Over the past ten days, a single sentence has been recycled across crypto Telegram groups with the persistence of a settlement oracle: predict.fun is about to launch, so what is each point worth?
The sentence is doing something clever. It poses a question that sounds numerical โ "what is each point worth" โ while quietly skipping every input required to answer it. There is no published total supply. No conversion ratio. No snapshot date. No disclosed circulating float. No audit, no tokenomics page, no legal wrapper. The question is not a valuation exercise. It is a narrative device, and like most narrative devices in this market, it is more useful to the person asking than to the person answering.
I have watched this genre mature across three cycles now. In 2017 it was "what is the fair value of the EOS airdrop" โ a question I dissected at length from the inside, spending four months on a comparative tokenomics study of delegated-proof-of-stake models while everyone else argued about price. In 2021 it was "what is a generative Art Blocks mint worth," which I dismantled with twelve thousand mints of on-chain provenance data proving that secondary volume was decoupling from creator royalties. The nouns change; the structure does not. A pre-market asset with no disclosed supply curve gets a number attached to it by someone who benefits from the number existing. The number is not a forecast. It is a marketing artifact wearing the costume of analysis.
That distinction matters more in a bear market than in a bull one, because in a bull market the artifact is subsidized by reflexivity โ price rises, the narrative validates itself, and nobody audits the math. In a drawdown, the artifact gets tested. Survival matters more than upside, and readers want to know whether the thing they are holding is an asset or a promise. So it is worth doing the thing the original headline refused to do: decomposing what a "point" actually is, what it can plausibly be worth, and why the honest answer is a range so wide it functions as a warning rather than a valuation.
Context
Prediction markets are not a new primitive, and this is the first fact the discourse tends to bury. The concept โ trading binary outcomes on future events, with prices expressing aggregate probability โ has existed in academic form since the Iowa Electronic Markets in the late 1980s and in crypto-native form since Augur's 2015 token sale. Polymarket scaled the model to genuine volume during the 2024 US election cycle, where event contracts on the presidential race briefly rivaled the liquidity of mid-cap altcoin order books. Azuro took a different route, building liquidity-pool-based event markets designed for composability across multiple chains. Neither was first, and predict.fun will not be last. The technical primitive is mature; the competitive question is entirely about distribution and settlement.
What is new is not the market structure. What is new is the distribution mechanism wrapped around it: the points program.
The points meta is a post-2020 invention with a traceable lineage. Blur popularized the aggressive, bid-to-earn points race in 2023, turning loyalty into a speculative instrument with a live leaderboard. EigenLayer converted restaking points into a distinct asset class of their own, with secondary markets pricing them before any token existed. By 2024, points had become the default customer-acquisition primitive for anything that intended to launch a token but wanted to delay the legal and market consequences of launching it. The logic is elegant from the issuer's side: points manufacture speculative demand before you have a security, a float, or a price. They are a pre-token with none of the obligations of a token โ and none of the transparency either.
History rhymes, but the code doesn't. The ICO era sold a future token with a whitepaper and a smart contract that anyone could read. The points era sells a future token with a leaderboard and a Terms of Service clause reserving the right to change the rules. The financial shape is similar โ capture user attention and capital before the asset exists โ but the technical and legal substrate is fundamentally different. An ICO allocation was an on-chain claim, auditable and sometimes enforceable. A point is, in almost every implementation I have examined, a row in a centralized database that the operator can edit at will, retroactively, without disclosure. That asymmetry โ on-chain aspiration, off-chain control โ is the load-bearing fact the "what is each point worth" question never touches. And it is where any honest analysis has to start.
Core
I want to be precise here, because precision is the entire point of the exercise. To price a point, you need five numbers, and predict.fun has published none of them.
The first is total points outstanding. Without it, you cannot compute the denominator of the conversion. A program that has issued ten million points and one that has issued ten billion produce radically different per-point values even if every other variable is identical. The second is total token supply at the generation event. The third is the conversion ratio โ how many points map to one token, if they map at all. The fourth is initial circulating supply, because a point's realizable value depends on the float you can actually sell into, not the fully diluted headline. The fifth is the lockup and vesting schedule applied to the converted allocation, because a point that converts into a twelve-month cliff is worth something very different from one that converts into immediate liquidity.
I have run this arithmetic for other points programs. In one case I reconstructed the outstanding points supply from on-chain claim events and a public leaderboard endpoint, then cross-referenced it against snapshot announcements to bound the denominator. The resulting per-point range โ depending on whether you assumed a 5% or a 15% initial float and a generous or stingy conversion โ spanned two orders of magnitude. That is not a valuation. That is a random number generator with extra steps. When I built the ETF flow model for "The Liquidity Premium" in 2024, I had a decade of comparable traditional-finance instruments to calibrate against. Here there is nothing to calibrate against, because the issuer controls every parameter and has disclosed none of them.
A point is not an asset. It is an unsecured, non-transferable, revocable liability that the issuer holds an option to convert โ or not convert โ at a ratio they have not defined, at a time they have not committed to, subject to rules they reserve the right to rewrite.
Read that sentence twice, because it is the entire risk surface compressed into one line. Everything else in this article is elaboration.
Consider the mechanics from the issuer's perspective. A points program is a call option sold to the issuer, not to the user. The issuer captures the user's time, capital, and attention today, and retains sole discretion over the payoff tomorrow. If the token launches into strength, the issuer can honor points generously and claim credit for a fair launch. If the token launches into weakness, or if regulatory pressure materializes, the issuer can dilute the conversion, impose eligibility cliffs, or quietly extend the snapshot window until the marginal farmer gives up. The user has no contractual recourse in any implementation I have seen. This is not cynicism; it is the plain reading of the standard points Terms of Service, which almost universally contain a clause permitting unilateral modification.
