You can also support the show at Patreon.com/TheSwordGuy Patrons get access to the episode transcriptions as they are produced, the opportunity to suggest questions for upcoming guests, and even some outtakes from the interviews. Join us!
Kari Baker is a data scientist and swordswoman from Arizona who writes appallingly advanced data analysis articles for Sword STEM. In our conversation we discuss how data helped increase women’s participation in events and whether we can predict a double in tournaments.
You can find Kari’s Sword STEM articles here: http://swordstem.com/author/kbaker/
If you have any interesting research questions, or datasets you want to ask questions of, send them to Guy or pop over to the Sword STEM Facebook page.
Audiobook Bundle Special Offer!
As mentioned in the introduction to this episode, check out my audiobook bundle for The Theory and Practice of Historical Martial Arts. This includes the ebook and the audiobook version, narrated by Kelley Costigan. You can find it at guywindsor.net/tsg22. That link will get you 20% off the list price until 15th September 2022.
Transcript
GW: I'm here today with Kari Baker, who is a data scientist and swordswoman who writes appallingly advanced data analysis articles for Sword STEM. I'll be asking her to explain them here. So without further ado, Kari, welcome to the show.
KB: Thanks, Guy. I'm really excited to be here. It's an excellent opportunity to just talk about statistics and swords and all sorts of fun stuff.
GW: And I think you're going to be explaining quite a lot of it because I haven't studied maths for about 30 years.
KB: That’s OK. We'll get you up to speed.
GW: Splendid. And just to start off, whereabouts in the world are you?
KB: So I'm in Mesa, Arizona. The club I fight with is Mordhau, and that's under Brittany Reeves and Kyle Griswold.
GW: And regular listeners to the show will recognise Brittany's name because she has been on the show and what an excellent guest she was.
KB: Yeah, she's a great instructor, too.
GW: I can imagine. So is that how you got started? You just showed up to her class one day or is there a back story?
KB: No, there's definitely a bit of history and back story. I actually started HEMA in 2014. I went to a Comic-Con and one of the sword fighting groups in the area was doing a demonstration there and at the time I was doing Wing Chun and really into martial arts. So just seeing the sword fighting looked really cool and my martial arts mindset just went, ooh, I really want to try that out. And so the very next weekend, I was at the sword practise. I left for various reasons. And back in September, October of 2021, I kind of went, oh, I kind of miss this, but let's see what's going on and found out that Mordhau has been around and they were close to me. So just signed back up.
GW: What kind of swords do you prefer? Do you have a particular period or particular weapon?
KB: Mordhau is mainly a KDF school studying longsword. Occasionally we bring out the Messers or the sword and buckler, some polearms, occasional dagger thrown in there. On my own though, I've been studying smallsword, rapier, and I've been getting into dual wielding, and I do that with either arming swords or basket hilt swords.
GW: Oh, cool.
KB: It's really fun.
GW: Okay, so you're studying smallsword and rapier on your own. What sources are you using?
KB: So it's actually not 100% on my own. Justinder Singh out of New Mexico, he came to Mordhau when I was trying to figure out smallsword and he was like, oh, let me help you. So he sends me videos and instructions on how to do rapier and smallsword techniques, and I'll send him video back. He'll mark it up and analyse it and tell me what I'm doing right and wrong. So kind of like a distance learning thing. It’s really fun.
GW: That is a lot of work he's doing.
KB: Yes. And I really appreciate it. You know, Mordhau, we're not a rapier school. We're not a smallsword school, despite the fact that I'm trying really hard to start our rebel contingent of smallsworders.
GW: Brittany's a nice person. She'll say if you want to have a study group inside the club, I'm sure she'd be fine with it.
KB: We have free sparring days. We have time after class where it's pretty much anyone gets to do whatever they want to do. So Combat Con is coming up and they have a smallsword competition. And Mordhau, how despite the fact that we don't teach smallsword has the most people in the smallsword competitions.
GW: That’s classic. You’re starting a rebellion. Well, I love smallsword.
KB: Brittany says that none of us are her children anymore. It's a good time, but, maybe we'll pop off and do a little smallsword study group. It'll kind of be everyone figuring it out from text or you know what I can explain that Justinder has taught to me.
GW: Yeah, I have never produced anything on smallsword because I find the original sources, like Angelo's School of Fencing, for example, they're so absolutely transparent and easy to use. And it feels to me like any work I would do to explain it and videos or in a book or whatever would be kind of a waste of time. You can pick up that book and just copy it.
KB: For me, I feel like I'm a visual learner or I have to be like doing it as I'm reading it. It's really hard to, like, hold a book in your hand and…
GW: Book in one hand, sword in the other. It's tricky.
KB: Yeah. So having the videos makes it so much easier for me. So I'm really appreciative of all the work that he puts in.
GW: Oh, absolutely. I mean, it is easier to copy a movement on a video than it is out of a book, that's for sure. But it's funny, I've got a 1740 Girard behind me and that's in French. But it's been translated by Phil Corley, or at least most of it has, and it's like Angelo, you can just read it and it is so much easier than working with the medieval stuff. They just explain everything, they go into to all this detail about hold your hand with the nails down and the parry from tierce to quarte moves four inches.
KB: Oh, it's really that detailed? That's interesting.
