Collins Coalition

View Original

Scrabbler Q&A: Team Woogles!

Welcome back to our Scrabbler Q&A series! I'm excited for this one. This past week brought big news in the Scrabble community - Woogles.io, a revolutionary new platform for online word gaming, was officially released to the public.

To get the inside scoop on Woogles, I sat down for an interview with... the entire team. Below is a chat with all seven members of the Woogles squad - Cesar Del Solar, BriAnna "Lola" McKissen, Josh Castellano, Jesse Day, Conrad Bassett-Bouchard, Doug Brockmeier, and Will Anderson.

EVANS: I thought I'd start by asking about all of you guys and what you each do for the team. Could we just go around the group and could each of you briefly summarize what role you've played in building Woogles?

CESAR: Yeah, so it's my brainchild, but I wouldn't have been able to get nearly as much of it done without the help of everybody else. I've been wanting to have a platform for Scrabble players - and not just Scrabble, but word games in general. Like, one big platform. I remember I saw lichess.org back in 2015, and I was like, I want to make that site, but for Scrabble. And then this year, once the pandemic started, we were like, OK. Tournaments are moving online. We need to actually build this.

It started as an AI project, really. I didn't have the thought of building a platform for playing Scrabble at first. It was really just Jesse and me - we were like, "Let's build an AI." We were working on it on and off, and we made some good progress. Then Conrad reached out to me, and he said "Well, you'll need a UI [user interface] for that AI. So we said OK, Conrad can design the UX [user experience] and the UI for this. At that point, we were like, "Why don't we just build a platform where everyone can play?" The rest is history.

LOLA: I'm a front-end engineer, which means all of the buttons and actual things you can see on your screen, I built those. And we've had a lot of extra contributors helping with that over the last couple weeks, which has been really great. Andy Kurnia from Singapore especially has been contributing a lot of stuff.

But yeah, I just happened to see Cesar post once on Facebook asking for someone with exactly my skill set that wanted to help build the site, so I was like, "Yeah! I'm in."

JOSH: As for me, before this, I had been doing a lot of work on RandomRacer - pulling stats from cross-tables, and showing all of this stuff, though not in a very pretty way. And then when I heard Cesar was doing this thing, I thought it would be better to just consolidate all our resources and our knowledge.

I mean, even before, I had been thinking - not actually doing it, because I don't have the skills, but thinking - about how it would be nice if we had just one place for everything. All the stats, a place to play, everything, it's all in one place. That would just make everything better. So that's why I hopped on, and I've been working on back-end stuff - the ratings, the stats, and a lot of back-end work on tournament functionality. So that's my deal.

JESSE: I do everything I can to let our builders build stuff unabated and give them direction. Eventually, I'll build some stuff of my own.

CONRAD: I have always wanted to design a thing for Scrabble. Scrabble has basically never had anything professionally designed for it since its inception. So I figured, yeah - I've only been doing this a couple of years, but let's give this a try and see if I can design a thing from start to finish and hopefully not screw it up too bad. So far, it's just been a ton of fun.

EVANS: And then, Doug and Will, you guys joined the team recently?

DOUG: Yeah. Like, about 10, 15 minutes ago.

EVANS: (laughs) Well? What have you done so far in your 15 minutes?

DOUG: I don't do anything! Will, as you know, has his own stream which is super professional and cool and informative and helpful...

CONRAD: Will, do you wanna explain what Doug does?

WILL: Doug and I are going to be kind of a two-headed monster in most ways. The only way that I'm distinct is that I have my Twitch stream. That's kind of an independent project that we're going to use, however we can, to get the word out about Woogles being such a better platform.

But in all other ways, the two of us are going to be focused together on just making sure that people that are already aware of online Scrabble play get the word that there's an amazing new option available to them that's definitely better than anything they've tried before.

EVANS: So, if I could start from the beginning: I remember when you guys first announced this project, you had an introductory statement on your website that said something like, "Let's face it, the status quo is pretty bad." I was wondering if you could elaborate on that a little bit. What were some of the shortcomings you guys saw in online word gaming, and what was your plan to improve on them?

CONRAD: Should I take that one? Because I wrote that.

I would start by saying that, like, things have never really been designed for Scrabble players. No one's ever really done any sort of research to understand what Scrabble players want. You can design this gamified crap - which works, and it's addictive! - and that's Scrabble Go. You can do that if you want. You can have these little side projects, that are amazing, that create things like ISC, or everything that Cesar's done, or everything that [John O'Laughlin and Jason Katz-Brown] have done [with Quackle].

