Something that a DM did on the fly was to take the average number of attacks and AC for each side and got a computer to generate the massive number of rolls, in the example below 2000 rolls for the first fight alone.
Each difference in average HP added 1 to the AC for that group. From this, each hit on the opposing AC would be a kill despite higher HP, etc. This result would be from one day (8 hours) of battle, and would scale down the losses if an attack lasted less than that. The DM knew which side was most likely to win, but did this to get a rough idea of the number of troops we would have left at the end.
So using 1000 gnomes versus 1000 goblins:
Gnome average HP = 5, AC = 16, no attack bonuses
Goblin average HP = 4, AC = 15, no attack bonuses
HP difference(5-4) =1, add it to the gnome AC so it is 17 to hit.
Making a bunch of rolls to see how many hits, given 1000 attacks for each army in an open field with no protection or other strategies:
Goblins hit 143
Gnomes hit 253
So after a full day of fighting, the gnomes now have 857 troops left, and the goblins 747. The goblins decide to keep attacking for another hour to see if their luck will change:
Goblins hit 101 x 1/8 = 12.625, or 12 (round down)
Gnomes hit 199 x 1/8 = 24.875, or 24
The rolls are made like those above and are scaled back to 1/8 the losses for each side. The discouraged goblins would probably fall back as they think another full day of fighting will only yeild more losses. All of the rolls took only a minute to setup and execute, so all of this could be done in a quick time out (with a computer)
This adds the realism of not knowing who has won until the end of the fight. The players can also add to the events with magic, traps, tricks, whatever to help change the score at the DMs descresion. Also not included are bonuses given for defensible positions, ideal conditions, high ground, weather, etc. so there is a lot of room for change, but this can give you a very quick response to a battle you might not have time to a more detailed outcome to.