GW: Oh, my God. Angelo particularly. Yeah, it's incredibly detailed.
KB: Yeah. I'll admit, I haven't looked at the smallsword manuals, so I'm not sure about them. But that's so interesting that it's so detailed that it gets down to the number of inches you move. That's amazing.
GW: Yeah. And sometimes just for fun in class, if I'm teaching smallsword I'll get a ruler out and get people to go from tierce to quarte and is it exactly four inches? Well of course, it doesn't have to be exactly four inches but it's like when they have the text that clear it's just super nice to be able to see how it works and that level of detail makes life much easier. Anyway, so you got started in historical martial arts by basically seeing a demonstration at Comic-Con. I've spent quite a lot of time demonstrating at events like that. And it's surprisingly effective, because many people going to events like that are like, oh, my God, I had no idea this even existed. Now I know it exists I want to go and do it. I only ever went the first time because I was invited because some of my students were organising the event and I was like, yeah, sure, I'll come and do a demonstration. And then it became a regular thing because it was so effective. So it's always nice to hear that people these are brought into the fold through this kind of outreach into other communities.
KB: Yeah, I know. I wasn't the only one that kind of got pulled in from that same demonstration. There were probably about half a dozen people or so that showed up because they saw it at Comic-Con. So I think it's definitely an effective marketing strategy.
GW: Excellent. So how did you get into writing articles for Sword STEM?
KB: So Sean Franklin and Brittany Reeves, they've known each other for a very long time. They were in Blood and Iron together when they were in Canada. And so Sean is kind of like a distance instructor for Mordhau. He did a bunch of videos during the pandemic and that sort of thing. But we have this discord server for Mordhau, and on there we talk about anything and everything. Very little of it actually relates to sword fighting. Sean and I seem to have the same taste in video games. So we just started talking, like private messaging about video games and stuff like that. And one day he's like, I hear you like data. Here's some data. And I was just like, oh, cool, nice dataset. Let's see what I can do with this. And so I threw together some graphs for the data and showed it to Sean maybe a couple of weeks later. And he just said, oh, wow, I didn't actually expect you to do anything with this data. Do you want to write a Sword STEM article about it? And I had never written articles before, so I was very apprehensive about it at first. But after about a day or two of thinking about it, I was just like, yeah, let's go ahead and take this opportunity. It sounds like it could be a good time. So that translated into the first article I wrote, which I think the title is something like, Hey, Where the HEMA Women at?
GW: I have it here to ask you about the details of that one. From the article you wrote, in the U.S., women make about 15% of the event’s participants, if a women's tournament is offered. If all the events offered a women's tournament in 2019, there would have been about 155 more participants across these events. That's fascinating to me. So what is going on there?
KB: Well, let me explain a key point about that article. So with the data we have, we don't have any documentation on if the person is male or female. That's just not a piece of data that's collected. You don't indicate your gender when you register for tournaments. It's not gathered. So the way we identify women is if they participated in a women's tournament. So, there could be a woman who participates only in the mixed tournament. We wouldn't be counting her as a woman for that.
GW: Because the data is not disaggregated by sex. Okay.
KB: Yeah. So that that's one of the key points about that article is we're identifying women as being people who entered a women's tournament. So the numbers in the article are a little under representative because there's some women who just don't participate in women's tournaments. They like the mixed ones.
GW: Or they go to the event and they don’t do the tournament at all, they do classes instead.
KB: Right. So that that's how we determine who's a woman.
GW: Honestly, it's as good a definition as any.
KB: Yeah. But some people when they were reading it, didn't quite pick up on that nuance. So I just wanted to make it clear that that's what’s going on here. So basically, the idea is if we look at tournaments that have both a mixed and a women's tournament, the number of participants in the women's tournament makes up about 15% of the overall tournament participation. And so if you look at tournaments that didn't have a women's event, you just add the extra on to make it be that 15% additional participation. And that's how we got to 155 number.
GW: So you're a data scientist by profession, right? That's your actual job. Health care data guru, according to my research.
KB: Yes. That is what the bio on Sword STEM says.
GW: Right. So what exactly does that entail?
KB: So I've been working in health care data analytics since I graduated undergrad, and I've worked in everything from health care quality to finance to insurance. I've worked for health care systems. So operational, clinical. Currently, what I do is I work with something called ‘social determinants of health’. And what that is, is it's things that affect your health that are not necessarily part of your physical attributes.
GW: Like income?
KB: Yeah, income.
GW: Rich people live longer, generally speaking.
KB: Yes. So like income. One of the ones that I like to think about is where you live and what kind of transportation you have access to. Maybe you live 20 miles from a grocery store, don't have a car. There isn't a good bus line to get to the grocery store. So you have to go to the gas station and that's a five minute walk. What can you get at the gas station? Well, chips and hot dogs or something like that, that's not going to manage your diabetes. That's the kind of thing that I work with.
GW: Okay. So would you choose an area to live based on how long it's likely to make you live?
KB: You know, you could do that. Make sure you're in a walkable distance to the necessities like your grocery store or your doctor or something along those lines. Make sure you're close to a public library. They have plenty of resources for people who might not have Internet access. So there's just tons of considerations when you're thinking about, where do I want to live? Maybe you want to go live in Portland, Oregon, because they actually have reliable public transportation, versus Arizona, where it's not so great.
