How The Political Machine lets you roll the dice on Kamala Harris vs. Donald Trump | Brad Wardell

The Political Machine 2024 is the latest political strategy game out from Stardock Entertainment for anyone who wants to see how the U.S. presidential election is going to turn out this year.

And in recent weeks, the simulation of the election has undergone major changes as Kamala Harris replaced Joe Biden as the Democratic nominee running against Republican Donald Trump. As a result of the updates and changes in the race, the game sold 27 times as much in August as it did when it debuted in May 2024, said Brad Wardell, CEO of Stardock, in an interview with GamesBeat.

Americans are disturbed by politics and many would rather avoid it. But it is rather unavoidable and many of these same Americans intend to vote in the upcoming election. And that tells you about the mixed feeling people will have about playing games about politics.

I played The Political Machine 2020 and I enjoyed revisiting the remake of the game for this new political season. I played a couple of rounds of the 2024 presidential campaign strategy simulation game on Steam, first as Joe Biden and then as Donald Trump. And now I’ll have to play it all over again, given how different the race is turning out to be. Stardock has had to stay on its toes, tweaking the simulation with the best data it can muster about how voters are feeling, where they stand on the issues and other election possibilities.

This game reminds me of my civic duty and the fun I had playing the old Parker Brothers board game Landslide from 1971. I first learned about presidential politics in that game, and you can do the same with your family with this title, either in a single-player game or in multiplayer. And this new game gives me an excuse to talk about one of my favorite side hobbies: politics.

I interviewed Wardell of Stardock in Plymouth, Michigan, about the creation of The Political Machine 2024 and the major updates the company had to launch as Trump dodged losing a lot of time to legal trails and Harris wound up replacing Biden as the Democratic nominee.

Here’s an edited transcript of our interview.

GamesBeat: Where are you based now?

Brad Wardell: We’re still in Michigan, just outside Ann Arbor. I always used to enjoy going out to the bay area, though.

GamesBeat: I was curious about Political Machine and the insights you get every few years when there’s a presidential campaign happening.

Wardell: The modeling technology gets better every time. This election season, though, has been insane. All of our carefully planned modeling had to get tossed out the window and redone. In April we thought it was going to be easy. We knew both parties, both nominees early on. It was the guys who ran last time. It was going to be so easy to do. Then it all had to be redone.

GamesBeat: What did you have to do? Did you have to create a fairly new game?

Wardell: We had a bunch of models on how different demographics react to different decisions. Biden had the advantage of being in public life for a long time. We knew all that data. For Kamala Harris, there just wasn’t very much of that. There was a lot of, “Well, I think her position is this, but–” We’d try to get the numbers. One of the things that’s been different about Political Machine versus a game based on polling is that it models enthusiasm. It’s always been based on how different demographics vote and figuring out how enthusiastic those demographics are in a given election season.

GamesBeat: How sophisticated are you compared to other kinds of simulations that are really meant to be simulations, as opposed to just for fun?

Wardell: If you run AI versus AI, it’s pretty easy. We get good results with that. In the past we’ve almost always gotten, if not every state–I remember last time, in 2020, we took some grief. It said that Biden was going to win Arizona and Georgia was in play. We were told that obviously we were wrong. We usually don’t know exactly why it says something like that. It could be a data error. Maybe we put in the wrong number. Georgia wasn’t supposed to be in play, historically. The fact that it was, according to the model–well, why is that? It had to do with enthusiasm in Georgia, who came out to vote.

During COVID we had to do a lot of shenanigans, so to speak. It came down to figuring out–it’s a game at the end of the day. The model isn’t running on a supercomputer. But what percentage of the voting base can vote remotely, can use mail-in ballots? As soon as you put in those numbers, that’s when a lot of these states started to flash into the colors we didn’t expect.

GamesBeat: Did you do a significant update for the game as Kamala came into the race?

Wardell: We had to go and basically build her, so to speak. There are different levels of data we put in for the election. You can run as Jimmy Carter, but we don’t put the same level of effort into figuring out how different characters will perform. We had to do a real pass on her and see–one of the challenges this time, different states have different mail-in ballot rules. Pennsylvania, for example, they’ve already started taking mail-in ballots, and it’s only September. When we made the original game we didn’t have to think about that kind of thing.

