I think they really wanted to use the Wins-above-replacement model but realised they couldn't apply it without knowing the prior probabilities of victory. I agree completely that training of troops matter, morale matters, food and water matter, terrain matters, the technology available matters just as much (steel vs bronze, longbows vs composite bows, cavalry vs infantry, pikemen vs cavalry, musketmen vs pikemen etc). But they don't have any hard data on any of these so they made a simplistic estimate based on troop numbers. An example where this breaks down is in the Battle of Zama, where Hannibal had a numbers advantage over Scipio (40k to 35k) but not in cavalry (4k to 6k). 2k Doesn't sound like much, but it made the entire difference between a small defeat and a catastrophic end to the entire war.

But there are a couple of other issues with the model

1. This model gives more points for fighting multiple small battles and winning them unconvincingly. Whereas it's obvious that a single crushing victory that completely eliminates an opponent should count for more. By that metric, Alexander's few victories should rank well above Napoleon's multiple small ones.

2. How much of the General's success can be attributed to himself and how much to having able aides? How much did Murat contribute to Napoleon's success, or Antony to Caesar's? Unlike a modern basketball team where you can measure a player's impact by seeing the team with and without them, it is difficult to quantify how much or how little Murat contributed to Napoleon's victories.

Beautifully said. I was appalled to realize the shallowness of the author’s methods. Then upon this sandpit they attempt to construct the un-caveated thesis: “Napoleon is the best general ever”.

Is WAR even useful for comparing agents with different numbers of data points?

I completely agree.

What if losing the war causing you to go into exile multiplies by 0.

Irresponsible IMO. I run a data website and I find quantifying things like Protein Per Second somewhat hard to quantify given things like Beef Jerky, McDonalds and the variable distance to a location, etc... But the claim isnt that this is 'the best ever'. The claim is that you can use the data to make your own decisions.

Too much buzzfeed quality content exists, I'm glad people are objecting to irresponsible data.

> Anyone with passing familiarity of military history will tell you that professional soldiers can easily outperform a force of untrained fresh conscripts several times their size.

And in open terrain, until at least the mid-19th century, horse nomads could utterly dominate even professional soldiers.