The system had a difficulty X roll a dice pool of N d10 and count all the dice greater or equal to the difficulty X BUT subtract 1 for every 1 rolled.
Also if the skill was equal or greater then the difficulty the result was supposed to be an automatic success.
Combining the results and the thumb rules I converted the Vampire dice system to a strait d% with skill=difficulty=>100 EML and -10 EML for every skill level less.
Rather than sum up the -1's to to get results totaling less than 0 to determine "dramatic failures" I made it into the Harninc max EML is 95 and failures ening in 5 or 0 resulting CF (dramatic failure).
The result was immediately familiar to Harnmaster players, was a lot faster and easier than the Vampire system and produced similar results.
In the Vampire system rolling Damage or Effect though was often done separately from the initial roll: first roll dex+melee of 7d10 vs difficulty X and add it up >0 to determine if a success and then if successful roll str+melee of 8d10 vs difficulty of 6 to get damage. Instead the Success was a HarnMaster like d% roll and then the damage was rolled as in Vampire.
Although rolling Nd10 and adding together number of dice greater than X can be a bit bulky to determine Effect/Damage it does produce some beautiful probability distributions. With difficulty 6 (the default) it produces a binary distribution bell curve like flipping a number of coins and counting heads (not unlike rolling 3d6 or whatever). But altering the difficulty higher or lower than the 50/50 coin flipping situation shifts the bellcurve in ways rolling dice and adding would have difficulty reproducing.
Difficulty vs Effect example.jpg
In some ways what Harn Master achieves thru multiple success levels and xd6+weapon impact this other system accomplishes thru altering the bell curve. Imagine overlapping and combining the 1d6,2d6,3d6&4d6 impact dice as a rough bellcurve weighted to the low end - shaped similar to the difficulty 9 damage curve above for 10 dice (.1 near 0, peaks at 2, slopes slowly to 0 at 6).
(If you haven't guessed from the above our Harn Group is currently playing a round of 1992 AD Vampire.)
Anyway, even if you can't calculate it directly repetitive testing and a sufficient sample size can often produce an accurate model/estimation. We were actually using similar techniques a lot in my advanced engineering math courses in college - though obviously if you build a bad model with wrong assumptions you get results that have no bearing on reality (for example the numerous global warming model predictions).