As a military historian who has done some research on probably a significant percentage of the battles on his database, I can tell you there are serious problems with the data. For many battles, especially in the ancient/classical/medieval time periods, we don't have any reliable figures for total numbers involved or casualties suffered by both sides - or either side. In many cases, we know who thought they had won, but we don't know who actually did better on the battlefield - many subsequent accounts sound suspiciously like Propaganda rather than reliable accounts.
Even modern combats are not immune from data managing. I have read, for instance, both German and Russian/Soviet accounts of the same battle in WWII in which both sides claim to have been outnumbered, and can prove it statistically. In fact, both sets of statistics were wrong, but they were the basis for most accounts of the battle, so, if that was one of Wiki's list of battles, just about everything about the data from it was Bogus.
Most important, however, is the false notion that Battle is everything, that winning 'on the battlefield' is the sole criteria for military success. This is simply Not True. Napoleon may have won the great majority of his battles, but the ones he lost cost him his country, his crown, and his freedom.
The German military won more battles than they lost in both World Wars, but they lost both wars. Contrary to their own excuses, they were not overwhelmed by superior numbers, they were simply beaten at the strategic level, because they didn't concentrate on winning at that level, and that is the level where wars are won and lost.
When Dupuy's HERO Institute did a study of battles to test and compile their mathematical model for predicting battle outcomes, they used a database of 100 battles, because that was how many they could be certain of having enough valid data for - and even then the statistical data for many of the earlier battles was 'estimated' (read: guesswork). They did not even attempt to study the effects of generalship on the battles, just the 'combat power' of the opponents. Even so, they did come up with one pattern that emerged solidly from their data:
Throughout history, from 1500 BCE to 1973 CE, the one constant is that raw numbers of troops Do Not Matter.
That is, out of 100 test battles, the smaller army won most of them.
What that means, of course, is that factors other than numbers: leadership, morale, training, terrain, weather, - and yes, Generalship - all had more affect on victory or defeat than pure numbers of troops.