The sybil dimension compounds this. Points programs are permissionless to farm and trivially automatable, and by the time a program reaches the "what is each point worth" stage of discourse, the sophisticated farmers have already been running scripts for months. Based on my experience auditing airdrop distributions, the top decile of wallets in a mature points program routinely captures 40% to 60% of the total allocation, and a meaningful fraction of those wallets are controlled by a handful of operators spinning up thousands of addresses. The ordinary user calculating their per-point value is computing a number that the farming cohort has already arbitraged down before the snapshot was even announced. This is the quiet theft of the points meta: it markets itself as a meritocratic reward for early participation, and it settles as a subsidy for whoever can generate the most wallets the fastest. The leaderboard cannot distinguish a genuine user from a script, because a point is defined by activity, not identity.
Then there is the oracle question, which the discourse ignores entirely because it is not about upside. A prediction market is only as honest as its settlement layer. Polymarket routes resolution through UMA's optimistic oracle, itself a token-voting dispute system with its own escalation-game attack surfaces and dispute bonds. Azuro embeds resolution logic into its liquidity pools. Any prediction market must answer a deceptively hard question: who decides what happened, and what happens when they are wrong, slow, or captured? For binary political and sports markets, the answers are usually clean, because the outcome is publicly verifiable and the dispute window is short. For anything ambiguous โ "did the protocol ship before the deadline," "was the event a recession," "did the merger close" โ the oracle becomes the entire product, and a disputed resolution is an existential event for the market in question.
If predict.fun has not published its oracle design, its dispute-resolution process, or its fallback for contested outcomes, then it has not published the single most important piece of its technical architecture. The original news brief disclosed none of this. Neither, as of this writing, has the project. A prediction market without a disclosed oracle is a prediction market with an undisclosed single point of failure, and in a bear market that is not a footnote โ it is the headline.
The regulatory layer is the final and most underweighted input. Prediction markets sit at the intersection of two regimes that each independently dislike them. In the United States, event contracts fall under the jurisdiction of the CFTC, which has repeatedly moved against platforms offering political and sports markets, and the SEC applies the Howey test to any token distribution. A points-to-token conversion triggers the second test almost by construction: users contribute capital and labor (money staked in prediction positions and trading volume generated), in a common enterprise (the project), with an expectation of profit (the anticipated airdrop), derived from the efforts of others (the team's development, listing, and liquidity work). That is not a subtle legal question. That is the Howey test reading itself aloud in four-part harmony.
The practical consequence is geographic. Polymarket restricted US users for years precisely because of this exposure, and MiCA in the European Union imposes its own disclosure regime on any asset issuance. Any project intending to launch a token with an airdrop faces the same fork: exclude US participants and shrink the eligible base โ which mechanically raises each remaining point's value but introduces eligibility risk for everyone โ or include them and take on enforcement risk. Neither choice is free, and the original brief mentioned neither.
Layering these inputs, I can construct a defensible per-point range for predict.fun only by assuming values I have no basis to assume. If I assume a mid-single-digit-percent initial float and a stingy conversion, the per-point value rounds to noise. If I assume a generous conversion and a large float, it looks meaningful. The honest output is that the range is wide enough to contain both "worthless" and "moderately valuable," and a range that contains both is not a forecast โ it is a confession of ignorance dressed as a spreadsheet.
Contrarian
Here is the part the discourse gets backwards. The conventional worry is that predict.fun's points will be worth less than expected โ that the airdrop disappoints, the conversion is stingy, and the farmers dump into the listing. That is a real risk, but it is the second-order risk. The first-order risk is more uncomfortable to state: the project may never have committed to a conversion ratio at all, in which case the entire "what is each point worth" conversation is pricing a contract that does not exist.
The framing is itself the tell. If predict.fun had published a conversion schedule, the question would have an answer and no one would be asking it rhetorically in Telegram groups. The fact that it circulates as open speculation rather than a lookup means the ratio has been deliberately left undefined โ which is the rational strategy for an issuer, because an undefined ratio preserves maximum optionality and minimum legal commitment. The user's corresponding position is maximum exposure and zero optionality. The symmetry is perfect and entirely in the issuer's favor.
The second blind spot is competitive, and it is structural rather than narrative. Prediction markets exhibit powerful network effects and liquidity moats. Polymarket did not win the 2024 cycle because its technology was superior; it won because event-contract liquidity is winner-take-most, and a trader routes to the venue with the tightest spread and the deepest book. A new entrant with a ".fun" brand and a points program can manufacture attention, but attention is not liquidity. The code that matters is the matching engine and the oracle, and neither is improved by a leaderboard. The same fragmentation logic that leaves dozens of Layer 2s fighting over an unchanged user base applies here: adding venues does not add traders, it slices the same scarce flow into thinner books.
History rhymes, but the code doesn't. Every cycle, a new cohort discovers that distribution can be bought but liquidity must be earned, and every cycle, the cohort learns it after the token has already repriced. The points program is a distribution machine, and distribution without settlement quality is a leaky bucket.
Takeaway
The question worth asking is not what each point is worth today. It is what predict.fun publishes next โ because the moment a total supply, a conversion ratio, and a snapshot date appear in writing, the speculation collapses into arithmetic, and the arithmetic will be ruthless about whatever the leaderboard was actually rewarding. Until those three numbers exist, treat the points as a lottery ticket the issuer has not yet priced, and treat every confident per-point estimate โ including the ones you will read tomorrow โ as a number someone needed to be true. Watch the oracle design before the token design. In a prediction market, the settlement layer is the product, and everything else is a spreadsheet waiting to be tested.