But things are usually designed by the person who's making them, for themselves - and to the best of their knowledge, what they think other people might like. What we're trying to do is to take all of our collective knowledge of what it's like to play Scrabble online and design something that makes sense and is intuitive and also teaches people. So you're going to see things throughout, like numbers of vowels and consonants remaining, or other little hints like that. This sort of app can teach people. It can also power people up, like if you're playing blitz and you want to use shortcuts. We're trying to design something that works for both high-level players and people who just play casually. We're really designing for both.

LOLA: So for me, I think I'm the oldest person on the team. I remember 2002. In 2002, we had this huge burst when "Word Freak" came out. A whole bunch of people came out and said, "Ooh, I want to build stuff for Scrabble. I want to build tools. I want to build Zyzzyva. I want to build ISC." And all of these things suddenly appeared, and they were super cool! For 2002.

Things kind of died down after that. And everything that's out there right now, you know, it's still stuff that's left over from 2002. We really haven't seen any of the advances that Conrad talks about, like how to build a user interface, make it to Scrabble at all. I think that's the biggest thing - there's been a lot of research done into how to make a game easy to play, how to make it engaging for users, make it fun for users, and that just isn't part of our world yet.

JESSE: Yeah, I mean, we've seen during COVID that chess and chess streaming have gained major popularity. We think that with this, if we clean it up, we can expand the reach of Scrabble beyond what existing platforms have.

Also, another thing I think about is the journey of expertise. This is Josh's point, but I really like it too. The tools are so scattered right now! Why shouldn't they all be in one place? My dream is that a future champion from Pakistan, or wherever, finds our website and has everything they need to go from rank novice to expertdom.

JOSH: Also, I think it's worth mentioning: There are some technical deficiencies, mainly with ISC, that could impede the growth of the game and the legitimacy of the game. Namely, that you can see your opponent's rack and you can even draw your own tiles if you finagle the code enough. These are unacceptable. They're really, really bad. It's not the way it should be done at all.

What we've done, or really I should say what Cesar has done, with his socket server, and with his knowledge that really not very many people have, is the right way to do it. If we want this to be something legitimate that happens online all the time, then this is the way it needs to be done. You can't have people drawing their own tiles. It's egregious.

EVANS: So... you've already begun rolling Woogles out to your Kickstarter backers, right? How's all of the initial testing going so far? Are people liking the site - and are they coming up with any good suggestions to improve it? How's all of that going?

CESAR: It's been great. I think every time we've done testing, the app has become more and more stable. I remember at the beginning, it was bad. Like, it was just crashing. We would start games, and then all of a sudden they'd just stop, and I'd be like, "Oh, crap, I need to restart the server." It's become much more stable now. People can play many simultaneous games. When we got the beta testers in, we were playing like, 14 simultaneous games and everything was fine.

Overall, the testing has gone well. We're getting a lot, lot, lot of feature requests and suggestions and bug reports, across many different channels. People post on Facebook, they post in Woogles chat, they post on Discord, and so on. But we're handling them!

Andy Kurnia, this kid from Singapore, he's fixed like 30 bugs for us. There are all these little things that are annoying, and he just goes and fixes them. I think somebody today, for example, said they were playing with Caps Lock on and they accidentally played a blank instead of the actual letter. Within 20 minutes, we had a bug fix from Andy for that. 

LOLA: And it had to have been 3 in the morning his time.

CESAR: Oh, yeah! He doesn't sleep. He just builds stuff for us. It's pretty amazing.

It's really nice. I really feel validated to have that support, because it means we built something that people like. Why else would people devote so much time to our app? I mean, look at you, Evans. You joined, and right away, you were like "I'm addicted to hyper-speed with bots."

EVANS: I've played over a hundred games in 4 days! It's incredibly addictive.

CESAR: Yesterday, we saw Mack Meller play a series against the bot, and it was like 70 or 80 games in a row. He just kept losing! It was so funny.

JESSE: Real quick, I want to shout out some people. We're open source, so anyone can contribute. Andy has been awesome. Martin DeMello built our bot, which everyone is now addicted to. Martin is amazing.

Also, I'll shout out the MGI crew. We asked them to run a tournament on our platform as a testing thing. The first time we had them use it, it was still pretty broken, and they caught a lot of big stuff for us. We were like, "Damn, we didn't have a logout button, because we're all attached to our computers by an umbilical cord and we're not used to logging out of stuff." So shout out to MGI for giving a lot of their time and resources to help us.

