Researching Spells: They have to be new

So, I wandered back to an excellent discussion of B/X spellbook limits at Ode to Black Dougal.

Looking at my comment I think I left an important idea out but researched spells, which I allowed to exceed the spellbook limitations, CANNOT be spells in the B/X rulesbooks or that you can be taught by other magic-users. Basically, the idea is there is a common legacy of magic, the spells TSR provided, that make up the limited spells. To become more than a copyist of others you much research your own.

I didn't have this in mind back then, but I think it is a crucial addition. In fact, I could swear this exact limitation (spell research means spell creation) was in the original Dungeon Master's Guide in conjunction with the Intelligence limits on spell (which I believe research allowed you to exceed). Maybe it is in some other early book but I could swear I read it back in the day (maybe one of Gary's rants in The Dragon).

Regardless, I think if you're going to allow Magic-Users to exceed whatever limit your rules set offers (as do all versions up to second edition where it became optional) you should only allow it by researching of entirely new spells.

2d6 Expldding on Doubles

What are the odds of scoring any value three through 20 on 2d6 if they explode on doubles? This, is of course, the same as saying "what are the odds on a T&T Saving Roll?" Because I'd like to convert the thief's skills to T&T style talents I needed to come up with values for levels one through 20. I'd have to look up the math to do an exact probability calculation so I wrote a Perl script to do 1,000,000,000 simulations instead. The results for those who are interested:

Roll	Odds	Odds <=	Odds =>
3	8.0%	8%	100%
4	8.0%	16%	92%
5	16%	32%	84%
6	16%	48%	68%
7	17%	65%	52%
8	9.0%	74%	35%
9	9.7%	83.3%	26%
10	1.3%	84.6%	15.4%
11	2.0%	86.6%	13.4%
12	1.4%	88%	12%
13	2.1%	90.1%	9.9%
14	1.4%	91.5%	8.5%
15	1.9%	93.4%	6.6%
16	1.2%	94.6%	5.4%
17	1.3%	95.9%	4.1%
18	0.1%	96.9%	3.1%
19	0.1%	97%	3%

Twenty was under a tenth of a percent even rounded so I excluded it.