Dan Snow's History Hit - Battle of Britain 'What Ifs'

Episode Date: March 30, 2020

Dr. Jamie Wood and Professor Niall Mackay at the University of York are mathematicians who love history. Sensible dudes. They released a paper which sent the rest of the history world into a meltdown when they tried to use the statistics of airframe losses from the Battle of Britain to test just how close Germany might have come to victory in the battle. Essentially (I think but then again I am totally innumerate) they tested what would happen if the loss ration on certain days had been replicated consistently. Anyway I wouldn't read my take on it, give it a listen and see if it makes sense to you. I loved these guys and I hope we get to work together again. For ad free versions of our entire podcast archive and hundreds of hours of history documentaries, interviews and films, including our new in depth documentary about the bombing war featuring James Holland and other historians, please signup to www.HistoryHit.TV Use code 'pod1' at checkout for your first month free and the following month for just £/$1.

Hi everybody, welcome to Dan Snow's History Hit. It is the 80th anniversary of the Battle of Britain. As a result we're going to be talking about that battle through the spring and summer, lots of big anniversaries coming up. All of our plans completely thrown into chaos as with everyone else in the world. Looks like we're not going to be going over with the little ships to Dunkirk this year which was the big hope but we are going to bring you lots of great audio and video content both on here on the History podcast, which is currently being listened to by record numbers of people, and also on our sister channel, our TV channel, historyhit.tv. That works like Netflix. It's just for history. Lots of you have been signing up. I think lots of parents keen to get their kids a safe place to go and immerse themselves in history while schools are not open. place to go and immerse themselves in history while schools are not open. It's great to have you all on the system. If you use the code POD1, P-O-D-1, you will get one month for free. So you

can use it as much as you like for free for the first month and the second month for just one pound, euro or dollar. So it's two months for one buck or one pound. Please go and check that out. We've got a new documentary out there today on some of the great speeches in the history of the House of Commons. I got allowed in to that famous chamber with its green benches. I've never been in there before in my life. So it's a huge honour that's going up this week. Lastly, we've got the audio podcast going out. We've got the TV channel up and running. We've also launched a three times a week, we're doing a history hit live, please go and check it out on YouTube. We're on timeline channel on YouTube. It's the biggest history channel a history hit live please go and check it out on youtube we're on timeline channel on youtube it's the biggest history channel on youtube so please go and check out

timeline and history hit live we've had our first one on friday on the black death we've got margaret millen coming up talking about at the end of the first world war and we've got lots of other wonderful historians on there as well so that will be live on youtube three times a week monday wednesdays and fridays the Timeline YouTube channel. So please go and check all that out. So speaking of the Battle of Britain, I talked to Dr. Jane Wood and Professor Neil Mackay at the University of York. Now they are mathematicians. They're not historians, but I think they're pretty good historians if they turn their hand to it.

They are mathematicians. They love history. They're passionate lovers of history and they have brains the size of planets. And so they have released a paper which sent the rest of the history world into meltdown. They tried to use the statistics from the Battle of Britain to test how close Germany might have come to victory. Essentially, I mean, I am obviously totally innumerate. So listen to the pod, and I may have got this wrong, even though I talked to them. They tested what would happen if that loss ratio of airframes was different because that's one thing we can measure very precisely is the number of aircraft on either side that were lost during that battle day by day and by switching days around like they said if if there were more days like the 23rd of

August or something would the outcome perhaps have been different and by doing that they're able to come up with a range of outcomes and see which ones were more probable than others that's what i think happened you'll now have to listen to the podcast to see if i have a clue what was on my own broadcast so enjoy this podcast with these two guys it was such an interesting one i'm going to get them back on looking at another part of history i think soon but all the back catalogue of podcasts they've come off itunes now they're available exclusively only on history hit tv you can use the code pod one you get lots of second world war documentaries on there if you're interested in that we've had we've had james holland victoria taylor um victor greg

