FFTA2 or Final Fantasy Tactics A2 is a an old game for the Nintendo DS. It has a pretty dull and featureless plot and is probably the least popular of the demi trilogy. It's not particularly difficult, even on hard mode, and there are a handful of adequate guides and an ok wiki already on the internet.
Despite this, I spent well over a dozen hours feverishly crunching figures to see the best way to make number go up. Below is the summary, and under that I'll talk about methodology etc.
summary table sorted by weighted total stats
I've listed the best way to increase a combined weapon attack (WATK) and speed (SPD) or magick power (MPOW) and speed for each race. For example, a classic powerhouse for Humes is to take a base Ninja, and then give it 83 Ninja levels and 15 Parivir levels. The result is a unit that hits both the SPD cap (149) and WATK cap (999), a feat only rivaled by the pure Trickster, which maxes out SPD and MPOW.
Green highlights are for WATK and blue are for MPOW. The C and D columns are highlighted for visual clarity, as job1 levels are confusingly right next to job2.
Methodoloy
I ended up with a fairly simple approach, though it took me some time to figure out. So even though I'm presenting my steps neatly here, the process was more messy.
First, I calculated the best base jobs. These are jobs that give the best combined starting stats. As the speed cap is 149, each point in speed is roughly 6.7 times each point in another stat with a cap of 999. So weighting speed and combining SPD and WATK and SPD and MPOW we get a single figure for comparison.
For example, the Viera Assassin has a base SPD of 65 and WATK of 77. Weighting the speed by about 6.7 gives 435.81 for a total base weighted SPD+WATK of 512.81. Speed dominates base stat comparison which makes sense as about 1/3 of a unit's total maximum speed comes from its base. For this reason, base jobs are usually fast, even if their WATK or MPOW isn't as good. 
assassins are the best viera base both for fast MPOW and WATK
Next, I used a similar method to evaluate growths, weighting speed about 6.7 times MPOW and WATK gains. 
nu mou growth comparison
Then I "simulated" leveling by making a big spreadsheet with all jobs, to see if a secondary job was necessary to avoid excess stats, as it often is. For example, a pure Seeq Berseker will reach the WATK cap before the speed cap, so by mixing in a few levels from a faster job with lower WATK, like the Ranger, an effective higher combined SPD+WATK can be reached.
This table is too big to reasonably screenshot, but here is a portion of it anyway. 
I made two big tables, one for MPOW and one for WATK
I had to semi-manually "tune" most level combinations. Doubtless I could have used an equation for this but I'm not sure it would have been much faster.
I was able to use some heuristics to save time. For example, I knew that all secondary jobs with both a lower SPD and lower secondary stat than the primary job couldn't be optimal. In cases where there were still several potential jobs to tune, I picked the job with the highest speed stat for the race, and then eliminated all other jobs with a lower MPOW or WATK than this job.
Then I took the jobs with the highest combined scores and make the summary table seen at the top of the post. I thought I was done but there were two edge cases remaining. When optimizing Viera for MPOW or Gria for WATK, the base job doesn't match either of the two "growth" jobs. This is fine on paper but in practice you can't recruit a level one character. Recruits are your clan's average level which is never level 1 even at the start of the game. So in practice your recruits will be, say, level 10 and throw off the arithmetic.
To address this I wrote a script which compares all two job combinations and used it with Gria and Viera. It probably would have been faster to just use a spreadsheet but I thought there were more than two edgecases at first.
correctly applying level caps was especially fiddly
Finally I added these two practical alternatives to the theoretical best in the summary. They are close to the "mismatch" optimal classes, especially the Viera. There is a level threshold where the mismatch job combination matches the two job combination in strength but I hope the reader will forgive me for not calculating this marginalia.
Perhaps I would have been better served by simply looking at all two job combinations in the first place, but the approach I took is more thorough.
It's important to note that this optimization approach doesn't take into account things like HP or weapon defense. Nor does it account for powerful skills like double sword. Still, if you have the patience to only grind abilities on the bench and focus on arguably the most important stats which are SPD and a damage stat, this is theoretically the best way to do it as far as I can tell.
Oddities
- Even though Black Mage has the "best" growths for Nu Mou MPOW, because the Alchemist hits the MPOW cap with a faster speed, 0 levels of Black Mage are optimal.
- Even though the Nu Mou Beastmaster has excess WATK, it is still the fastest nu mou, so it cannot be improved upon in terms of SPD+WATK. It probably makes sense to mix in another job with better defense or magic or something if you are wanting a physical nu mou for some strange reason.
Comparison
There is already a stat optimization guide for this guide on gamefaqs. I think parts of it might be a little inaccurate, however, and I wanted a more robust analysis even if the basic approach of "this unit's growths are good they are probably the way to go" mostly works.
Misc
That I spent so much time on this is extra funny as I don't like grinding in games, and have little intention of reaching any of these theoretical maximums. There is some satisfaction in feeling like my characters are advancing as quickly as possible but even then I don't like the tracking of fussy details while playing.
