Sep 9, 2026 · Prediction markets · Desk briefing · Written by GammaPredict Research · Snapshot date 8 Sep 2026 · All news

DefiLlama TVL is not our ranking

People paste a DefiLlama screenshot into Slack and call it “the real top 10.” We published the same snapshot on the ranking page so you can see the number next to the editorial list. They measure different things. Mixing them is how Rain and Predict Fun get sold as Kalshi substitutes.

Updated 9 September 2026: Rain, PredictStreet, OPINION, Sport.fun, Liquidity House, Pascal, Overtime, Trueo and Limitless Exchange are watchlist listings after the new Llama pull. They are still not Gamma Score top-10 substitutes for Kalshi or Polymarket.

The 8 September pull

Source: api.llama.fi/protocols, category Prediction Market. 121 protocols in the dump, 78 with a TVL figure, category total $427.4 million.

• Polymarket International — $361.1M, +6.0% over 7 days, Polygon. 84.5% of category TVL. Same product family as our Polymarket review.
• Rain — $29.6M on Arbitrum in that 8 Sep snapshot ($28.5M on 9 Sep). Watchlist row now — still not a flagship review.
• Predict Fun — $11.3M, BNB + Blast. Watchlist only. Story: why it is not a Kalshi alternative.
• PredictStreet $4.4M · OPINION $3.5M · then a long tail.
• Azuro — editorial flagship, Llama row about $1.30M in the same dump. That is not a contradiction. Protocol TVL is not sports-book handle.

Kalshi and Polymarket US show up in some Llama UI sorts with empty TVL. Empty is correct: a CFTC cash DCM does not lock dollars in a named pool the way a Polygon market does. Sorting that table by “rank” in the UI is not sorting by capital.

What TVL is good for

Locked collateral, pool size, a public time series. If you are asking whether an on-chain venue has any capital behind the screenshot, Llama is one of the few honest feeds. We say so on Trust: liquidity is 25% of Gamma Score, and Llama is one public signal for on-chain names — not the rank.

Worked example. A $11.3M Predict Fun TVL is real money in listed pools. It is about 3% of Polymarket International. It is not “top 4 prediction markets in the United States.” It is not a second DCM. It does not tell you KYC, geofence, or whether the UMA dispute path matches the contract you clicked.

What TVL is useless for

• Regulated cash venues. Kalshi and ForecastEx will look small or blank. That is a category error, not a quality verdict.
• Fees. TVL does not print the taker line. Use the fees comparison.
• Access. A pool on BNB can be huge and still be the wrong legal product for a US cash reader. Access and KYC.
• Resolution. Capital in a pool does not name the settlement source. Resolution briefing.

Why Gamma Score did not move

We listed Predict Fun on the watchlist after this snapshot because omitting a $11M venue was a catalog hole, not because Llama outvoted the desk. Rain is a watchlist listing as of 9 September 2026 — still not a flagship review we would put above Kalshi. Azuro stays a flagship because sports liquidity and protocol design are the job of that page, not because its Llama TVL beat Polymarket — it does not.

If a thread says “#1 on DefiLlama so #1 on GammaPredict,” send them to the table on top prediction markets and to the liquidity guide. The first is the editorial list. The second is how we talk about depth. Llama is a footnote with a date stamp. When we refresh the JSON, the date on that section moves. The Gamma Score formula on /trust does not.

Editorial analysis. TVL moves daily. This briefing is tied to the 8 Sep 2026 snapshot, not a live widget.