TS Imagine, the risk-and-order-management vendor formed by the 2021 merger of TradingScreen and Imagine Software, has quietly wired prediction market probabilities into its institutional risk platform. The signal is not exotic — event-contract implied odds sitting alongside vol surfaces and factor exposures in the same risk blotter a portfolio manager already stares at. What matters is not the feature. What matters is that a vendor whose client base includes hedge funds, asset managers, and bank prop desks has decided the probability curve coming out of a retail-heavy prediction venue is now a data layer worth paying for. That is a shift in what counts as institutional-grade signal, and it deserves reconstruction rather than applause.
What TS Imagine Actually Shipped, Stripped of the Press Release
Let us strip the announcement to what the pipes do.
The delivery is a data feed inside the RiskSmart / TradeSmart consoles — the same panes that already carry Greeks, VaR contributions, and factor exposures. The new column is an implied-probability read on named event contracts: political outcomes, macro-print thresholds, Fed decision paths, and a handful of geopolitical binaries. A portfolio manager who wants to see how a "Fed cuts in December" contract is trading can now render that number in the same row where a EUR/USD 25-delta risk-reversal sits.
That is the whole product. Not a trading rail. Not an execution venue for the event contracts themselves. A read-only market data layer.
The distinction is doing more work than the launch copy admits. A read-only feed means TS Imagine is not on the hook for order routing, best execution, or reg oversight of the event-contract exchange. It is on the hook for one thing — that the number displayed at 09:31:22 reflects the mid, or the last, or the composite as documented. Everything downstream — hedging, position sizing, PM commentary — is client-side.
The Probability Feed: Where the Numbers Come From, and Where They Do Not
OK, here is where it gets interesting, because "prediction market probability" is not a single object.
An event contract that settles at 100 or 0 has a last trade, a best bid, a best offer, an order-book midpoint, and a volume-weighted average over some window. Each is a defensible "probability." Each gives a different number on a thin book. On a contract with $80,000 of resting depth and a two-cent spread, the difference between midpoint and last can be four vol points on the underlying implied probability. That is not a rounding error. That is the entire alpha of the read.
We have not seen TS Imagine publish the composition methodology in a public technical note. The reasonable assumption — and it is only that — is that the feed exposes both a raw venue midpoint and some volume-conditioned filter. Serious risk consumers will need the latter documented before they can put the number into a factor model with a straight face.
The other absence worth naming is the venue count. Kalshi and Polymarket price the same politically-shaped contracts, sometimes with double-digit basis-point gaps, and the arbitrage between them is capital-constrained rather than free. A composite feed averaging two venues is a different animal from a single-venue feed. Which one TS Imagine ships determines whether the number is a market read or a house index.
Anatomy of a Risk Spread: Vol Surface vs Event-Contract Implied Odds
Let us take one contract and unpick the spread the way a rates desk unpicks a swap.
Say a "Fed pauses at the December meeting" contract trades 62 bid, 65 offer, on 800 lots of depth each side. Midpoint says 63.5% probability of the pause. Same afternoon, the Fed Funds futures curve implies a pause probability of, call it, 58% — pulled from OIS math on the December contract.
Five points of spread. Where does it come from?
Peel the layers. First, the venue's raw markup — the spread the market maker on the event contract is charging to warehouse binary risk into settlement. That is 300 bps of the 500. Second, the liquidity premium — the fact that you can hedge Fed Funds futures instantaneously in size, but you cannot exit the event contract at midpoint on 800 lots without moving the book. Call that 150 bps. Third, the reg premium — the event contract's counterparty risk against a CFTC-registered DCM is not identical to CME cleared-and-margined futures. Small, but real. Fifty bps.
That leaves zero for "the crowd knows something." Which is the point institutional users need to internalize before they treat the probability as signal. Most of the "predictive edge" being sold in the retail narrative is, when you strip the layers, a liquidity and reg premium, not a Bayesian posterior. TS Imagine's contribution is putting the reader in a position to see that decomposition rather than take the top-of-book number at face value.
