Saturday, 8 December 2012

An analysis of casualy rates in late Napoleonic battles

A discussion erupted in a Facebookgroup on the battle of Waterloo about the quality of the French army during the Waterloo campaign and the ability of Wellington as a general. I was intreagued when Barry, one of the participants came up with statistics to prove his point that the French army of 1815 was as good as any, or indeed better. On the other hand the French always lost to Wellington. This meant to him that only the superior generalship of Wellington had saved the day for the allies.

Now, the second part doesn´t actually follow logically from the first, but I want to focus here on Barry’s first argument and use of data. Barry put up this database to back up his argument, which is based on Digby Smith´s The Napoleonic Wars Data Book.

Didn´t have a picture of Digby Smith´s book at hand...

There were a few things in Barry´s analysis that had me suspicious. Firstly, his choice of battles for the Prussians seemed like an odd selection to me. At Jena and Auerstedt we are dealing with a different Prussian army than that of 1812-15, while Eylau and Leipzig were fought by multinational armies in which the Prussians had a limited role. This made the comparison look weak.

I am also not keen on comparing ‘Prussian’ battles against battles with Wellington in command. The former were led by 5 different commanders of differing ability on the Prussian and allied side.

But the main issue for me was that Barry only looked at it from the side of the rate of casualties the French had been able to inflict on their opponents. I felt that ignored the fact that this only brings insight when compared to that of the opponent. In a hotly contested battle, on a cramped battlefield, it was logical that casualties would be high for both sides. And so it turned out. In most cases where the French rate of inflicted casualties was high, the allied rate was high too. 

Offering an alternative

The real test of French quality would be how much better the were compared to their opponents. So I chose to divide the French rate by the allied rate to generate a measure of relative effectiveness.

I decided to expand the database with as many battles and smaller clashes from Digby Smith´s book as possible. I restricted my research to 1813-1815 for time’s sake, but I collected more battles and grouped them together. By increasing the number of battles in the analysis I have hoped to reduce the effect of outliers and give a more balanced view of the armies. I have 9 combats for the 1813 spring campaign, 18 for the autumn campaign, 25 for 1814 and 7 for 1815. 

Of course it is possible to challenge Digby Smith's data. For the few battles I’ve checked, the force strengths mentioned in other sources could be off by tens of thousands of men, depending on sources or Smith’s selection of forces he deemed to have been in contact. These choices will undoubtedly colour the outcomes, but for the sake of speed I have used what’s available, rather than repeat Smith’s work.

I have left out the number of prisoners in the losses where possible, because these don’t apply to the killing power of the armies, which is what Barry set out to analyse.

The data for individual battles I have attached here, so you can all check what I’ve done, do your own analyses or add new data to the database.
Now to the outcomes...

The French and allied armies of 1813-1815

I’ve presented the results per campaign in the table below. It shows that during the spring campaign of 1813, Napoleon had an advantage in numbers employed on the battlefield of almost 40%. However, he also suffered 80% more casualties in what was a campaign of moderate intensity.

Together this means that only one in eleven of his men employed for battle killed an opponent. The Russians and Prussians managed to kill a Frenchman for every four of their men employed. This was mostly the result of choosing good defensive positions at Lützen and Bautzen. Napoleon’s lack of cavalry was also a hampering factor in exploiting his victories.

Spring 1813
Autumn 1813
French : Allied strength
138 : 100
67 : 100
78 : 100
72 : 100
French : Allied Casualties
180 : 100
96 : 100
86 : 100
88 : 100
French relative effectiveness
40 : 100
157 : 100
150 : 100
158 : 100
French loss rate
Allied loss rate
Allied Casualties per French soldier
1 / 11
1 / 6
1 / 10
1 / 4
French Casualties per Allied soldier
1 / 4
1 / 10
1 / 14
1 / 6

By the autumn, the tables had turned on Napoleon and on average he was outnumbered on the battlefield by 3 to 2. But this time he managed to inflict slightly higher losses than he received himself. This meant that 6 French soldiers now accounted for one opponent, while 10 allied soldiers where needed to kill one Frenchman. 