and we've had paul beaver all on a recent documentary about air power in the second world war that was a very interesting one so go and check that out on the channel in the meantime enjoy dr jamie wood and professor neil mckay interesting ones. So go and check that out on the channel. In the meantime enjoy Dr Jamie Wood and Professor Neil Mackay. Guys thank you very much for coming on the podcast. You made the internet melt down because you a talked about the Battle of Britain, you b tried to apply some mathematical interrogation which historians get very nervous about because we're all innumerate, and c you it was dangerously counterfactual. Tell me, how did the adventure begin? Well in my case, I remember hearing a lecture by a chap called John Kingman on Lanchester's

models which were a simple way of trying to capture some of the maths of warfare. And my journey began by setting that as an exam question for undergraduate students and then setting it as a project for an undergraduate student. And that particular undergraduate student, I said, okay, find me some nice wartime data that we can try out these equations on. And the answer was very close to home. He came to me with the Battle of Britain. And that was ten years ago. And in fact, so there was a first paper ten years ago and then the new one that's caused the trouble just recently. But it's the same data basically and the reason it's good data is because one day is pretty much like another which is very good from the point of view

of statistical cleanness. The types of planes used, many aspects of the battle didn't change day by day through the battle and so it gave us a nice clean data set to play around with some maths on. I had a very peculiar routine that I was a graduate student studying physics and I was very privileged to meet somebody called Joseph Rotblat who at the time was only one of two surviving members of the Manhattan Project and that got me interested in World War II history and the atomic bomb and the development of it and from that I had a conversation with Neil over coffee when I joined the maths department at York and we started talking about methods and maths and history and well we've published several papers

now together on that. I think we were perhaps both slightly embarrassed to be history geeks underneath our mathematician exteriors and we gradually discovered that actually each of us was obsessed by historical stuff. I know those furtive conversations only too well. What do you feel is sort of missing from the way history is often told that you guys can help with? So I think the counterfactual history has a really bad name in history and frankly with some justification but what we can do with some of the mathematical techniques that we bring and being quite from the more applied end of maths we're quite happy to throw whatever we can at something to make the best answers, is we can actually start to put some numbers and some quantification on

the kind of distributions of things that might have happened. And that's something we find that historians kind of do but don't admit to, because you need to understand that to differentiate between the people who got lucky or the people who actually did the right thing and became unlucky. And history judges you very harshly because it only takes one mistake, one horse tack or whatever it is. Horseshoe nail. Horseshoe nail to send you off. When you make very sensible decisions as a historical actor,

you might end up getting a terrible reputation because of the proverbial horseshoe nail. So there's this debate in history, of course. Many people really don't like counterfactual history. Michael Howard said grown-up historians don't do counterfactual history. And that's an entirely natural and reasonable position. So what has been described as exuberant counterfactual history where you pile supposition on supposition is just not good historical analysis at all, of course. But then at the other extreme, if you write any history at all that is not just one thing after another, then you end up using the kind of language that at least implies that decisions were made and alternatives were possible. And somewhere in the middle, the people

who try and resolve this, Neil Ferguson, Richard Evans, the way they write about it, you have a sense that they don't really know quite how to occupy the middle ground. So there's a point at which Neil Ferguson says rather plaintively, how are we to distinguish probable from improbable unrealised alternatives? And then he just stops there and leaves it. And I guess our take is, well, let us at least try to apply some of the modern mathematical and statistical computing techniques that are available. And then we work with historians who have the correct reaction to counterfactual history. Be very careful. And their take would be that we do as much as we can reasonably with the numbers and then take that as a jumping off point for quite conventional

historical analysis. And the thing with the Battle of Britain story was that that was exactly what we did. And we arrived at absolutely the consensus view about the Battle of Britain, actually, which from our point of view was quite a good thing, because as test of the methodology if it came up with something really weird then we really would have had more work to do. The fact that the quantification was consistent with what most people believe from our point of view meant that the method was reasonable in retrospect. And so let's talk about what you came up with. Where did you come down on how close fighter command got to being

unable to carry on intercepting German raids? We were very careful not to adopt a position. So what we did, there's a mode of thinking in statistics, Bayesian thinking, where what you try and do is you begin with some view, you challenge it with evidence, and then you see how that view is modified on the basis of the evidence. So what we did was first of all to identify trained monoplane fighter pilots as the crucial constraint, not airframes, Britain was building plenty of airframes, but pilot availability. And we said, OK, so suppose you're the kind of historian who thinks that it was a very narrow margin indeed, that the Battle of Britain was won on a coin toss.