GW: I've never been to Arizona, but I have been to Texas. And my feeling in Texas was basically, if you don't own a car, you're dead.
KB: Oh, it's not quite that bad in Arizona, but a car, I would say, is almost a necessity unless you live in very specific pockets of the different cities.
GW: Okay. And it's funny because here in the UK we are spoiled for public transport, we have like trains and buses and all sorts of ways of getting around the place. And the distances are a lot shorter. My parents live 300 miles away and that is a really, really long way in this country. That's like a six hour drive and that's incredibly far. You know, I have friends who drove 10 hours just to get to a sword fighting event.
KB: Yeah. My threshold for whether or not I fly or drive to a sword fighting event is if Google Maps says it takes more than 16 hours.
GW: Wow. So if Google Maps says you can drive it in less than 16 hours, you won't bother getting a flight. Wow.
KB: 14 to 16 I look at the prices, but 16 is like, no, I absolutely must fly.
GW: That is probably the most American thing I have ever heard, although actually the Australians would probably feel pretty much the same way.
KB: They've got a pretty big country there too.
GW: Yeah. So you have these data science skills, which is something I am completely unversed in. I mean I should know the difference between mean, median and mode, but that's about it. Oh, and I am familiar with standard deviation. But I'm guessing that quite a lot of the listeners have no idea of what they are really. So what you're doing with your articles is you're taking these data sets. Here’s a thought. Where does the data actually come from? How is it collected and how do you get access to it?
KB: So fortunately, Sean built HEMA Scorecard. So he has all the data from all the tournaments.
GW: What is HEMA Scorecard?
KB: Oh, HEMA Scorecard is a free tournament software that, at least in the US, is very prevalent, has very high usage at tournaments. So basically you can keep score, you can keep time, you can say specifically, “oh, it was the thrust to the head versus the cut to the body”. Not all event organisers use it the same. So sometimes there isn't that specific level of detail of thrust versus cut or head versus arm. But at minimum what you'll have is the number of points per exchange.
GW: So I had a sudden thought. You have the names of the competitors, right?
KB: Yes.
GW: Okay. And most names are gendered.
KB: Well, you'd have to make some assumptions.
GW: You would have to make some assumptions. And there would have to be some that you couldn't determine. But if you were looking to get more granular information on the genders of people coming to the tournaments have you thought about using separating that by name?
KB: Um, you know, it would be a really, really big assumption to do that. I don't think I would necessarily be comfortable doing that. But I suppose if someone wanted to do that, they could do that.
GW: That was just an idea that came to me in passing because if you're trying to figure out certain things and data hasn't been collected, it might be there some other way.
KB: I mean, there was one thing that I was thinking about on how to get better information on if someone is a woman. So with the data set I used for the article we had discussed, that data actually came from HEMA Ratings. HEMA Ratings has more data than HEMA Scorecard, more tournaments send their data to HEMA Ratings than use HEMA Scorecard.
GW: And just for people who don't know what HEMA Ratings is. Could you explain?
KB: Oh, yes, sure. So I forget the guy's name, but it's basically a website that tracks performance at various HEMA events and ranks you amongst all other HEMA competitors. And they have it broken down by weapon type. So you can see what your ranking is for longsword. You can see what your ranking is for single stick. You can see what your ranking is for rapier, etc., etc., down the line. So that’s HEMA Ratings.
GW: Okay. And that's completely separate to the HEMA Scorecard thing, which is software that Sean developed for basically running tournaments on. Is that correct?
KB: Right.
GW: Okay. Quite a lot of the listeners aren’t involved in the tournament scene at all. I mean, for some, they're going to be like, oh, yes, we know that already Guy, shut up. Come on, let's get on with the interesting stuff. But actually, I know because I get emails from listeners telling me these things that quite a few of them don't actually practise historical martial arts at all. Quite a few of them just do research and what have you, they are not interested in tournaments. So there's quite a broad range of listeners. So I'm going to be asking you to define quite a lot of things.
KB: No, I'm happy to. That makes complete sense. I mean, at Mordhau we probably have about 40 to 50 people in our club at any given time and maybe about ten of us go to tournaments. So it makes sense that, not everyone would be well versed in tournament lingo. But yeah, the data set for the women's article that came from HEMA Ratings. So what we had was the name of the tournament, the number of people in the mixed longsword event and the number of people in the women's longsword event and the number of people who were in both mixed and women's. So that's what the data set looked like. We didn't have specific names of people. I know the EU has like much stronger privacy laws than the US when it comes to data and data collection and data dispersion. So Sean doesn't tend to get data from HEMA Ratings, that's like by person. So it's rolled up to protect privacy. But one thing that I was kind of thinking of in order to maybe better capture women's participation is look at the history of a particular fighter and see if they had ever participated in a women's tournament. Because just because they're not participating in the women's tournament at Event X doesn't mean they didn't participate at Event Y.
GW: So you can use the Event Y to inform the data from Event X.
KB: Exactly.
GW: That’s not a bad idea. I've tried to read your articles on Sword STEM and they're fascinating and well-written and to have amusing memes and stuff in them as well. So I would totally recommend that people go and read them. But honestly, just like for you, getting smallsword from a book is maybe not optimal, you'd rather have someone send videos, reading your articles to understand what's in them is not optimal, I'd rather have you explain it to me. So you're using data from Sean’s app to talk about predicting doubles. How do you do that?