At the level of the code, it means the enthusiasm threshold that triggers someone to vote in a certain demographic is different. If you have 90 days to vote by mail, your threshold to put in the effort to vote is a lot lower than if you have to get up and drive to a polling place on election day to vote. That’s been a big challenge this time.

Probably the biggest thing we’d hoped to get in, which we don’t get in this time, is a bunch of AI stuff we’ve been working on. I was hoping–if they could just delay the election a little bit, we’re building an SDK that lets us have a micro-AI model that can go up to the cloud. We’ll obviously have that for the next one. We want to put it in all of our titles. I was really hoping to get that in this year.

GamesBeat: Do you still have a scenario where Biden runs? I wonder if that’s affected by things like poor debate performance.

Wardell: We still do manual updates on Steam and elsewhere. When we do an update, we tweak numbers like that. The enthusiasm for Biden is a bit less. The amount of energy he has to campaign is a bit less. When we did the update–obviously there are no guarantees, but there was no scenario where he was going to win. I can’t imagine what the background polls–these guys have some pretty sophisticated polling out there. I imagine they showed the same thing. Certain states just became out of reach for him.

There are certain states he has to win. A Democratic or Republican candidate has to win certain states. Michigan, for example, went off the board for him after the data went into the sim. Kamala Harris coming in puts Michigan back in the Democrats’ favor. That makes it a lot more interesting. It’s harder to predict who will win now.

GamesBeat: I did a playthrough with Biden where the Trump trial had a dramatic impact. I forget which trial it was. But it took him out of campaigning for two weeks or so. That stopped him from influencing particular states at critical times. Was that too big a thing for the Democrats to be able to play?

Wardell: The expectation, when we put it in–sometimes we’re just trying to predict the future, and we’re often wrong. There was a case, the one where he got convicted of various felonies, and he was supposed to be sentenced in July. We predicted this would have a significant impact on his ability to campaign. That didn’t turn out to be the case. It’s looking like nothing will happen between now and the end.

The bigger issue now that affects the model is going to be things like the upcoming debate. We have no idea how that will play out. The enthusiasm threshold for Kamala Harris is on the edge. The support for her on the issues is weak, but the enthusiasm is pretty strong. Her scores on various issues aren’t great, outside of a handful, lower than Biden, but her score is much higher than President Biden on the enthusiasm. If she loses enthusiasm she’s in trouble. You can imagine it as simple math. You have your score on the issues that animate certain voters in our demographics, and you multiply that by their enthusiasm score.

Ironically, Trump’s problem is the opposite. On issues he does pretty well. But people are so anti-enthusiastic about him at times, he has a problem where he generates enthusiasm for whoever is running against him. It’s a little like Magic: The Gathering or Hearthstone. You do five damage whenever you swing, but then they get two back. He was the first candidate since we’ve been doing this where we had to implement a reverse enthusiasm concept. The other candidate gets enthusiasm just for running against him.

GamesBeat: It becomes a more interesting game now that Kamala is more competitive.

Wardell: Definitely a lot more competitive. We will almost certainly have to do at least two more updates before release. Certainly one after the debate to adjust the enthusiasm. In July we were getting ready to say, “This is what the map is going to look like.” We were confident. When you do it at the state level–predicting that California will go blue, you don’t really need an advanced simulation to figure that out. When you look at Wisconsin and Pennsylvania and Georgia and Arizona and Nevada, that gets a lot more interesting.

GamesBeat: What kind of impact does social media have in terms of the need to go to every state and campaign in so many spots?

Wardell: It’s definitely made it harder. When we first shipped the game it still had newspaper ads and radio ads. Remember radio? This box that had knobs and stuff. That’s made it a lot tougher from a game mechanics point of view. It’s much more complex than in 2004. This time around we introduced the political action cards. The candidates do a lot more stuff now than they used to. It’s not just, “Here’s our one message that we pound on.” There are lots of things they have to do all the time.

We have a political action card in the game that says, “Target opponent is considered weird.” You play that on them. For a period of time it lowers enthusiasm for them, or lowers their ability to generate enthusiasm. When the world changes in one way you get it back in other ways. That’s where the political action cards come in.

A lot of people ask us why we don’t localize this in other languages. It’s because we would have to do so many updates during the course of a campaign season. Every couple weeks we’d have to run all this new stuff through translation. We could do it with AI, but it wouldn’t be very good.

GamesBeat: Do you have big things happening with your other games?

