Paul Hollywood's Got A Type: Using Count AI to Predict British Bake Off Winners - Final Cut

I guess, normally, we would take this opportunity to introduce the show and the person. I'm hoping that most people watching this will actually know not the show, which also has a working title. Honestly, maybe the show should be working title. This is a mess. Ollie Hughes, father of the year, boss of the month, CEO at Count. Ali, how are you? Spoiler your ridiculous flattery. I'm fine. Thank you. I feel like our subject today is something that you and your children maybe bond over. The Great British Bake Off, or as as it is known in America, The Great British Baking Show for reasons that I can't understand. And I'm proud to be an American. I don't know why. It's certainly playing my my a hidden passion of mine, having seen almost every season, and I cannot stop myself. Like, every I'm like, I'm not gonna watch this time, and I do. I bake twice a year for each of my son's birthdays. And what do you think? Just literally funny enough, literally this week, I baked a pizza cake. So a cake which then looked like a pizza. I guess the answer to this next question is is pretty clear. Do you do you think that going on the show, you could have some success? Definitely not. Because I have a full time job. So your job influences your ability to have success on the show, does it? I have a strong hypothesis that if you are retired or a student, that helps you a lot. Your crude hypothesis is very close to mine. It's actually the perfect segue. I do have a dataset. I have a dataset with every contestant. I have some demographic information about them, including their occupation, as you mentioned, and then some success metrics. Do they win the technical? Do they win the star baker? Are they a finalist, either runner-up or winner? Knowing that this dataset exists and that we're about to look at it, what is your very first thought? What comes to mind? I think there's also a weird weird thing about STEM people, people who do, like, science or are in, like, engineering. Like, the precision piece of this makes a difference, so I think there's something about that. Like, I remember Raul, if you're if you're Bake Off fan, Raul was insane. I can literally cut a number of butter like that. I love it. The way it just goes down your throat is amazing. He was like, come on the way on the way he did it. He worked in a, you know, chemistry lab or worked in a university, and he was crazy good. Like, I think he was one of the best bakers ever. And he, you he started baking a few months ago, he's basically just doing chemistry experiments in a in a in a in a kitchen. If you've got a bit of cultural heritage, like, you're, like Your your, like, family's from a different country than the UK, and you actually have can pull on experience or find recipes from other parts of the world, like just a bit of extra, you know, spice, literal, to put in your cake, that might help. I guess I did write the word success metrics, but I don't think it's a coincidence that you and I maybe look at the Great British Bake Off through the lens of like business terms. Is this you're making you're making an excuse for why you looked at Great British Bake Off data Yes. Trying to bring it back to b to b I'm trying to justify my salary. Correct. So thinking about all these contestants in the tent, they're all vying to win that strange glass cake platter. There are some parallels. Right? Like, winning star baker in a dataset is could be seen as just like another success metric like a closed deal. Contested demographic relationship to success. Right? Where people live, what they do, exactly what you were talking about is very similar to our customers, how they might look at the demographics of their customers, of their customers, and how that relates to Yeah. I see great. Exactly. And how that relates to conversion. Ideal customer profile for those of us who are more baking fans than b to b SaaS fans. Right. For sure. Or if you're really into the insane clown posse. Woo woo. Are baking skills really the only predictor of your ability to win? Like you said, there are some demographic ones, but I wondered if there were some that were less predictable. So maybe it is where you live or where you were born. Perhaps it is even the number of letters in your name, and there's these, like, weird unconscious biases that Paul and Pru have. And then generally speaking, can we see statistical correlations between the data that we have, demographics, also the things that you win, like technical versus Star Baker, etcetera, and getting to attend that ceremonial circus that I find extremely strange at the end of the at the end of each season. Okay. So, Ollie, as promised, this is your data. Take a look at it. Poke around. Maybe if you would wouldn't mind thinking out loud about some questions you have. Just a I might ask you a few questions, actually. Hey, Count AI. We've got a dataset here. I'd love you to help me understand what correlates with the final result or people being a winner. Could you help me understand how many if they're if they're winning technicals, if the number of technicals they win influences their winning streak? And can you also help me understand what makes you most likely to win based on your age? You can dictate in count. Clearly, Ollie's not just typing that quickly as