So you think Britain had a 50% chance, say, of winning in some sense, to be made precise, not losing too many pilots, winning the Battle of Britain. And then what we did was to play around with different contractions and expansions of the phases of the battle or different targeting and see how that historian ought rationally to change their probability of victory on the basis of those changes. So 50% might, in a very adverse situation for the British, change to 10%, let's say. If, on the other hand, you were a historian who believed that Britain was always going to win

and won easily, suppose you think that Britain had a 95% chance of winning, then that might come down in these counterfactual scenarios to 50%, something like that. So we don't adopt a view on whether Britain actually won the battle with ease or not, we simply adopt views on how, given your beliefs about that, you should change them after altering the conditions and phases of the battle. So when you alter conditions, what you're saying is on the 15th of August, for example, or 15th September, Battle of Britain Day, you just ratcheted up the number of hurricanes and spitfires that might have been lost?

Not quite. So what we're doing is we're doing this thing called bootstrapping. So the easiest way to think about it is the analogy we use there. So let's just imagine a simple bootstrap. What we do is we take every single day in the Battle of Britain and we write them almost literally on a bingo ball, okay, on a ball, we put them into a wheel and we spin it round and we pull out another battle of Britain, okay, so if we just do that then we just get a reordered battle of Britain.

Okay so you've got the profit and loss of each day, that's a horrible way of talking about it, but you've got the casualty figures on each day and then so the day the 15th of August goes back to the 11th of July? Could do, so that's a reordering but what we do instead is we we allow replacement so that means on many different battles of britain we could pull out uh bad days several times and pull out good days not at all so you pick the ball out of the urn with its numbers on it you write them down you chuck the ball back in and then pick out another one. So doing that with literally tens of thousands of times builds you up a distribution of possible battles of Britain. So that is going to be one of these bell curves

because you're not doing anything clever with it. So what we did then in addition to that is that we can say that certain days there was a particular strategy that the Germans were adopting. They were attacking airfields, they were doing reconnaissance, they were attacking London, whatever. And if you like, what we're going to do is we're going to weight those balls differently. We're going to make it more or less likely

that we choose those balls in different ones of the counterfactual scenarios. So what that means is that on days on a more kind of airfield heavy day, we pick out more of the balls where they attacked airfields. Now the problem with that, of course, is if we only have a small number of days where they did a particular tactic you are repeatedly pulling the same day out multiple times and you will inevitably do that and that means that you may be distorting it because of individual things that happen on the day. The weather could have been weird or something. Oh no, the weather's very weird. But one of the things that enables us to do this is the surprising

fact that each day is quite not very related to the one that happened previously or the one that happened, which is very unusual and enables us to do this. So it's a good, you love the data set. Yeah, I think that's a nice point to make about trying to resolve the dispute between people who like and people who don't like counterfactual history. You see, what we've got in the days of combat data that actually happened is essentially everything that actually happened, but nothing that didn't.

So if we play around to make counterfactuals out of that, well, we're not supposing anything that didn't actually happen in the battle. We're doing as much as we can, wringing as much information as we can from what actually did happen. So there's absolutely nothing in there about genuine counterfactuals for different planes, different tactics. Goring coming through sensors or yeah. Yeah. It's just saying how might the actual combat data that we see happening in the battle have been rearranged differently and caused things to play out a bit differently.

And so there's a version of the battle where the Germans relentlessly pounded the radar installations, but this is one of your infinite possibilities. Right. We don't do the radar because there weren't enough days when they really did pound the radar installations. That's the problem, I was going to say. There were enough where they attacked airfields. Now, a crucial danger in what we did is that I think there were 16 days on which the British fared particularly badly and the Germans were attacking airfields and we extended that I think to 43 in one of the scenarios.