Wardell: We just released Sins of a Solar Empire II. That’s done far better than we ever anticipated. We’re working on another game that makes use of our micro-AI tech that’s not shipped yet. It doesn’t generate writing or art. It’s way more grognardy in terms of–you’re asking it questions. Here’s a set of data. Very small models that can be run locally. We think that’s going to be a big deal coming up. We want to do much more immersive stuff.

When you want AI characters to do interesting things–you could have A talking to B. All we have to do is give it a bunch of data. How does he respond? Is he angry, happy, sad? You can get all kinds of emergent gameplay out of that.

GamesBeat: Back to Political Machine, was there much change that might happen if Harris selected Tim Walz as a running mate versus another option?

Wardell: There was a candidate in Pennsylvania that we’d already 3D modeled. We were so certain it was going to be the governor of Pennsylvania that we already had the artists do the whole thing. Minnesota seemed like a pretty safe state for Harris, while Pennsylvania was a better choice, at least purely from a video game perspective.

GamesBeat: Do you feel like other sources of prediction should pay attention to this?

Wardell: Well, you can look at our past performance. We predicted that Trump would win in 2016. That was an outlier. We were shocked, but it comes back to the enthusiasm effect. In 2004 we were wrong about the election, but we got every state right except Ohio. We were pretty sure Ohio would go for John Kerry. Ohio was close. It could have gone either way. We were right other times, but those were too easy. No one was shocked when Obama won both times. The one time we were wrong and it was close was 2004.

GamesBeat: What is the sales pattern like for this game in a year like this? Do people tend to buy it right before the election, or just whenever you release it?

Wardell: Usually it comes in two waves. When we release the game, we don’t usually sell very many. In fact, relative to the rest of the game industry the model is completely messed up. Normally, if you sell 100 units in your first month on a typical game, the next month will be 24 units. The month after that you take whatever you sold in the previous month and multiply by .6. We have so many models on game sales, it’s crazy.

But this game, it’s almost nothing until the primary season heats up. This year we didn’t have a primary season, so it started around the conventions. To give an idea of how extreme this is, our sales went up 27 times in August over where they were in, say, April. That’s obviously not normal for a game. With most games, it’s instantly up and then down. With this game there are two big humps. This month will probably be less, and then October it will jump again to a crazy number. After November it goes down normally.

GamesBeat: Does it make sense to invest more heavily in the prediction or simulation model? You mentioned AI.

Wardell: That’s the thing we want to do in the future. We want to rely more on AI. We have a lot of knowledge we could use to train an AI. Then we wouldn’t have to manually update a giant data structure that we’ve had going on for years and years. Some of this stuff is pretty obvious. Eventually we could have something where it just scrapes the news in real time. Right now we still have to update it on Steam. I suspect there will be another election in four years, so we’ll have our chance.

This year is probably the biggest update we’ve done to the game since we originally released it. We added the political action cards, which changes the gameplay dramatically. We updated the graphics engine. The older versions of the game always looked really bad. Very indie. Now it’s starting to look like a real game.

GamesBeat: Is there anything you do in the way of marketing to get it noticed more in election season?

Wardell: We’re teaming up with some of the other companies on Steam that do political games. We try to do a big combined sale with those games in October. We’ll do a publisher sale in October. I can bundle almost anything with Sins of a Solar Empire right now and get sales going. It’s been a crazy summer for Stardock. A lot of our focus coming up is going to be in the area of on-chip AI stuff.

GamesBeat: How do you expect that that’s going to change things for you?

Wardell: It is going to allow non-player characters in the game to feel a lot more intelligence. The kind of AI we’re working on isn’t generative in a sense. It’s not writing. It’s not creating art. It’s AI in the old-school sense. I have a decision tree. I have five things I can do. What’s the most reasonable thing to do? The AI tends to respond with one, two, three, four, and then five, or however many choices we give it. It just generates a number. In the old days you’d have to have a guy like me go in and roll dice, if you ever played Dungeons and Dragons. AI allows us to have much smarter weighted dice for that kind of thing.

GamesBeat: What are your biggest games to date? How does Political Machine fit into that?

Wardell: On the game side, the biggest one by far is Sins of a Solar Empire. I could be wrong, but that might actually be the best-selling strategy game ever for the PC. Political Machine is much smaller. But in August Political Machine outsold Galactic Civilizations IV, which was shocking. There’s been so much political news lately that we probably shouldn’t have been surprised by skyrocketing sales. Galactic Civilizations continues to be a pretty good seller for Stardock, though.