he's Shouting at you. I do wonder, do you do you think you you talk to Count Count AI more or less than your mother? That's a very difficult question to answer a public show. Alright. And but I definitely speak to My mother. Definitely my mother. Also, my English accent's amazing. Alright. Did you did you send it? Is it going? It's going. Alright. I'm thinking away. I mean, I asked it a few different questions, so to be fair, it's gotta do a bit of work. Yep. Alright. So let's let it think. Let's do the I'll take you through my analysis. So you come down here, let me just walk you briefly through how I made this dataset. So you're right that someone on the Internet had fourteen seasons worth of data. For whatever reason, they separated it into fourteen individual tables. So that lives somewhere else on the can oh, you're finding it. So, yeah, if you follow the connectors, you can see all fourteen of those datasets. And Wow. Instead of joining those myself, which, you know, I I could have done, I did ask the agent, and I'll I'll pull open this agent here. And I basically said, hey, like, here's the structure for the first two. Join those. Let me confirm that it's correct. And then if so, join the rest. It did that. We have the dataset that we were looking at earlier, a hundred and sixty eight contestants, and also consolidated things like whether they were a runner-up or a winner. So if you come down here, if you recall the original question was, like, what are the biases? You mentioned age. You mentioned occupation. I was like, gender. It's always gender. You know? Like, unfortunately, it's always gender, but when you come down here, turns out British Bake Off literally Fifty fifty. Fifty fifty on the nose. But another thing that I noticed when I was just sorting through the data is Richard from season five is the only person, unless it happened with Jasmine, I don't think it did, to win Star Baker five times. Do you do you remember this man? I do. Yeah. He didn't win, though. He didn't win. Oh, yeah. He came second. Yeah. He didn't win. He win Star Baker five times and came in second. That is that is just disrespectful. Think he I I don't know if I don't know, like, I'm sure someone's gonna come tell me I'm just gonna go this wrong. I think he, like, had a really bad final. He must have. Like, it wasn't like, it was, like, he really kind of had a bad day. He the bed. You're I don't know how you're gonna feel about this, but when I saw Richard and I and I and I looked closely, I thought to myself, okay. This guy maybe looks like he shops at the same places as Paul Hollywood. That geographically, they might be, like, bumping into each other at the same Brooks Brothers. And then I thought, okay. Where does Paul live? Paul lives in I'm probably gonna say it wrong. Cheshire? Cheshire. Yeah. Lil' Paul Lil' Paul grew up and I believe still lives in Cheshire. And so the next thought was, okay, maybe this is a proximity bias that Paul nor Prue know about, that if these contestants live near where they live, their chances of winning go up. I went and found where Paul lives and where Prue lives. I then added their latitude and longitude as data points, and then I found all of the contestants, like, where their hometowns are as longitude and latitude points. And then within count, you can easily plot that on a map, which I have done. Turns out that there is, like, no correlation in in proximity. I asked the agent, hey. Like, I gave it the same thesis I gave you. It's all about proximity. And I said, but before you actually run the analysis, can you tell me what your plan is? The reason I did that was I was afraid that it would do the pure, like, oh, there's more people within this location, therefore, that means success, without taking into account the population in London or the number of contestants that are bunched in a certain space. And it did, without me telling it, group contestants into bands. So if they're within fifty kilometers, that's one group. If they're within fifty to a hundred kilometers, that's another group and so on. So down here, if you see success metrics by distance, in short, proximity to a judge means almost nothing. Like, nope. Not at all. In fact, there's you can see there's this weird skew if you're, like, far away but not too far away, a hundred to two hundred kilometers. But it seems kinda arbitrary, doesn't it? Yeah. It's basically noise, isn't it? Exactly. So that theory was put to rest. But then I followed up with the agent and said, hey. Instead of proximity to Paul and Pru, can you do proximity to one another? And this is sort of my way of being like, are contestants who win often in the same places, are they frequently within major cities? You know, it's that kind of analysis. It found, I'll just read this off to you, roughly one to five other contestants within a hundred kilometers of them, so, like, semi remote, have the highest winner rates of eleven point five four percent, whereas very isolated contestants, and there aren't many of them, but they have never won, the ones who live really far away, like, way up in the north or in Ireland. So while the majority of contestants, a hundred and twenty of them out of a hundred and sixty eight, came from high density places like London, these only produce winners at an eight point three percent rate, which the agent has dubbed this middle