunks. There's no need to, of course, but the way search engines work, people will find that perhaps that would not find the complete document otherwise.