We have another big game that’s about to come out. I can’t talk about that yet. Give me about four weeks and I can talk more. That’s going to be very interesting. Our biggest product by far is in development right now. It’s about 60 people with contractors. It’s a god game, if you ever played Populous and things like that back in the day. The world is very simulated. Lots of AI to try and make the people behave like real people. How would people behave if there was an actual god messing with the world?

GamesBeat: When do you plan to make the last big update for Political Machine?

Wardell: There’s supposed to be a debate this month, September 10. It’ll take us a few days there to gauge reaction and update. Probably about a week after that we’ll have another update. We’ll do a final one in October, in case of any October surprises we need to take into account.

GamesBeat: Does the topic generate a lot of discussion in the studio around politics?

Wardell: It’s funny. The Stardock staff, we’re all over the place politically. We have people who are pretty far to the left and we have people who are pretty far to the right. It’s all very jovial. These are people who’ve known and worked with each other for decades. There’s always a concern around making sure the game is even-handed, evenly abusive to both sides.

GamesBeat: It’s good to see that you can help with political education.

Wardell: The game has been used a lot in schools over the years, in universities. Every season we get orders for classes. This time, we added a feature where you can simulate the primaries. In multiplayer you can have five people playing together in one party and see how they do. It’s really interesting when people play the game, seeing how quickly they will change their positions to win. “If I ran for president, I’d do…” Well, then you’d lose. You can’t afford to lose certain states. “I wouldn’t support farm subsidies!” Then you’ve lost the primary.

GamesBeat: Is there anything you have to do now to get ready for four years from now?

Wardell: Not until after the election. There are two or three big things we do. When a new census comes out we have to pay close attention. One thing that’s been interesting is the consistency of aging demographics and how they evolve. I’m biased on this stuff. The data ends up revealing my biases. The baby boomers, when they were younger they were more to the left, so they’ll stay that way, right? Nope. Age is much more predictive of how people will vote than anything else. I never would have thought that would be the case 20 years ago. But age is a much stronger predictor.

GamesBeat: How careful do you have to be about the question of bias? At some point you have to make judgment calls and assign weights to certain things. How important is a particular event or characteristic?

Wardell: This time around, because we did such a major rewrite–the biggest issue that comes in when it comes to subjectivity, what’s the enthusiasm around a given issue? Very early on, in alpha, it was apparent that some of the subjectivity that went into those issues was completely messing up the model. People who are really into politics, they think these issues here are the ones that everyone cares about, but when it comes to what the mass of people vote on, it’s a lot more humdrum. People vote on things like jobs, the economy, Social Security. They’re not voting based on trans rights, Black Lives Matter, or whether a certain book is allowed to be in a school library in Florida. These issues that get all the media coverage don’t affect the outcome of the election.

When you get new people working on a project like this for the first time, it’s hard to get over that. It’s discouraging for them. “Everyone really cares about this, right?” No, that doesn’t really matter on election day. You may wish it mattered. The exception is abortion rights. That one can move the needle. But even that issue–it might affect Congress, which we don’t have to worry about in our game. Donald Trump and Kamala Harris are not going to win or lose any given state just because of that issue. That’s why you always see candidates trying to scare you about Medicare and Social Security. Those issues will affect the election. You can win or lose states on that.

GamesBeat: Can you get to predicting whether the House or the Senate will go red or blue? It seems like adding House races would be a massive undertaking.

Wardell: That would probably be outside of our ability. You get into the size of the market that would be interested in something like that. Elections at the local level, fortunately, don’t have a lot of impact on the presidential election. Their ability to raise money and affect turnout might have an impact, but they’re so dependent nowadays on national political trends. Which has to be frustrating. Voters only care if you’re a Republican or a Democrat. They’ll vote based on how they feel about national issues. It wasn’t always that way, but it is now.

GamesBeat: How many people work for you now?

Wardell: Just over 100. We’re very busy at Stardock now. Contrary to the rest of the industry, Stardock has been hiring as fast as we can across the board. Everything from artists to engineers to AI and everything in between. We’re out in the midwest. There aren’t a lot of studios in the midwest. We’re starting to pick up a lot of people who don’t want to live on the coasts. Maybe they want to move back to be closer to their family. That’s been helpful for us.