Why the Buy Side Asked for This Before the Sell Side Did
The sequencing here is not accidental.
Hedge funds and family offices — TS Imagine's buy-side book — have been running unofficial event-contract feeds through internal scrapers for two years. Every macro discretionary desk we have spoken to during the reporting window mentioned in passing that a junior on the desk had a browser tab open to Polymarket during CPI print mornings. Not for signal. For context — an anchor point on how the retail-plus-crypto crowd was positioned into the number.
Fieldnote: two macro PMs described the workflow identically. Print lands, they check the ES reaction, then they check whether the event contract moved *before* the ES did. If yes, they mark it as a leading-indicator moment. If no, they mark it as retail catching up. Neither of them called it a model. Both used it.
The sell side did not push for this integration because their prime brokerage franchises do not touch these venues yet — the counterparty setup around Kalshi and Polymarket is not compatible with a standard bank-cleared workflow, and reg treatment is unsettled. Buy-side clients could ask because they own the risk end-to-end and route through their own custodians. The vendor is following the buy-side signal because the sell-side committee that would have blocked this five years ago is not the buyer here.
That is the interesting shift. Institutional data budgets used to be gated by sell-side data governance. They increasingly are not.
The Historical Precedent Nobody at the Launch Mentioned
There is a version of this that ran once already, in a different asset class, and it did not end where the participants expected.
In the late 1990s, the Iowa Electronic Markets and, briefly, several CFTC no-action-letter political futures venues were the object of similar institutional curiosity. A handful of macro shops — the public record around LTCM's positioning discussions and the published memoirs of several 1990s macro traders reference this in passing — pulled feeds from political-outcome venues to overlay against currency positioning ahead of the 1996 and 2000 US election cycles. The academic literature that followed, particularly the Wolfers and Zitzewitz working papers of the early 2000s, argued the venues were informationally efficient in ways sophisticated participants underweighted.
What happened next is the part worth remembering. The venues did not become institutional infrastructure. They stayed thin, stayed retail-adjacent, and the information edge — such as it was — got competed away by the very act of institutional attention. The academic finding proved right and commercially uninteresting simultaneously.
We are not saying that arc repeats with Kalshi and Polymarket. The venues are larger, the contract set is broader, the regulatory posture is materially different — Kalshi is a CFTC-designated contract market, which is a legal category the 1990s venues did not occupy. But when a vendor integrates a feed and the pitch is "institutional-grade signal from a retail-heavy venue," the historical record says the signal quality tends to compress the moment institutions can act on it at scale. Something to price into the subscription cost.
Red Flags in the Data Layer That Institutional Users Should Interrogate
We would ask the following questions before wiring this feed into a live risk model.
First, what is the tick-level latency between the venue's matching engine and the number in the RiskSmart pane? A one-second lag on a thin binary book is enough to make the display number a fiction during the moments that matter — election-night hours, Fed-print minutes, geopolitical breaking-news windows. Vendors quote average latency. Insist on 99th-percentile latency during volatility.
Second, how are contracts with settlement disputes handled in the historical file? Prediction markets have a resolution-committee layer that occasionally reverses or delays a settlement. If those events silently rewrite history in the backtest data, any strategy calibrated on that file is calibrated on a survivorship-adjusted reality. Ask for the pre-resolution and post-resolution files separately.
Third, is the feed reconciled against the venue's own end-of-day settlement print, or is it a live-only stream? The distinction matters for any client running overnight VaR — if the risk system's close print does not match the venue's, the exposure roll is broken from day one.
Fourth, what happens on venue outages? A binary contract with a stale price during a market-moving hour is worse than no price — a stale 50/50 can look like reality until it is not. Documented failover behavior is the difference between a data product and a data hazard.
None of these are unique to this integration. All of them are the questions that get skipped when a launch is exciting and the pipes are new.