Of course at the extreme you might say if that was a surprise attack, well you can never repeat a surprise attack so it's not a good thing to do to repeat it many times. The correctness of this depends to a large extent on the systematic advantages that the Goethe Germans get from attacking airfields. So it is a much shorter flight across the channel to the targets near the coast. It's a more difficult situation for the British to defend. Nevertheless what we can't do from the actual data is to try to observe whether or not the British could have found better ways as such an intensive attack developed to combat those tactics. Sure, move to different satellite airfields and whatever else.

And I would add to that that one of the things that's interesting as you do that is that you're always going to get this gaussian, this bell curve that's going to appear. And I had interesting discussions on social media afterwards about it, about people worrying about the fact that you as you get more and more days into it you're repeating the same day more often then somehow it's not going to go wrong i mean my answer to that is that yes you're right but if we're having a discussion about the tales of distributions of probabilities around the event that actually happened i think we've done something valuable because i think that's that's our interest actually is that can we can we have a discussion about how likely it was and how not likely it was and what

it looks like for a rare event to happen in these systems. It's crucial to say no black swans. So in Nassim Nicholas Taleb's phrase black swans, really weird things that you just couldn't predict and didn't know were going to happen. Our technique has absolutely none of those because it only uses the days that actually did happen. So you're not going to get strange, peculiar, unforeseen events turning up. So assuming the days that did happen, what days needed to be repeated for the Germans to have an even vague chance of winning the Battle of Britain? Well, we were just looking at this actually on the way there, and one of the clear ones is the airfield days. Yeah, attack on Biggin Hill. We were just looking at this actually on the way there and one of the clear ones is the airfield days.

Attack on Biggin Hill. Yeah, so the 15th of August is the clear one, which is very different. And I think the issue there was we were not unsurprised to see that the whole kind of London switch is a bad idea. I think what we were surprised about is that the shift of not continuing the attacks on the airfield is a similar shift to London. It's a similar level of mistake in terms of the distributional shift. So I think what we found, which was a surprise, is that it's the lack of any single coherent strategy, not the specifics of any given strategy, was what was causing the distribution to shift. One comment on Twitter was that, you know, surely you can't, there's only so many times you can bomb Biggin Hill into rubble. But we don't care about that.

What we care about is pilot numbers. If the RAF had continued to contest the skies to the extent to which Keith Park did on days of attacks on airfields, and we'd carried on getting the same kind of results, then Britain would have run short of pilots fairly quickly. I was having a discussion, actually, on email with Stephen Bungay, who wrote a lovely book about the Macmillan. He loves his data. Yeah, and I think so that the modelling he used relied more on airframes and of course the British were building more planes.

Essentially every pilot who was able to use one could have a shiny new monoplane fighter pretty much immediately. Going back to the counterfactuals, I said it was a nice jumping off point for historical analysis. We realised that once you leave the data behind, the trouble is that to get these things to happen you really need a very different mindset from Hitler and from Goering. And Jamie made the point really that what the Germans didn't have was a strategy. They knew they had to defeat the RAF but they had no idea how to do so or what the right thing to do was. Should they hit the factories or the airfields? Should they destroy the RAF in the air or on the ground? They didn't know.

And our feeling is that almost any strategy would have given them better results. I mean, the British played a blinder, as you commented, essentially in a kind of reverse slope defence. But Britain defended very well, especially because Keith Park was very careful to use his planes economically, sparingly, to disrupt German raids. Germans didn't really know what to do. It's an information asymmetry. It's often said that Britain had an information system which won the Battle of Britain, and the Germans trying to decide what to do over enemy territory just had no clue. Just making it up as they went along. Now, was there anyone doing anything even remotely similar to the job that you guys had done during the Battle of Britain or during the

Second World War? Varies a lot depending on the the area so we actually had a discussion about that so we actually struggled to find contemporary analysis like what we've done at least at that stage of the conflict there was an interesting thing we were reading about that you read about pro as well that lee mallory did some work after the battle of britain of doing the region because there's all this big wing sort of antagonism that was going around. And it's quite hard to really unpick exactly what took place in terms of the contemporary analysis.