The big defensive battles like Dresden and Leipzig weigh heavily in this analysis and the limited use of casualty figures is illustrated by the fact that battles like Katzbach, Kulm and Leipzig were lost with large numbers of prisoners (not counted here) despite favourable rates of casualties inflicted on the allies. So improved relative effectiveness didn’t lead to better results.

The frantic action of early 1814 saw the French with a slightly better balance of forces (3 to 4) and lower casualties than the allies. However, the intensity of the battles seems markedly lower, taking into account the loss rates. This is probably because there were few fights where both sides decided to stick in.

Allied casualties were now one in ten of all French soldiers employed, but it took the allies 14 soldiers to kill one Frenchman. Relative effectiveness of the French was a notch lower than in the autumn of 1813.

Enemy loss rates

In 1815 the French again faced superior forces (2 to 3) on the battlefield, but suffered fewer casualties than their opponents. The losses were heavy on both sides though and this is reflected by the fact that one allied soldier became a casualty for every four French soldiers on the battlefield and in reverse that Allies needed six to kill or wound a Frenchman. 

This means that the French relative effectiveness was as good as in the autumn of 1813. All this despite the fact that the Allies fought all their battles on the defensive. 

The 1815 campaign was the most intensive of the four, with 19% of all participants in battle becoming casualties, with 14% in spring 1813, 12% in autumn and 8% in 1814. 

The Waterloo campaign in detail

The French were almost as effective against the Anglo-Dutch armies as against the Prussians. In both cases one casualty for every four soldiers. The Anglo-Dutch  rate combined for Waterloo and Quatre-Bras was almost as high. The Prussian ability to inflict casualties on the French was lower however, with nine Prussians required to inflict one French casualty.

Enemy loss rates in the 1815 campaign

At Waterloo, the Anglo/Dutch troops did one casualty to the French for (slightly more than) every three of them. Battles with similar rates of casualties inflicted on the French are Borodino and Lützen, like Waterloo defensive battles in strong positions. Both Waterloo and Borodino were small and cramped battlefields. 

However, the French exacted a heavy toll on their opponents in these battles as well. Even against the Anglo-Dutch at Waterloo, the French rate of casualties inflicted was only slightly lower than the rate they sustained. This was on par with their performance at Leipzig where it took a bit less than four Frenchmen to kill or wound an allied soldier.

Battles with high enemy loss rates 

Desperate battles like Kulm and Vauchamps saw the rate between three and four soldiers. At Ligny the rate was better than one in three, at Borodino five French soldiers inflicted two casualties, and around Plancenoit every other French soldier accounted for a Prussian casualty.

The interesting feature here is that some of these actions were offensive, such as Waterloo, Borodino and Ligny. The French were thus able to inflict high rates of casualties even in attack.


All in all this shows that the relative effectiveness of the French army was about the same from the summer of 1813 to Waterloo. Only the 1813 spring campaign shows weak French performance. And with Waterloo only being second to Borodino in terms of crowding and bitterness, high casualty rates on both sides are not surprising.

On the other hand we have also seen that French numerical superiority in the Spring of 1813 didn´t automatically result in great success. Defensive terrain and cavalry shortages stood in the way. French relative effectiveness later also didn´t always win them battles. Other factors might be more important. A crowded battlefield, prepared defenses, artillery concentrations and generalship all play a role. Some are possible to quantify. But without emulating the work of Trevor Dupuy in depth, that´s going to be extremely hard. 

It is clear that the Prussian army wasn´t all that good in 1815, with a pretty consistent but low rates of enemy casualties. These low rates are comparable with the rates in the autumn of 1813 and 1814. The anomaly in the Prussian case, as well as in the French case therefor seems to be the spring campaign of 1813.

This makes me interested in the Russian rates for 1812 and the Austrian rates for 1809. I also haven’t gone into the Wellington comparison here. I will continue with a comparison of battles from the Peninsula and the 1809 and 1812 campaigns, at some later time. 


  1. Nice one Jur,

    Just wondering would the declining casualty rate not be more a case of morale ? The longer the war lasts the less men are willing to stay in the line of fire ? (something which can be seen in a lot of wars)The sharp increase in 1815 would then also be explained as being the final battles to end the war "for good" ?
    I would be very curious if you would count prisoners among the casualties if the results would be the same ? Since you really dont care if they die or surrender as an armycommander...

