ROLR and the Gap Between a Full Arena and a Thin Order Book

On a morning in Incheon, I read the interview with Seth Young — a former professional CS2 player, now CEO of ROLR — and stopped at exactly one sentence. He said the esports betting market in the United States is “not there yet”. What made me stop was not the content, but the timestamp: he said the same thing seven years ago. Seven years is long enough for a flawed xG model to be discovered, repaired and rewritten from scratch. It is also long enough for a market either to mature, or to prove it has a structural ceiling.

In those seven years, esports arenas in North America stayed packed. Tickets kept selling out. Streaming platforms kept logging millions of viewing hours per event. But trading volume on esports prediction boards remained on a completely different order of magnitude from professional basketball, football or baseball. A full arena, a thin order book. That gap is the subject of this piece.

Two concepts merged into one

To read the story correctly, one has to separate two concepts that are routinely merged. Traditional sportsbooks — DraftKings, FanDuel, Fanatics — sell fixed odds and answer to state gaming commissions. Prediction markets — Kalshi is the clearest example — let users trade event contracts and answer to federal oversight. ROLR chose the middle ground: no head-on collision with the fixed-odds giants, but an event-trading product built for esports.

Seth Young is not an outsider. He competed professionally in CS2 before moving into operations. Before ROLR, he ran High Roller — an esports betting product — in markets he himself describes as “not nearly as strong as the United States”. Spike Up Media, the current partner, is both a large shareholder and the user-acquisition engine. That is the structure of a company that does not burn money to buy growth, but buys growth in measurable units.

The central question is concrete: if the U.S. esports audience is that large, why does trading money not match it? And can a company with five years of positive returns in weaker markets replicate that result in the largest market of all?

Three layers of data to separate

The first layer is audience size. That is a stock, not a flow. An arena packed with 15,000 people is a stock figure — it measures interest at a single moment. Trading volume is a flow — it measures repeated behaviour week over week. The two do not convert into each other automatically.

In football, I once got this exactly wrong. In March 2026, I built an improved xG model for Ulsan Hyundai; the model predicted a 2-0 win over Jeonbuk, and the match ended 1-3. Three weeks of auditing the pipeline exposed an encoding error in the “key passes” variable. K League 2026 taught me this: pioneers do not fail because they look far ahead, but because they look far ahead and miscount one column of data.

The column that may be miscounted in the ROLR story is the product itself. People measure viewing hours, fan counts, ticket revenue. They rarely measure the conversion rate from “following” to “placing an order”, and they almost never measure order-book depth by hour. A market only truly exists when there are participants on both sides of the order. An audience only stands on one side.

The second layer is return on ad spend. ROLR says it has five years of positive returns with Spike Up Media in markets weaker than the United States. That is valuable data, but it must be read with conditions attached. Positive returns in a market with little competition, few substitutes and high margins do not guarantee positive returns in a market where four major brands are competing for the user’s wallet. Customer acquisition cost is the most sensitive variable in the whole model, and it depends more on competitor density than on product quality.

There is another problem with that five-year record: it was collected in markets ROLR managed to survive in. Markets it did not survive in contribute nothing to the dataset. This is classic survivorship bias, and it systematically makes the return figure look better than reality.

The third layer is how money is spent. Seth Young’s language is technical: “surgical”, “measurable”, “getting our fair share” rather than “taking the whole pie”. That is the language of someone who has paid the price for a wrong campaign. In esports, blockbuster marketing deals rarely produce a deeper order book — they produce one week of spiking traffic and then silence. A trading platform lives on users who come back next Tuesday, not on users who clicked an ad during launch week.

The arithmetic of “fair share”

One thing corporate language tends to hide needs to be said plainly: a share-taking strategy only means something if the pie exists at a certain scale. If total esports trading volume in the United States across a season amounts to roughly one night of professional basketball, then a third of it still cannot support a company with a full operating structure.

That is why “not there yet” matters so much. It is not a warning; it is an input parameter for the entire valuation exercise. Seth Young is telling investors that the denominator is still small, and that his company is built to outlast that denominator.

The paradox of patience

Here I have to argue against myself.

Based on my experience following matches and transfer cycles, a statement repeated verbatim over many years can carry two entirely opposite meanings. Meaning one: the speaker is consistent and the market genuinely needs time. Meaning two: the market has a ceiling the insider cannot yet name, so they repeat the old line to keep expectations safe.

From the outside, I do not have enough data to distinguish between them. I will say it plainly: I cannot verify it. Markets do not move on news. They move on the gap between two reports.

There are three competing hypotheses for that gap, ordered by how verifiable they are.

The legal hypothesis: esports betting has no unified regulatory framework across states, so the product cannot scale. This is the easiest to verify — just track legislative progress in the large states.

The product hypothesis: esports fans care about match outcomes but do not want to put money on them, because the esports community carries a deep culture of suspicion toward betting — a direct consequence of past match-fixing scandals. This is the hardest to verify but carries the greatest explanatory power.

The liquidity hypothesis: users arrive, see a thin order book, find no counterparty to match against, and leave. A thin order book is not a symptom of an immature market — it is the cause of the market not maturing. This is a self-reinforcing loop, and it fits the data best.

The blind spot sits on the human side

There is one variable no dataset in the interview can measure: player motivation.

During the 2026 pandemic, I ran a study across 200 matches in K League and Bundesliga while stadiums stood empty. Home win rates fell from 45% to 38%, while average goals rose from 2.4 to 2.8. I wrote an 8,000-word report proposing an indicator called the Pressure Index. That report measured the effect of crowds on on-pitch behaviour. It could not measure what happens inside a player’s head when they know the stands are empty. I once thought I was reading the map of the match; it turned out I was only looking into a mirror reflecting my own fears.

With the esports betting market, the same question appears: if a segment of the fan base considers betting on their own discipline an act of betrayal toward the community, no user-acquisition campaign fixes it. Customer acquisition cost will not fall, and it will not fall for technical reasons.

I have seen this shape of data elsewhere. In 2026, when Son Heung-min suffered a hamstring injury and diagnoses pointed to eight weeks, I built a regression model from injury data on 47 European players from 2026 to 2026, and it produced a significantly shorter “recovery window”. The model was right about the number. It said nothing about what that player endured to return early. The same limitation applies to every model forecasting betting-market liquidity: behaviour can be measured, will cannot.

Signals to watch in the coming cycle

Three concrete indicators.

First, order-book depth by hour, not total volume. A platform can report rising volume because of one major final, while the order book for group-stage matches stays empty.

Second, regulatory progress in large states. If New York, California or Florida opens a clear framework for esports prediction products, the addressable market could grow exponentially within a quarter.

Third, ROLR’s own customer acquisition cost. If that figure rises by more than 30% without a matching return, the “surgical” model has hit its ceiling.

My conclusion is not a prediction. If the U.S. market matures within 24 months, ROLR sits in the right position with the right cost structure. If it does not, their caution itself is what keeps them alive longer than competitors who burn cash. Neither branch is bad. The only branch that worries me is a third one: the market fails to mature for human reasons, and nobody in the data room is measuring that.

Applause in an empty stand is not noise; it is a signal from a future we have not yet been brave enough to index.