ground the Goldilocks Zone, which is great editorializing from a robot. Right. So what we're saying is, so if you don't live in London, if you live outside of major city cities, it's the kind of idea here, live in the suburbs, you've got more Exactly. You're not you're you got less for life. That was David's theory. Right? It's like the city slows down where you are, and so the patience it requires to bake well is like a part of your daily life. You buy that? Comes back to, like, they just have a they're less less busy so they can just actually crack on with baking during the week. This kinda leads us to this this final analysis, and then we'll pop back up to see what your agent has come up with, is what are the other possible demographic indicators, and can we formulate a perfect profile for a contestant? At the start of the show, if they check all of these boxes, do their chances of winning go up, up, up? Part of me was like, is there just some strange bias outside of these demographic markers that we can latch onto? Turns out your chances of winning, your chances of getting a handshake, and your chances of winning Star Baker, and Star Baker and Winner are obviously very tied together, but go way up if your name has between seven and eight letters. Really? As in first name? Correct. Correct. Granted, we probably need a thousand years' worth of Great British Bake Off data in order to find something statistically significant. Do you wanna hear David's theory on this? Is that if you have a longer name, you're more likely to get a nickname, and that makes you more endearing. Wow. That's interesting. Could be. Interesting. I wonder if it was more, like, more cultural that you just got back to that kind of you've got more influences to play with because you haven't got a name like Tom. Right. Again, kind of a nominal thing. Whereas exactly what you predicted, those that are homemakers and retired win the show at the highest rate. They win Star Baker at actually, not the highest rate. It was the other group you said win Star Baker at the highest rate. Really? So what? The academics? Academics and technology professionals like Raul. I knew it. I knew it. Right. So if you know how to do science, you can bake, basically. Right. If you can use a pipette tube, you can Yep. Do it. The last bit is all about age. And I don't think you predicted this. Let's see if your agent actually found the same thing. It should. But, essentially, if you're under thirty four, your chances of winning go way up. Yeah. Which I won't reveal Don't know whether they reveal either of our ages, but I don't like my chances. No. Not me. We're past our baking primes. That's that's there. Until until we retire. But that that that that will make till we retire. Well, even then, it's not great, is it? Oh, it's it's not ideal. Not for winning at least. In any case, we have our perfect profile. This is a person whose name has between seven and eight letters. They are in academia or they are in technology and young. And if they're not those things, they're retired or they're a homemaker. And then, ideally, they don't live too far away from major cities, but they also don't live in major cities. So that makes sense. So like so Jasmine is like the perfect candidate. He's seven. She's under thirty She was She was a student in a science. Wait. I think I feel vindicated that my apologies are generally correct. I think so. Alright. Let's go find out. Where'd you put him? Oh, there it is. Put up here. I think that's right. I think we can see the age profile we discussed here that to get a star baker to be a runner-up or winner, you're younger by quite a way. It's the average number of technical wins by final results. So since as a winner, you did win more technicals, but not by much than a runner-up. So it's like you need to have two technical wins on average to do well, but you don't wanna look I looked over here. It quite interesting that you you don't want too many star bakers That's so interesting. At all. Like, you wanna have what looks like one or two, maybe three, but you don't want four. No one's actually won the game who's had four or five. Like That's so weird. Our friend, our friend before you mentioning Good old Richard. Richard. This makes this makes sense. I think I think I feel like I feel happy. This would be this this fits my mental model of the show. Okay. Good. And we now know that we are never gonna be bakers of the Great British Bake Off. My my my pizza cake's a mess I'm gonna do. Your pizza cake would stop every show. If if there was a pizza for your son's themed week in the tent, you would succeed. There you go. Wow. I I'll say it's so good. I'll you're missing some crust there, but it's so good. This was done with my son, to be fair to him. He did a good job. To be fair to him, I mean, you got a star in the making. And listen, if he if he wins, you win. So maybe we can just nudge him toward a career in homemaking. Yep. Get him into sciences, start him young, make sure that we're too close to poor Hollywood. This is awesome, Josh. Good work. You too. So if you're watching this in a place where you're able to leave a comment or a question, please do. Otherwise, you can you can contact us. Visit count dot co or you can even reach out to me, josh at count dot co. And for Ollie, hi, I'm Josh. And I guess I'll say thanks and and take us offline.