And it's something that really interests us because, of course, Lanchester published his equations, which is where we started from in 1916. So this stuff was out there. And they were certainly interested in this prior to this. And it was very much of the vogue in the First World War era and just after it and in the inception of the vogue in the First World War era and just after it and in the inception of the RAF but 20 years on it's not quite clear where that appears in its...

This was the first couple of papers we wrote, this was actually me and historians and students, Jamie wasn't involved in those 10 years ago, was exactly on that. So Lanchester's equations are about what are the effects of numbers in battle. And of course, the big wing is precisely a question is, are mere numbers advantageous, all other things being equal? And again, we played around with some actual bootstrapping at the time, same kind of statistical model, and deduced that the Germans did better on days and in raids with large numbers than the British did.

So the big wing was probably not a good thing. And again, that's pretty much historical consensus. That seems to be the consensus, yeah. Land a Viking longship on island shores, scramble over the dunes of ancient Egypt, and avoid the Poisoner's Cup in Renaissance Florence. Each week on Echoes of History, we uncover the epic stories that inspire Assassin's Creed. We're stepping into feudal Japan in our special series Chasing Shadows, where samurai warlords and shinobi spies teach us the tactics and skills needed not only to

survive, but to conquer. Whether you're preparing for Assassin's Creed Shadows or fascinated by history and great stories, listen to Echoes of History, a Ubisoft podcast brought to you by History Hits. There are new episodes every week. What else have you applied this process to? So the the Lanchestrian idea is what, 1916? Of course, that's also the year of the Battle of Jutland. The impact on that is really one of the key things that we've looked at, or whether there

was any explicit impact. But it's certainly clear that if ever was there something that the Lanchester equations were applied to, first of all, naval battles were definitely in that bracket. I should say something for viewers about what the Lanchester equations are. In one sentence... Yeah, I mean, obviously I know, but I think the audience might not.

There's a lot of rubbish that's written in operations research papers. At its simplest, it's that both sides perceive a target-rich environment. It's kind of a fighter pilot's joke when you're outnumbered that there's a target-rich environment. But it's very rare to have situations in which each side can aim fire against a chosen target at any moment that there is no kind of overcrowding and the point Jamie's making is that one of the few situations when that clearly applies is dreadnought battleship fighting in lines because you've got lines of exactly so lines of battleships and every gun can find a target. The Battle of Britain and air combat is a bit different.

In fact, so, Lanchester's simple laws don't really apply to air combat, which is more like a sum of duels really. But the First World War at sea is especially and unusually interesting because it's one of the few situations where this idea that you cause damage in proportion to your numbers, which is the heart of Lanchester, really do apply. And they were definitely thinking about this naively. So there's this wonderful essay in the US school in 1905 with Bradley Fisk, where he has a spreadsheet in basically of this

cumulative effect of firing. And he has this idea of the first five minutes that it's so important for a battleship commander to fire unopposed for five minutes and then that will sway the battle and this this is exactly what you would get as a kind of rule of thumb from the lanchester solution so if we're looking at jutland what is the bits of data it's number of shells fired and the number of ships destroyed so what we actually think is the interesting one to look at which is what we did in great detail is dogger bank so we went back 16 months to look at the Battle of Dogger Bank. Now the reason why Dogger Bank is interesting is that it's essentially, it's not quite this, but it's not far off a trial run for the run to the south phase of the battle.

And we have good data as well for the Battle of Dogger Bank. So what we did with the modeling of that was we did a full reconstruction using a Bayesian framework which is relatively recent, approximate Bayesian computation so we took the idea that we know what the distributions are of the various parameters of the firing and then we tested them to see how compatible that was with the outcome and what we found was if we look across all the data that we've got is that actually what happened at Dogger Bank was extremely unlikely. So you have a situation where the British believed they'd won, yet their two lead ships took something like 33 shell hits between them, whereas they only hit the lead German ships three times in total.

But one of those caused a magazine explosion on the sailors, whereas all of the hits in the sunken ship was from the trail ship, which wasn't of the same class as the rest of the combatants in that battle. So it was written off as a victory, Doggerblank and there wasn't many lessons learned from it on the British side so there was this still obsession with rate of fire not of accuracy and there was no real attempt to address inherent vulnerabilities in the ships which they were very fortunate if you look at what happened at Jutland, they didn't happen at dog-a-blank. So we were able to actually quantify that.