    I am not convinced that the effectiveness of an army should be measured at killing ability in itself but I can see why it is done (given actual figures to compare in the first place).

    1. Hi René,

      I fully agree that these numbers have limited value. The best they can do is give an impression of its killing and wounding power. It doesn't explain it however: it might result from training, fire discipline, tactical ability and the length and intensity of the battle.

      The strange thing is of course that the French rates were about the same from late 1813 to 1815 and higher than in spring 1813. If war fatigue had been the most important factor in combat effectiveness, then you would have seen a decline.

      You're right that killing and wounding people is only part of winning a battle. Especially if you agree with Napoleon that the moral is to the physical as three to one.

      Capturing prisoners is one of the best signs of a defeated enemy, together with losses from desertion. So on a campaign scale these numbers would offer a lot of useful hints on morale.

      If somebody did the same kind of analysis for Napoleonic units as Geoffry Parker did for the Army of Flanders, we'd have a much better idea of the state of the army. Maybe we should crowdfund a couple of PhDs.

      Your best point is of course that we do this kind of analysis because it's the best we have to work with. We don't have data that can measure battle effectiveness because it is an academic construct. An attempt to catch in one number a much bigger and more complex story of why an army fought as well or as bad as it did.

  2. My original very crude spreadsheet was based upon killed wounded missing as this was the data in Digby I used.
    There was no ulterior motive in selecting the particular battles, the point was that the theory had been put forward that the French Army at Waterloo was rubbish. My counter to this was if the army was so poor how did it cause so many casualties? To prove and extend this it was necessary to get a comparison. Thankfully Jur had taken the concept much further and has produced an excellent model.
    There are however two data questions.

    1.0 What casualty estimates did you use to establish the French Battle casualties?
    2.0 how did you get the ratio of casualties caused by the French at Waterloo to be the same for the Prussians and the Netherlands army?
    I make it 14,000 French causing 7,000 Prussian casualties = 2 per casualty
    58,000 French causing 15,000 Allied casualties = 3.8 per casualty
    Then of course one ought to take into account the hours fought. Prussians say 4 hours = 1,750 per hour
    Now of course the number of troops engaged varied each hour but that is a bit complex for this time of night.
    For the allies we have 9 hours which is 1,666 per hour but these were caused by 58,000 French. Complex but interesting stuff.
    Anyway well done Jur as the overall result supports my early assumption that in fact the Armee Du Nord was a fine instrument of war.
    Best regards

    1. Hi Barry,

      some answers first:
      1. It's all from Digby Smith
      2. I don't think I have the same rates of casualties inflicted by the French on the ANglo-Dutch and the Prussians. If you look at the graph the rate is 50% (1/2) against the Prussians and 26% (1/3.8) against the Anglo-Dutch.

      I wouldn't call the Armee Du Nord a fine instrument of war. It is just as bad as that of late 1813 and 1814. I would need to go back to the battles of 1805-9 to test whether they are as good. And of course, this depends on their opponents as well.

      Your addition of the length of the battle might help in explaining the amount of casualties inflicted, just like the size of the battlefield as I mentioned in my post.

      If I had that data readily available, I'd happily include it in the database. If you have a good source, let me know

  3. PS the earlier comment deleted was just a test message to get me registered

  4. So basically in the end it all comes down to a culmination of aspects beyond any measuring that on day one Army x wins whereas the same armies on another day might see Army y win.... I mean given a more prudent cavalry commander at Waterloo might see a very different result and, go^ forbid we would all have spoken French by now. Supporting my personal theory that every man counts ..... :-)

    1. Sure, models based on numbers are very limited in explaining, let alone preficting, especially when they are complex.

      Like in all these kinds of situations, a chain of small events can change the outcome. It is never just one thing you can identify that explains it all.

      And numbers are so manageable and comprehensive that it's very tempting to work with them rather than intangibles.

      Nevertheless, weather forecasts have improved markedly over time. So there is some value in them.


I appreciate comments. Let me know what you think!