EVANS: As I understand it, your goal is to create not just a place for people to play games, but also a platform for studying - is that right? What's your vision for that, and how is that aspect of the project going so far?

LOLA: That's something that we'll probably get to next year. The idea that we have is there'll be like, a learning mode. Like, if you're playing against the bot, it will be able to give you hints and ideas here and there. Not, "You should play this word," but maybe a certain area of the board will light up and tell you, "Maybe this is somewhere you should be looking to score." Or maybe things like telling you, when you play a word, that there are 5 front hooks for it. Not necessarily telling you what they are, but just kind of giving you hints about that information in a way that we've never had before.

It'd be great to be able to teach people the strategy of the game. We've had lots of cool word study tools like Aerolith and Zyzzyva, but actually having something that can teach you board vision, for example, that's something that doesn't yet exist. We're really excited about creating it. Probably next year.

CESAR: I think about puzzles as well. We can have the bot generate puzzles and have the user try to find the best play, or the highest-scoring play, or whatever. Or endgame puzzles, or just regular mid-game puzzles. Like if there's a star play with a lot of overlaps, or anything tricky like that. We could have categories. That's something that I've been thinking about.

EVANS: I should also ask about the AI. I've already got some firsthand experience there... I've already played against the bot about 125 times, and my record's like, 25-100. So I can say from experience, Macondo's already pretty good! Could you walk me through the process a little bit? What goes into creating a Scrabble AI, and where are you guys at in that process?

JESSE: To be totally frank, that was what Cesar and I got together over a year ago to work on, but we basically put it aside when Conrad got involved. That was the point in time when we really unified on a vision for what needed to be built next. We decided that what we needed most was just a better place to play. That was the biggest thing holding back the growth of the game - more so than building a great AI.

But in terms of the AI: We can use machine learning to build the best Scrabble AI there's ever been. We're already not that far away from having a Quackle-level Championship Player. It's just a matter of time. Quackle is a great tool; it's also 15 years old. There's no reason we can't harvest machine learning to build something better.

CESAR: So, I built the code that actually plays the Scrabble. And Jesse, really I just had him crunch numbers for weeks. Months, even. I'd be like, "Here, Jesse, here's a batch of 3 million games, give me some numbers." And then we would iterate, and so on. We did that for months. We eventually figured out a way to calculate superleave values that are actually better than Quackle's. When we matched them against Quackle, they actually win a tiny bit more - they have an edge of 50.5 percent to 49.5 percent. But it's consistent! It's well outside the margin of error.

So we've definitely found some basic values. And I think we can take a similar approach to figure out pre-endgame strategies and endgame strategies. There's a lot of things we can build that would make our AI even better than Quackle's. But I think machine learning is what's really going to give us a big leap forward - figuring out stuff like the shape of the board, and with the tiles left in the bag, what places are dangerous. You know - that kind of stuff. Being able to look at the board and know how to play. Or like, knowing what tiles your opponent might be holding based on their last 2 or 3 plays. I think that's the kind of stuff we're going to explore more next year.

JESSE: This is a big project. It's not going to end next year. We're going to be working on this for a long time.

EVANS: One last question to close things out. So you've got the site ready to unveil to the whole world, and that's super awesome. So... what's next for you guys? Do you have any near-term goals you're going to try to achieve moving forward? What are you focused on now?

LOLA: We've got a few things on the immediate roadmap.

We want to flesh out the chat features a little better so you can, you know, talk to your friends back and forth and have an ongoing chat. Right now, we've got the lobby and we've got in-game chat, and that's about it. We'd also like to have tournament stuff in place so that it's easy for people to have a separate chat channel for tournaments. But also, more importantly, we want them to be able to pair people, have matches set up, show live standings, et cetera.

JOSH: Think chess. Are you familiar with chess.com or lichess.org?

EVANS: Why yes, I am!

JOSH: Think Titled Tuesdays. That's the ultimate vision. We're obviously going to do the classic stuff, like King of the Hill where everyone gets paired. But we also want to do stuff where you can just pop in and play, and you don't have to worry about where everybody is - it's just an arena. That'll be a really big feature, and we hope to work on it soon.

JESSE: Also, I'll just add: There will be some announcements on the horizon soon about tournaments. Like, real-world tournaments, happening on Woogles, and soon. So stay tuned for that.

EVANS: That is very, very exciting. I can't wait.