And that has a really important lesson then when you go on to Jutland, that the Germans were able to learn from their loss at Doggerbank. I use the word in quotation marks. Whereas the British didn't learn from their win, which seems to be a kind of a classic situation. You learn more in defeat than victory. Absolutely.

So going back to the methodological point, the approximate Bayesian computation ideas, let me unpick that a bit. So the essence of the Bayesian idea is you begin with an idea, you challenge it with evidence, and you see what happens. Okay, so what's our idea? Okay, my idea is that this drug works. Think about testing a new drug,

and I have an idea that it works in a certain way. I do a trial. I discover that actually I need to tweak my ideas a little bit about how this drug works. Or in a military context I've got a new missile system, I know what I think its envelope of operation is, I do some trials, I get slightly surprising results and on the basis of that I change my ideas about the envelope of operation of this weapon. If you're a sane policy maker that makes an evidence-led policy. Exactly. So in the case of the First World War at sea,

what you do is you begin with some parameters, chances of hitting your opponent, the damage it will do, chance of destroying a turret, things like that. You challenge it with the evidence of the battle itself, and then, well, in this case, the point about Dogger Bank is that what actually happened is a really long way from consistency with the numbers. Because of the lucky hit.

Which means either your model's wrong, and we're fairly sure the model isn't wrong, there's ample evidence for that, or the parameters are far wrong, but the thing is that most of the ships and most of the conditions were the same at Jutland later as they were at Dogger Bank. So that gives you a very strong sense of what the numbers are, and that puts you in a position to deduce that what happened really was unlikely. Now, it's an unusual situation to be able to do that. In history, there's only one outcome, typically, or one situation. I think there's a danger with historians of a hindsight bias in which you think that what actually happened was, of course, likely, or was in some sense

inevitable, or is to be argued towards. And that, of course, as a taut sense inevitable or is to be argued towards. And that of course, as a tautology, is going to be true in most cases. But sometimes it's not and this technology of approximate Bayesian computation gives you a chance to discover if you've got nice clean data occasions when the unlikely happened. So the unlikely is the catastrophic flash that blew up the German battle cruiser? Or is the unlikely that the British ships took so much damage and didn't sink? Yes, the British got lucky. The British got lucky. The British got lucky.

To get it simplest, three battle cruisers blew up 16 months later at Jutland. Which was what should have happened? Much more probable. So Britain's victory was a catastrophe at Dogger Bank. I think that's quite a good way of putting it, actually. Or rather, OK, we lost three battlecruisers 16 months later at Jutland. In a sense, Jutland was still a strategic victory for the British, but it was certainly catastrophic for the crew on those three battlecruisers.

It would have been better to have one explosion at Dogger Bank and learn our lesson, certainly. And this, again, is a conventional view. And so the risk of getting into the weeds as a naval history geek, is it the fact that the British shells were rubbish, the German ships were better, or the British gunnery, or the command and control? So which bits did you have to tweak? Gunnery is one thing that I really focused on. So in the BCF, of course you've got this kind of strange dual system. The Battlecruiser fleet that that BT was commanding the BCF had this obsession with

rate of fire all about accuracy don't start me on that and the issue there is that the accuracy is a really key part of it they only landed three successful hits when they're actually in a kind of a fair battle in a parallel line with their opposing numbers and so we what we're based on this we actually have done some reconstructive events of the Jutland Clash, the Run to the South phase of the Clash. And it's interesting then to then see how you might vary the tactics or even the strategy of the Beatty and Hugh Evans Thomas to have a better outcome against them, given that you've got an ability to now hit them further. So remember that everybody doesn't know really how to use these things at that stage in 1914. It's still a bit of an unknown and there's things that were done like the closing of the range that he did persistently

which meant that he was actually allowed the German ships to fire even on him even though he actually had the ability. He had longer range. I mean useless men. Don't start me on BT. But German ships were hard to sink. Andrew Lambert has a nice comment that the German ships were Mercedes whereas British ships were Fords. They were very, very subdivided. When they were in harbour, the German crews actually stayed in barracks ashore. They weren't intended to be ocean-going ships. They were very survivable ships.

