Predictions should reward accuracy, not luck

Aditya Kapur is the co-founder and CEO of Gradient, a prediction platform that pays users based on how close their forecast is to reality, not whether they were on the right side of a line. He started the company while studying business economics at UCLA, after watching friends lose money on sportsbooks and prediction markets despite making smart calls. We talked about how Gradient works, why he thinks the rest of the prediction industry is built on a flawed assumption, and what he's learned launching into one of the most crowded categories in consumer fintech.

What is Gradient, and what made you want to build it?

Gradient is a precision-based prediction platform where users place predictions on quantifiable real-world outcomes, ranging from sports and finance to politics and entertainment, and are rewarded based on how close their prediction is relative to everyone else in the contest. Unlike traditional sportsbooks, there's no house taking a rake or profiting off user losses. Unlike over/under betting or yes/no prediction markets, outcomes aren't resolved in a binary win-or-lose format. Gradient is built around forecasting the actual outcome itself. The closer your prediction is, the better you perform.

The idea came after watching a lot of my friends lose money on sportsbooks and prediction markets. What stood out to me was that a lot of the time they weren't even making bad predictions. They were just losing because the system only rewards binary outcomes. If someone predicts a player scores 29 and he scores 30, that's a strong prediction, but on most platforms that can still be a complete loss if the line is 29.5. That felt broken. Gradient was built around rewarding accuracy, not just being on the “right” side of a line.

What were you seeing in the market that made you think predictions need to work differently?

The biggest thing was realizing how similar the current platforms are once you strip them down. Sportsbooks like DraftKings and FanDuel are built around spreads and over/unders. PrizePicks and Underdog simplified that into prop entries. Prediction markets like Kalshi and Polymarket built tradable contracts around yes/no outcomes. Different formats, but the same binary structure underneath.

The issue is people don't naturally think in binary. If I think a player scores 31 points, or inflation comes in at 3.2%, that's my real prediction, not just “over” or “under.” Current platforms flatten that into a simple right-or-wrong system, which misses the quality of the forecast itself. We thought there should be a better way to measure that.

Who are you building for, and what does a successful user experience look like?

We're building for people who already love making predictions and think they're better than the average person at it. That could be sports fans, fantasy players, bettors, finance people, political junkies, really anyone who likes forecasting outcomes and testing their judgment.

A successful experience on Gradient is when someone makes a prediction, sees the result play out, and immediately understands how they performed. Not just whether they won or lost, but how close they were, where they ranked, and how they compare to the rest of the field. Over time, that builds into a real performance history, which gives users something current platforms don't really offer: a measurable track record of how good their predictions actually are.

The prediction market space is crowded. Where does Gradient fit, and what sets you apart?

We're definitely entering a crowded space, but we're structured differently than the existing players. Sportsbooks like DraftKings and FanDuel are house-backed, so users are betting against the book. DFS platforms like PrizePicks and Underdog still revolve around beating preset lines. Prediction markets like Kalshi and Polymarket are built around buying and selling contracts tied to outcomes.

Gradient is a peer-to-peer contest platform. There's no house taking directional risk and there's no trading market. Everyone enters the same contest, submits their forecast, and payouts are based on who was most accurate. That changes the whole user experience. It feels less like gambling and more like competition, where your performance is based on how good your prediction actually was.

One of the clearest examples of that difference is Gradient Chains, which is our version of parlays. In a normal parlay you're stacking binary outcomes and one missed leg wipes out the whole entry. With Gradient Chains you're stacking multiple predictions, but instead of needing every leg to hit exactly, your payout is based on how close you were across the full chain. That changes the experience a lot. It rewards consistency and precision, not just perfection.

That's really where Gradient fits in. It borrows familiar mechanics from sportsbooks, DFS, and prediction markets, but the core model is different. Instead of betting against a house or trying to beat a preset line, users are competing directly against each other on prediction quality. The better your read on what's going to happen, and the closer your forecasts are to the end results, the better you perform.

How have you gone about growth entering this space? What strategies have worked?

So far it's been very community-driven. We've focused on the places where prediction behavior already exists, mainly Reddit, X, Discord, Instagram, and sports communities. The good thing about this market is we're not creating new behavior. People already spend hours debating games, props, politics, and market outcomes. We're just giving them a better format for acting on those predictions.

Content has been one of the strongest channels for us, because prediction content naturally drives conversation. Hot takes, stat lines, and controversial forecasts create engagement fast. Right now the focus is less on scaling aggressively and more on learning what drives repeat behavior, what users understand quickly, and where the product creates the strongest retention.

What's been the hardest part of building so far?

The hardest part has honestly been category definition. Because Gradient doesn't fit neatly into sports betting, DFS, or prediction markets, people usually try to compare it to one of those right away. A lot of the work early has been figuring out how to explain the product clearly enough that people understand why it's different. Once people understand it, they think it's so much better.

On the product side, building the scoring and payout system has also taken a lot of iteration. Rewarding accuracy sounds simple, but creating a system that feels fair across very different categories like sports, politics, finance, and entertainment is harder than it looks. And on top of that, balancing product, legal, growth, and fundraising all at the same time has been the biggest adjustment as a founder.