KB: As I said in HEMA Scorecard basically the data that is always there is what the outcome of an exchange in a match is. So it'll say whether there were points, maybe penalties or if there was a double. So that's information that we have access to. And doubles are always very interesting to me because people can double out of tournaments. And I've been fortunate enough not to have doubled out of any.
GW: When you say ‘doubled out of tournaments’. Could you just explain what that is?
KB: Oh, sure. So some tournaments have rules that if you have a certain number of doubles, then you both lose the match. And for those people who aren't familiar with what a double is, it's basically you hit each other in the same tempo. So basically you're both killing each other at the exact same time. And some people say that it's worse to lose by doubling out than it is just to lose by not even scoring any points. It’s better to have zero points on the board than it is to double out, in some people's minds. So I just always found this to be a very fascinating idea. And I thought, well, maybe if we could predict which scenarios might lead to doubles, then we can train it out of people or be like, hey, you know, the data says that you're at risk of doubling, so maybe try not to. So that was kind of the idea that I had when I went down this rabbit hole of trying to predict doubles was to see if there's anything that can inform coaching or play strategies or mindset or something of that nature to help prevent doubling.
GW: Is there? Did you find anything?
KB: I was able to find a method that predicts doubles with a decent amount of success. However, the way that the method works is that it's a black box. You don't really know the factors that go into the decision that the predictive model outputs.
GW: Okay. So how does it predict doubles after the fact? Because you have the data after the fight is over, you have the data of all these fights. And you're using that data to say, well, okay, if this fencer and that fencer fence under this ruleset, they are much more likely to have to double out than, say, if this fencer was fencing this other fencer over here. Is that what this is talking about?
KB: Yeah. So basically the way that the predictive model is set up is that, well, the articles kind of looks at doubles in three different ways. So this is across two different articles on Sword STEM. The first article tries to see if we can predict how many doubles will occur in a match and how many doubles will occur in a tournament. And that attempt was completely unsuccessful. All of the models were terribly unpredictive. But we decided to write up the article anyway. And just because it's interesting. It's a good idea. You know, a lot of research really just goes into the successes and so no one really understands what was attempted and didn't work. There's a huge phenomenon in academia called positive publication bias where only positive results get published. So I thought it would be a good idea just to say look, this was tried, it didn't work. But, you know, sometimes even knowing what didn't work is helpful.
GW: Of course it is, it's like when doing research and you’re figuring out how a particular technique in a book is done, it's really, really useful to find out that it's not this way because it doesn't work. So you can just forget about it and move on to the next way of trying to do it. And eventually you find what actually does work. And that's maybe the right answer. Your negative results are least as important as the positive ones.
KB: Exactly. I fully believe that. And that's actually one of the things that I've complimented Sean on is because he regularly writes articles where he says, oh, I did this and didn't see anything. And Sword STEM is very unique in that it has a number of articles where non-results are published. So it's great.
GW: Even these non-results, somebody looking through it with the right kind of data sciencey brain might look at it and get an idea for a different approach that might actually get a result. And they don't have to go through all the stuff you just did because they know that doesn't work. And so it saves them time and effort and makes getting to the truth much faster.
KB: Yep. That's the idea. So those were the first two attempts I did. And then the third one was predicting if an exchange will result in a double. And everything that goes into the prediction is something that you would have at the start of the exchange. So, for example, something that you would know at the start of the exchange is who has what point? What's the point? Have they had any doubles before? What was the outcome of the prior exchange? For example, was the prior exchange a scoring exchange? Was it a double? Was it a no exchange? Was it the start of the match? We also have information on how many matches each fighter had fought before their current one. We know the number of exchanges that have already happened. When I started thinking about this, I was thinking, well, I personally am more likely to double if I'm tired. So just kind of thinking about things that indicate whether or not someone is tired. Like how long has this match been going on? How many fights did I fight previously? Like sometimes you fight back to back matches. So are you in your second match of a back to back situation? So these are the kind of things that you have information on.
GW: Right. And you say it is a black box. What's that mean exactly? Presumably you understand how it works?
KB: No.
GW: Really? Okay.
KB: No. So basically, there are certain predictive modelling techniques where you will know exactly which variables are affecting the outcome. For example, in a regression model, you have an equation that looks something like y equals 0.5 x1 + 2 x2 + 3. And so you know that the point five is the degree to which x1 affects the model and you know that 2 is the degree at which x2 affects the models. And you know any variables that don't make it into the model, they're just not important. So for certain predictive modelling techniques, you will know exactly what affects the outcome. However, for the predictive model that determines if the exchange will end in a double, that's something called a support vector machine.
GW: Tell us about support vector machines because they sound fascinating. And can I build one in my shed?
KB: If you know Python, you can do it right on your computer. But I know it sounds really cool, right? It sounds like a really cool kind of sci-fi-sounding technique.
GW: With cogs and gears and levers and pulleys and things. Definitely.