And British shells were faulty as well. And British shells were faulty. Could you work out that British shells were faulty? If we didn't know that British shells were faulty, could you have worked it out? I think not. No. So one of the interesting things about the British shells were faulty. So could you work out that British shells were faulty? Like if we didn't know that British shells were faulty, could you have worked it out? I think not. No. So one of the interesting things about the British shells is the HE, even though it didn't penetrate very well, it did quite a lot of secondary damage, which actually had the same net effect on the turret firing ability of the ship.

So things like von der Tamm, for example, was virtually intact after the run to the south phase, but couldn't fire any of its guns. It could jam some stuff in the mechanism. If you filter out, if you separate out the issue of explosion, flash fire explosion, the chance that a British hit on a ship would destroy a turret was about the same as for a German hit destroying a turret. Angelico knew this. As well as the Doggerbank paper, we then have a Jutland paper that's in the journal History, so that's just straight narrative history. And the story there really is of British system triumphing again. So you have Jellicoe as Director of Naval Ordnance back about ten years before Jutland, who, he's

a technocrat, he knows that British shells aren't as good as they should be and he's very involved in a committee which decides on the shape of the fleet they want to build. And this is at a point when Jackie Fisher wants always to build the newest and best possible ship to vault the existing ships and build something newer and better and faster. And a spanner is thrown in the works by a committee that Jellicoe is on in 1905-6, just when Dreadnought is undergoing sea trials I think. And that committee effectively says, no, we want more big guns on the sea before we start thinking about quality again.

We really do need a bigger, better fleet. And in the end that's what wins Jutland. It's this sense that if you have a big fleet with more big guns on the sea, and if you keep it together, which usually means keeping it in line, then whatever the Germans do, and despite the terrible tendency of British battlecruisers to explode, Britain inevitably wins the battle which is offered but not taken up by the Germans. And there's a reason why the Germans practiced again and again, doing battlecruise fleet turns so that they could turn through 180 degrees. If they find themselves heading into the British fleet they want to be able to turn

around and run away. Of course at Jutland they had to do that twice. They had to do that. The alternative was destruction. So in this occasion as well your mathematical model is agreeing with the sort of historical... Yes, I think there's some more interesting nuances to it as well. I think that one of the characters that comes out very well from our mathematical analysis is Hugh Evans Thomas, who gets a... I think he's a legend. He gets a bit of flak from some contemporary historical analysis, but... No one told him anything.

And he does a very good job. And our reconstructions we've done afterwards show that actually he was doing the right thing. I mean, essentially he didn't know he was awaiting a plan that never appeared. And, you know, he was in the right to think that if he couldn't keep up with the battlecruisers of the first and second battlecruiser fleet, then they might as well have, he might think that there was a trap that was being sprung or something like that. So he has every reason to do that. And somebody's decision-making is brilliant.

And so I think he comes out very strongly from the analysis. So, I mean, it only reinforces a view, but it's not necessarily a majority view, some of the character analysis. So what is next for you guys? Well, we have this rather nice collaboration. So there's the two of us. We're mathematicians, but we work with two historians, Chris Price and Ian Horwood. It's rather lovely because essentially what's going on is that we're mathematicians but we're historians man qu'est

and they're historians but unlike some historians who say, oh, I've never got any good at maths at school, they know that some things can be quantified and should be and we try and have completely honest discussions where anything that we want, we should never pull the wool over their eyes any piece of maths they should be able to insist that we take it back to one plus one equals two and explain it so we've written quite a bunch of papers some of which are in operations research journals but some of which are just narrative history so straightforward words one of the Battle of Britain papers called safety and numbers about the the ideas behind the big wing and

then this paper on on Jutland. So there's all sorts of things we can do between narrative history and bringing in new statistical techniques. We've got all sorts of things in the pipeline. One is to do with the development of aircraft carrier warfare and war gaming at the US Naval War College in the 20s and the effects and implications of that actually in the Pacific War in the Second World War. I think you hit the nail on the head I