Fieldnotes From the Rollout Window
Fieldnote: three institutional consumers we spoke to during the reporting window had already been told by their vendor rep that pricing is "at parity with equity index data." Two of them laughed. The pricing gap between index data and event-contract data is the entire commercial question — one is CME-cleared and continuously two-sided, one is not, and pricing them the same is the vendor testing what the market will accept.
Fieldnote: the RiskSmart integration slots the new column between VaR contribution and stress-scenario deltas. Small UI decision. Large implicit statement about where the data sits in the mental hierarchy — next to the risk numbers, not next to the alt-data flags. Reads like a promotion the data has not earned yet.
Fieldnote: we asked one PM who had trialed the feed for two weeks whether she had ever executed a trade *because* of the event-contract probability. She said no. She said she had killed two trades because of it. Which is the more institutional use, and probably the honest measure of what the feed is worth in Q4 2026.
Fieldnote: the phrase "prediction market" appears exactly zero times in the CFTC's Part 40 rulebook that governs designated contract markets. The regulatory category that would house these venues formally does not yet exist as a named object. Every institutional integration ships before the regulatory language catches up. That is not new. It is worth noting.
FAQ
What data does TS Imagine's prediction-market feed actually deliver to risk consoles?
Based on the launch materials, the feed pipes event-contract implied probabilities into RiskSmart and TradeSmart panes alongside existing risk metrics. It is a read-only market data layer, not an execution rail. The client sees a probability number keyed to a named event contract; hedging, sizing, and downstream P&L handling stay on the client side. What is not yet publicly documented is the composition methodology — whether the number is a raw venue midpoint, a volume-weighted composite, or a filtered read.
How does an event-contract probability differ from a Fed Funds futures implied probability?
Fed Funds futures imply rate-path probabilities through OIS math on a deeply liquid, CME-cleared instrument with continuous two-sided markets. An event-contract probability comes from a binary contract on a thinner book, at a venue where counterparty and reg treatment differ. The two numbers frequently disagree by hundreds of basis points on the same underlying question. That gap is a liquidity and regulatory premium first, an informational disagreement second — a distinction that matters for anyone treating the difference as signal.
Is this suitable for regulated institutional use in 2026?
The feed itself is a data product, which sits in a well-understood regulatory category — vendor-delivered market data. The underlying venues are a different question. Kalshi is a CFTC-designated contract market; Polymarket's status in the US remains in flux. Institutional consumers using the feed as a display layer face negligible incremental risk. Institutional consumers building strategies that trade the underlying event contracts face a regulatory posture that is still being written, and should get counsel-signed clarity before scaling.
Why now, and why TS Imagine specifically?
TS Imagine sits close to buy-side risk workflows and has spent the post-merger period expanding its data-integration surface. Buy-side clients have been pulling event-contract data through scrapers for years. Bundling it into the risk console is a low-cost feature for TS Imagine and a workflow tidy-up for the client — one less browser tab, one more auditable data source. The competitive signal is that no bank-owned analytics platform got there first, which is consistent with the sell-side being downstream on this asset class.
What should a risk officer ask before enabling the feed in production?
Ask for 99th-percentile tick-to-console latency, not the average. Ask how contract resolution disputes are represented in the historical file. Ask whether the live feed and end-of-day settlement print are reconciled. Ask what happens during venue outages — does the last known price persist, does it null out, does it flag stale. And ask for the composition methodology in writing. Each question has a one-line answer if the vendor has thought about it, and a paragraph of hedging if they have not.
Does this compete with or complement existing alt-data feeds?
Complements more than competes, and awkwardly at that. Alt-data providers tend to sell derived signals — sentiment scores, positioning proxies, satellite-derived inventory estimates. This is closer to a market data feed on an adjacent venue. The mental placement matters: treated as market data, it belongs next to Bloomberg and Refinitiv feeds and should be priced accordingly. Treated as alt-data, it gets budgeted differently. TS Imagine appears to be positioning it as the former, which sets the pricing anchor high.