KB: Yeah. But this is this is something that I feel a visual would help with quite a bit. And I include one in the article. But basically if you think about a graph that has X and Y and there's data points on it, and let's say the data points belong to two groups, in this case, it would be double or not-double, what the support vector machine does is it tries to find a line that separates the two groups on the graph. And if the line is linear, so an actual straight line, then you actually can say which variables affect the model and to what degree. However, when I tried making it use a straight line, it crashed my computer. It couldn't do it. So I had to use a different, they call them kernels. So instead of using a linear kernel, I had to use what's called an RBF kernel. And that one doesn't give you like an equation for the line. So you don't get the point five or the two. You don't even know which variables are being used in the model. So basically you would just give the model the data and it would do its thing and it would output a result. But you have no idea what's going on for it to calculate that result.
GW: So have you actually tried seeing whether this prediction thing works, like at a tournament and then having the data in and seeing whether it will predict whether the next hit will be a double or not. That would be fascinating. I presume it is a percentage. There's a 63% chance of it being a double or a 72% chance of being a double. Is that how it works?
KB: So for certain models, most of the models will output a binary yes or no. And sometimes that's based on like a percentage. And you can set the threshold so you can say, okay, if there's greater than a 50% chance, then yes, you'll output yes, it'll be a double. Or you can fine tune it to say if there's a 75% greater chance, then yes, it'll be a double. But that's not quite how the support vector machine works. It just outputs a yes or no.
GW: Is it actually accurate? I mean, have you tested it?
KB: So I don't remember the exact numbers from the article, but yes, it's accurate. However, accuracy isn't always the best metric to use to determine if a model is successful. So basically, doubles happen only in about 15% of the exchanges. So if you have a model that says everything is not a double because that's what happens in 85% of the cases, then your model is going to be 85% accurate.
GW: I see. Okay.
KB: So there are other metrics such as they're called precision and recall, which basically take into consideration the balance between predicting something as a double versus not as a double. And so basically, the way that I'd explain it is you want to look at all three of the metrics to determine whether or not your model is a good model. And by those metrics, it is a good model.
GW: Okay. So how would you actually use it?
KB: That's the thing.
GW: OR would you not? Is it just like a thought experiment?
KB: Well, this was kind of like me trying to prove Sean wrong.
GW: Proving Sean wrong is very difficult. He's very clever. And he has a lot of data and he knows what to do with it. So, yeah. Good luck. I hope you were successful at tweaking his nose.
KB: The entire joke is that Sean he he's just an engineer and engineers are really bad statisticians. I'm a data scientist, so I know what I'm doing. And just because Sean couldn't make it work doesn't mean that I can’t. So that was the running joke while I was doing this was I'm only doing it to prove Sean wrong because I came up with this idea and I thought, oh, this could be a really cool article. And the first thing he said to me was, this is pointless. I've already looked into it. There's nothing that predicts doubles. And I said, I'll take that challenge. So the idea is, as I was explaining before, that, you know, having a model that's predictive is great. But if you can't learn what to do with it or how to apply it and the application would be, be able to coach certain behaviours under certain situations that show you're going to double then what's the point in having a model? Basically the only thing that you could do with this and I've joked with Sean that he needs to do this, is you build the model in Scorecard and after every exchange it'll indicate whether or not the next one is likely to be a double. And so the table staff during the event just shouts out like, “Look out for doubles!” Because it is that black box, you don't really have insight into what you can do to improve your sword fighting game.
GW: Okay. So it has at the moment, it therefore has a kind of academic interest rather than any kind of practical application for coaches.
KB: That is a good way to look at it. But I still take it as a win because I was able to predict doubles and Sean is wrong.
GW: Sean being wrong is the point. And here's a thought. Would it be possible, let's say you know there's going to be a big tournament, where there’s going be loads of data generated. Collect that data. Find all doubles in the data and then use all the data available before the double occurs, run into the machine. Does it correctly predict the double or not? That would be really interesting to me.
KB: Yeah. I mean these predictions when you use the model, it doesn't take a minute to run the data through the model and get your prediction. It takes a fraction of a millisecond to run it through the model. So it's really fast. However, tournaments also like to run fast, be on time. So I don't know if it would necessarily be a good idea to try to test it out at a tournament in real time.
GW: But you could test it after the fact.
KB: Yeah, you can test it after the fact. And the way that I built the model, it actually kind of does that already because what you do when you're building a predictive model is you split up your data into what they call training data sets and testing data sets. So let's just say you split the data 50/50, your model will learn. It'll build itself on the 50% of the data that's the training data. And then you use the testing data to say whether or not your model is accurate. So basically, if I were to run all of, Combat Con is one of the next big tournaments that's coming up. If I just run Combat Con’s data through the model that's essentially running it through another test data set. I have joked with Sean that if he wants to sit down and watch like ten different matches and call out before the exchange if he thinks something is a double versus not a double, we can record that and we can see if he predicts it better than the model.
GW: That would be interesting.
KB: See if his decade of HEMA experience and coaching does better than my model.
GW: That would be really, really interesting.
KB: Yeah. So if anyone wants to do that to see if they outperform my model, we can have an arrangement, we can sort that out.
GW: Excellent. If the event is running with Scorecard and your model is built into it. If the coaches have access to the screen that the bout is being run on, if they can look over the score keeper's shoulders. There could be a little thing that pops up to say, a double is likely. And they could have a very quick word with their fencer saying, don't you dare double this next time.
KB: I mean, it is certainly doable to build that into Scorecard. But Sean’s busy. We've joked about it but you know, it could work exactly the way that it's intended where the coach goes, you're going to double, better not do it. Or it could just make a fighter more self-conscious and more likely to do it.