think what's really interesting is it's not going back on it now with beautiful hindsight and just doing hitting it and doing basically wargaming it's about trying to get the history of how these things potentially made an impact on people's real decision-making during the actual events that took place and we've got good evidence that people were thinking about this a lot, particularly the naval one, because there was a lot of interest in how you should use battleships prior to the First World War breaking out. But the same is true with aircraft carriers between the wars,

that there was a lot of development about actually how deaths were used. And things like the Naval College informed the decision-making of Reeves, who basically invented the deck park because of the demands he realised were necessary from running carrier war simulations. Is there also something about the data set, so with battleships, your number of shells fired and your number of hits observed, it's sort of manageable. On the Somme, if you're firing a million shells, does that just make people's head, you can't plot each one's fall?

You'll notice that almost always, if I talk about the things that we're going to be working on, it's to do with air or naval warfare. Very rarely it's to do with land battles, because most people involved in a land battle are not concentrating on recording data. They've got other things on their mind. Kursk, for example, the great tank battle of 1943, I've had a student working on. But the data are just really not fine enough or well recorded enough to be able to do sensible modeling. Whereas as I say so you know naval warfare we can do, the Battle of the Atlantic is

an interesting one partly because of its implications for economic history and we've been looking at Vietnam. So Vietnam is interesting because of course McNamara was a great believer in data. You've got these rooms full of big computers, data on punched cards. You have something called the Hamlet Evaluation System, which is where the US tries to get data about the state of every little village, every little hamlet in Vietnam.

And can you imagine it? So the local boss says to some poor chap on a bicycle, will you go over to that hamlet and see who's in charge? He said, ask for the Viet Cong. And so he gets on his bike, goes 100 yards down the road, fills in the card, waits half a day and comes home. And what the US ends up with is garbage data, mostly. And there's a strong sense at the time of, you know, can we make better decisions with data?

Throwing all the modern statistical computing techniques at this data, in a word the answer is no, you can't. It's garbage in, garbage out, which I think perhaps is potentially an object lesson for people who would bring data in when it's not well understood and clean. So we want clean data. Anyone listening to this got a good idea of a good clean data set, let us know. Unfortunately, I'm trying. Anything before the 20th century is just hugely unreliable, isn't it, unfortunately? Yeah.

I should perhaps say, so the kind of techniques we're using are very unusual in history. Military history is quite good because often you do have clean data. In economic history, there's much more of this kind of thing. And typically, economics and econometrics, measuring economies, is a matter of what are called time series data. So things like the rate of thing. And typically economics and econometrics, measuring economies, is a matter of what are called time series data. So things like the rate of inflation. So you measure them month after month and day after day. There is an area of history that again is controversial, Clio dynamics. So Clio, the muse of history. And again, there's a

danger of trying to do things that are, I think in some cases, that are not fully warranted. But you can look back, for example, at inflation data and great inflations of the past. And with economic data, you can often do interesting, not so much counterfactual, but quantifiable things. Just my mind's worrying now. Is it the 18th century Navy just not quite? Frederick the Great did quite a lot of work with musketry and shooting at sheets and bits of paper to try and work out just how effective musketry was fired by an infantry battalion. But I can't believe they recorded, anyway.

I'll look it up and send it to you if I can find it. That's great. Well, that was just phenomenal. Thank you very much, guys. Stay in touch. Let's talk about your next one, whatever it might be. Thanks very much. Thanks a lot.

Thank you. let's talk about your next one whatever it might be thanks very much thanks a lot thank you i hope you enjoyed the podcast just before you go bit of a favor to ask i totally understand if you don't want to become a subscriber or pay me any cash money makes sense but if you could just do me a favorites for free go to itunes or wherever you get your podcast if you give it a five-star rating and give it an absolutely glowing review purge yourself give it a glowing review i'd really appreciate that it's tough weather that law of the jungle out there and uh i need all the fire support i can get so that will boost it up the charts it's so tiresome but if you could do it i'd be very very grateful thank you you