GW: It's not actually always helpful to be coached at the time. To my mind, the thing that is most useful in preventing doubles, I find, is getting the fencers to concentrate on controlling their opponent’s weapon. If that is their main objective, doubles are a lot less common because they're not thinking by hitting the person, they are thinking about controlling the weapon first and then hitting them. And you are only having a double when you strike without controlling your opponent’s weapon first. But that's boring, frankly. That's the problem, isn’t it? It works, it's the best way I’ve come across in my 20 odd years of doing this for a living to get students to not double, but it's got no glamour to it.
KB: Yeah, I mean, I think the best way not to double is to just think about offence and defence simultaneously. Just keep in mind that you need to make yourself safe.
GW: Yes, controlling the weapon and then strike. It's the same thing. So you obviously spent a huge amount of time on this predicting doubles thing, do you have any other projects in the work?
KB: Yeah. So I have two articles that I'm working on right now. One of them is actually goes back to the women's tournament aspect. And basically it's like a calculator of sorts where it says like, if you were to add a women's tournament to your event offerings, how many women would you get to sign up? And Sean's actually thinking about adding that to a Scorecard because here's an interesting story. So, Sean, he lives in Michigan. He teaches for a school called Ars Gladii. And they just had their first ever tournament called the AGO Open. And Sean was not going to add a women's tournament to it because it's their first year. He doesn't know how much interest there's going to be. So he wasn't going to add a women's tournament. But based on what I wrote in that initial women's article, he said, well, here's evidence that if I offer it, people will sign up for it. So he did. And it was one of the top ten largest women's tournaments ever.
GW: Wow. That’s fantastic.
KB: You know, he thought it would be a good idea to encourage other event organisers who are thinking, well, maybe you should add a women's tournament. He thought it would be a good idea to show an estimate of if you offered this, how many women would sign up. So that's one of the articles I'm working on. I've already built the predictive model for it. It's just a matter of sitting down and writing the article at this point. So that's one that's in the works. And then another one that's in the works, I kind of want to keep it a little bit of a secret, but I'll throw out the title that I'm thinking of and it's, Do swords make you look cool?
GW: Oh my God. Well, of course they do, everyone knows they make you look cool. That's why we do it, right?
KB: Well, but can we prove it with numbers?
GW: Well, that's an interesting question. How would you measure the coolness added to a person by a sword?
KB: Well, that's what you're going to have to wait to find out.
GW: When is this article going to come out?
KB: I would say another 1 to 2 months.
GW: Now, there was something else in your bio that completely baffled me. And so I thought I would just ask you straight out, what is a ‘trophy guide writer’? What is a trophy? What is a guide? What is a trophy guide?
KB: Sure. So my biggest hobby before I came back to HEMA is video games. I play a ton of video games and I play on the PlayStation platform and they have achievements in the games called Trophies. Basically, they're things like, oh, defeat X number of enemies and you get a trophy or get to this part of the story and you get a trophy. And so if you play games to get these trophies, you're known as a trophy hunter. And so I'm a trophy hunter. I like getting the trophies that makes me feel like I've done everything the game has to offer. It's also a little bit of a challenge because some of the trophies, only like 5% of the people who play this game, get them. So I just really like it. It makes the videogames a bit more engaging. And so I'm part of a trophy hunters community on a website called PSN Profiles. And people in the community will write guides on how best to obtain a game's trophies. So basically it's including strategies. It's including step by step instructions on how to do things. You get screenshots, you get videos of your gameplay, and you organise it in such a way that people can easily follow along and implement your strategies. So I think I have maybe 6 to 10 guides written at this point, and I've won awards for probably about half of them from the community.
GW: Okay. This is a world I know absolutely nothing about and I can't imagine playing a video game and wanting to follow somebody else's instructions as to how to get a trophy more easily. That, to me, would just obviate the entire point of the game. Because I'm not very much of a gaming sort of person, these sort of trophies and whatnot I don't find them physically motivating. So to my mind that it sounds to me like taking something fun and making it into work. But, the thing is, I perfectly recognise I am not representative and it's great fun for me to see that there's this entire kind of subculture that I know absolutely nothing about. So I'm guessing that the trophies themselves, they're like little icons or pictures or something that you get in your game account. There's no cash involved, is there?
KB: So yeah, you're exactly right about them just being like little digital images that are on the game account. Every game that comes out has the trophy list attached to it. And when you unlock the trophy, there's a little noise that plays. And in the corner of the screen, it'll say, you just unlocked this trophy, and then you can go to the digital trophy list and see which ones you have and which ones you still have to work on.
GW: But they're not like things you need to progress in the game, right? It's not like you need to get this special sword to kill this particular monster?
KB: Generally not. There are obviously different types of trophies that you can get in the game. And so some of them you will get without even trying for them and some of them you get just for progressing through the story of the game. But then there are others where it's like what you said, you need to get this special sword and defeat this boss with that special sword, and then you get a trophy for that.
GW: Okay? You don't need the trophy. The trophy itself isn’t a necessary thing that you need to get to a different part of the game, it is more like a just a marker that you've done a particular thing already.
KB: That's correct. Yeah. So it doesn't unlock things in the game. So no, you don't. No one has to go for trophies. But more than likely, if you play a game, you will get some of them just by playing naturally.
GW: And for some people, that's super motivating. It reminds me a lot of sort of habit creation things where people are trying to lose weight or get fit or study a language or whatever, and they're trying to build habits. For some people, having like gold stars on the fridge and they get a gold star if they do the thing that day, that really works. And other people, it doesn't work at all. And it strikes me that what these game designers have done is find a way of providing additional external validation for gameplayers. For the sub population of the game players who are really motivated by these things, this is like catnip. And clearly, you just said there's like communities around it which give out awards for providing guides for helping you go and get your trophies. So what are the awards?
KB: Basically, your trophy just gets a little digital emblem on it, saying that it's an exceptional trophy guide or it won first through third place. You also get a small cash award. We're talking like $25. Nothing exorbitant or anything. So you're definitely like, the number of hours that you put into these guides.
GW: You're not doing it for the money.
KB: You're not doing it for the money. You're doing it because it's something you enjoy. You're doing it to help other people. So yeah, that's the whole awards aspect to it. And you know, as you said, some people find the gold star to be motivating, whereas other people don't. There's a lot of people who start to feel really burnt out and start to find video games not to be fun anymore because they're so focussed on getting these achievements. Some people start to feel like it's a chore or it's a job. And so like this website that I'm a part of, it's a community, they have a forum and at least once or twice a month someone starts talking about how burnt out they feel about the trophy hunting.
GW: Presumably, if you're feeling burnt out about the trophy hunting, you should stop and go do something else instead.
KB: Yeah. There's different kinds of trophy hunters. Like how in HEMA you have the people who want to go to the tournaments and you have the scholars and the people who want to teach. Well, with trophy hunters, you have the people who want to get as many as possible. You have the people who want to get. 100% of the trophies, you have the people who only want to get the trophies that only 1% of people can get. So you have different flavours of people within the trophy hunting community.
GW: Let's say you get one of these very difficult trophies that maybe 2% of the players get. And then you go and write a guide to how to get that trophy. Wouldn't that piss off everyone who's already got the trophy? The whole point of having that trophy is that it's hard to get. And therefore, if you make it easy for people now, 5% of players get it and it's half as valuable to those people as it was before.
KB: So a lot of the times when a trophy has such a high rarity, it's not because there's some secret method and once you learn the secret method, it becomes easy. It's because those things take a lot of skill. So even if you give a method or show a video of you doing it or something of that nature, it doesn't necessarily make it easier.
GW: Yeah. Okay. Like, you know, I have videos of me doing stuff like the punta fasa from Fiore and still most people can't do it because it requires precision.
KB: Yeah, exactly.
GW: Okay, fair enough. Fascinating. All right. You clearly like you do a lot of things, but I am curious as to what is the best idea you haven't acted on yet?
KB: You did send me that and I didn't think about it.
GW: If this creates a big long awkward pause we can just snip it out. That's fine.
KB: Just give me a second.
GW: One quite common answer I get is you act on every good idea you get, so there are no good ideas you have not done yet.
KB: You know, I'm going to say the best idea I haven't acted on yet, but really should, is getting tested for ADHD. I have a lot of anxiety and I have a lot of a lot of things that really fall in line with the ADHD diagnosis. But I haven't gone and gotten officially tested for it yet. So I really should follow through on that because life can be better once you figure out what's going on in your head and getting whatever assistance you need to help through with that. So, yeah, mental health is important.
GW: Absolutely. Several of my previous guests have done that. The one I'm thinking of immediately, because I saw him just a couple of weeks ago in the States, is Kaja Sadowski, who got quite recently diagnosed and got started on one of the ADHD drugs, which is basically an amphetamine, which you'd think would just make you wired and crazy. But just the way he describes it, it's like suddenly all the shit you have to wade through to get anything done just isn't there anymore. And you can just go and do it. It’s your brain, your body, you do exactly what you want. But I would certainly, from the experience of my friends, I would encourage you to act on that.
KB: Yeah, I know. I mean, I have a therapist. I have a psychiatrist for my anxiety medication. And talking with my psychiatrist, she suggested that I go get this ADHD thing evaluated. And, you know, it's a really good idea. Just need to follow through with it, because as you said, it can really help when you get medications or figure out habits to help you. So like getting on medication for anxiety, I am a much better person for it. So, if things can be better by getting that sort of assistance, then you should do it.
GW: And I have a friend, it’s not ADHD, but he had basically panic attacks and things, really severe kind of work related, nervous breakdown anxiety. I mean, the ambulance people had to carry him off a train because he was basically curled up in a corner and couldn't move. And he's like, you know, if it was my liver that stopped working properly, I would just go to the doctor and it would be completely fine. And my brain stopped working properly and the doctors came along and now I'm doing these things and this medication and everything's a lot better and there's a whole lot of really unnecessary, I don't even understand really why it exists, but this is like stigma around well my brain isn't working quite right so I go to the doctor and the doctor gives me this, that or the other and now it is a lot better. If we were talking about your lungs or your liver, it would just be a no brainer. But because we have this sort of this weird cultural thing around mental health and how, oh, well, you should just muscle through or just calm down or take a deep breath or whatever. We are kind of steered away from actually just going and getting professional intervention when we need it.
KB: I mean, I think a lot of the times it's because it's something that you can't necessarily see, like your arm is broken, you can see your arm is broken, but you can't see that your brain is broken. There's like an entire discussion around invisible disabilities. Like, for example, some people who have seizures, they can get the disabled placards for their car to be able to park in the handicapped spot. And some people, when they get out of their car, they're not on crutches, they're not in a wheelchair. Nothing looks wrong with them. And so I've heard stories of people who go and say, that's a handicapped spot. You're not supposed to park there. And it's like, well, I have a handicap, you just can't see it. So I think that's where a lot of the stigma is around mental health disorders is you can't see it. And being that it's a mental thing, a lot of people think, well, if it's just a mental thing then I can mentally get through it, I don't need anything tangible to get through it.
GW: And sometimes some mental health things are fixed by just talking to a therapist or there are other interventions other than chemistry that that can work in some cases. But yeah, I'm in big favour of drugs in all of their forms.
KB: I mean, if it helps, why not?
GW: Exactly. Okay. So, the best thing you haven’t acted on is going and getting tested for ADHD. I think that's an excellent one to go and act on. My last question is, if you were given $1,000,000 to spend improving historical martial arts worldwide, how would you spend it?
KB: You know, because I'm a data person, I would say definitely anything to do with data collection. I think something that could be really interesting is to be able to record everyone's fights or be able to record practises and be able to hire people to document things that they see going on in these videos to get data points that don't get recorded. And something like HEMA score card, like for example, being able to say left handed people tend to do better in fights against right handed people. Well, you don't collect data on left handed people. You don't have that data point. But if you're watching a video, you can see that they're left handed. So I just think anything that would improve data collection and data analysis could be very insightful. So I think that's where I'd put my money.
GW: Okay. I'm struggling to see how you would collect that data.
KB: You would just have video transcribers who watch the videos and look for certain things within those videos.
GW: What would you do with the data?
KB: You could use it to help coach or train people. Like, for example, there was an article recently written by Steven Cheney for Sword STEM, where he did an analysis of what I think he calls the practise games. But basically he had two people. He had all of his students in pairs fighting against each other. The one person could only use one particular guard and the other person could do whatever it is they want to try to break that one guard. So in the end, he found that pretty much if you are relegated to just one guard, no guard is really better than any other guard. So I think that's a very interesting thing to think about because now you don't have to have some misconception that, oh, it's better to be in vom tag than it is to be in plough or ochs or something of that nature. So I think it will help swordfighters become better competitors if they have data that shows or that can elucidate certain guards or cuts. Or, should you be aggressive? Should you be defensive? Anything of that nature, I think would be very useful just to help people with their training, to help instructors and coaches with what they want to teach in their classes.
GW: I've often thought that I should have gathered a lot more data when I was teaching over the last 20 years or so, because there would be patterns that I would be able to see, like to be able to maybe anticipate when say someone who has been training a couple of years, is going to have patch where they're going to get frustrated and possibly that their chance of quitting goes up. But if I can just get them through that patch, then they will settle in the next couple of years of training or whatever. And so to be able to head off trouble before it starts. That might be useful, but I'm not a data person, so I didn't collate it. I have loads and loads of attendance data for my school and I've done absolutely nothing with it ever.
KB: Well, so, you said you would like to prevent people from getting frustrated and quitting. One project that I've worked on at my job was with nurse attrition at hospitals. So trying to figure out if there are any indicators that a nurse is going to quit their job. Nurse turnover is a huge thing in the U.S..
GW: Well it's very, very hard work in very, very difficult conditions for not very much money.
KB: Very hard. Very difficult. Very thankless. You don't get the glory like the doctors do. And yeah, they're just burnt out. They're overworked. So trying to figure out what kind of markers there are for a nurse to leave are very important because maybe you can improve those conditions. For example, one of the things that we found was an indicator was how many hours they have worked in the last six months. You work a lot of hours and a lot of overtime, you're going to feel burnt out.
GW: You are more likely to quit.
KB: Want to leave. So figuring out something like fighter attrition or what causes a fighter to want to leave HEMA. If you collect the right data, you might be able to figure it out.
GW: Okay. I think getting people to transcribe videos would be a very laborious way to do it. There must be there must be a simpler way.
KB: I mean, if anyone has any ideas send them to me, send them to Sean.
GW: We don't give out guests’ emails and things on the show because privacy. But if anyone wants me to forward stuff to you, they can send stuff to me and everyone can get my email from the show. That's fine. And so, yes, listeners, if you have ideas that you want to be sent to Kari, just send them to me and I'll send it along.
KB: Yeah, absolutely. You know, there's also the Sword STEM page on Facebook. You can ping through that. But yeah, if anyone has data or has an idea that they think could be explored analytically, I'm not going to promise that I'll look into it. But I don't know what everyone else finds interesting to look at. I mean, if I don't have the data then I can't do anything with it. So Guy, you just said you had attendance data. I can't think of anything that could be done with that off the top of my head. But I might be able to think of something.
GW: Okay. Brilliant. Well, you heard it here first, folks. So if you have an interesting research question or perhaps a data set to ask questions of, then send them to me. I will send them along to Kari. She doesn't promise, but she might take a look. Brilliant. Well, thank you so much for joining me today, Kari. It has been great talking to you.
KB: Yeah, thanks, Guy.
