Marijuana Rescheduling Odds: The Bet On A Judge With No Deadline

Updated August 28, 2026 · 16 min read · by Jake Hari

Somebody spent Friday evening buying the idea that marijuana leaves Schedule I in the next four months. Between 4 and 6 PM ET on August 28, about 1,800 contracts changed hands on the Kalshi market that pays if that happens before January 1, 2027, and the price ran from 29 cents in mid-afternoon to 37 by six. The last trade printed at 40. A week earlier it had closed at 25; 12 days earlier, at 22.

Nothing was reported filed that afternoon: no ruling, no published rule, no signed order. What did happen, over the previous ten days, is that the federal hearing on the question closed its record: final briefs on August 17, and a corrected 2,533-page transcript on August 25. The next step belongs to one man, DEA Chief Administrative Law Judge Derek Julius, and the regulations give him no deadline at all.

We handed that record to seven AI models, hid the price from every one of them, and asked each to put a number on all three deadlines Kalshi sells. Their middle answer on the near one is 11 percent. Buyers on Friday were paying 39 cents. The rest of this page is about that gap, and about a rescheduling that already happened once this year and paid this contract nothing.

What Traders Bought This Week

The market is a ladder of three cumulative deadlines, all asking the same thing: will marijuana be moved out of Schedule I, the most restrictive federal drug tier, before a given date. A "Yes" contract costs its price in cents and pays exactly one dollar if the answer is yes, so the price doubles as the crowd's probability. Buying the before-2027 contract at Friday's 39-cent ask turns $100 into about $256 if it hits, and into nothing if it does not.

Here is how that price got where it is. On August 16 the before-2027 contract closed at 22 cents. The post-hearing briefs landed on the 17th. On the 19th it closed at 33 on about 3,000 contracts, its busiest day in weeks. It gave most of that back, closing at 25 on the 21st, and sat at 25 or 26 through the following Wednesday. Thursday the 27th brought a 2,800-contract burst in the early afternoon that pushed it to 30 before it settled back to 26 on the close. Friday brought the move above, and a 40-cent print.

The two longer deadlines barely noticed. Before-2028 went from 69 cents on August 17 to 68 at Friday's midpoint. Before January 20, 2029, the end of the current presidential term, held between 73 and 77 the entire time. Read together, that is a crowd that has not changed its mind about whether this gets done. It has changed its mind about when, and pulled the date forward.

That is a specific claim, and the arithmetic makes it stranger than it sounds. At 37.5 cents for before-2027 and 68 for before-2028, the market is putting 37.5 points on the next four months and only 30.5 points on all 12 months of calendar 2027. The crowd thinks the window between now and New Year's Day is likelier to carry this event than the whole of next year.

New to event markets? How a prediction market turns a question into a price is the five-minute version of everything this page assumes from here.

These are market prices and model estimates, not predictions of fact and not financial advice. Kalshi is a CFTC-regulated event-contract exchange, and availability varies by state. Kalshi requires traders to be 18+. Prices on this page carry an "as of" stamp and get re-read when the story moves.

The Order That Paid Nothing

To see why seven models read the same news and landed at 11, start with what this contract has already refused to pay for.

Kalshi's rule for every rung is one sentence: "If marijuana is rescheduled from Schedule I to a lower drug schedule under the Controlled Substances Act before [the deadline], then the market resolves to Yes." The fine print is where the money is. The exchange's own clarification, dated December 12, 2025: "The Payout Criterion requires that marijuana is legally removed from Schedule I. The mere announcement or publication of a Final Rule in the Federal Register does not satisfy the Payout Criterion if the rule has not taken legal effect." And a rule that moves "only certain cannabis derivatives or specific cannabinoids" without reclassifying marijuana as a whole does not count.

Those two sentences broke a contract this spring and settled it in July. On December 18, 2025, President Trump signed Executive Order 14370, directing the Attorney General to complete the rulemaking "in the most expeditious manner in accordance with Federal law." Traders read it as a promise. The contract for before July 2026 traded about 22,500 contracts that day, five times the day before, and held a 78-cent close; the before-2028 rung is where the promise got priced, trading as high as 97 cents.

Four months later the Justice Department delivered something narrower. Acting Attorney General Todd Blanche signed an order dated April 22 that published April 28 as a final rule, effective the same day. It moved marijuana to Schedule III only "to the extent that any of these are included in an FDA-approved drug product or are subject to a state-issued license to manufacture, distribute, and/or dispense marijuana or products containing marijuana for medical purposes." Everything outside those two boxes stayed in Schedule I.

Watch what the price did. On April 22, the date on the order, the before-July contract traded as high as 67 cents and closed at 35. On April 23, as the text circulated, it closed at 6.4 cents on roughly 409,000 contracts, the heaviest day in the market's history. The before-2027 rung fell from a 66-cent close to 47 the following day. Traders needed about one day to work out that a partial rescheduling is not this contract's rescheduling. The before-July contract settled No on July 1 after about 1.63 million contracts of lifetime trading.

A federal order rescheduled medical marijuana in April and took effect the day it printed. The contract asking about rescheduling lost 29 cents in a single day. On an event exchange the rule text pays, and the headline does not.

Two more orders landed in the Federal Register the same day. One withdrew the 2024 hearing notice and terminated the old proceeding, ; the other noticed a new hearing on the broad question, to begin June 29 and "conclude not later than July 15, 2026." That hearing is the thing whose record just closed.

What The Hearing Record Changed

The new hearing ran June 29 through July 15 before Judge Julius, 11 days of testimony. He allotted no time for closing arguments and told the parties to file them in writing by August 17.

The government's brief is the most important document on the docket, because the agency that runs the schedules is now arguing for the change. The DEA wrote that marijuana "no longer fits the statutory requirements for Schedule I because it has a currently accepted medical use within the United States and it has an accepted safety for its use under medical supervision," and that it "most appropriately belongs in Schedule III because its abuse potential and dependency profile most align with those substances in Schedule III." The government's case for accepted medical use, as reported from the transcript, leans on the scale of the state programs: more than 30,000 practitioners treating more than six million patients across 43 jurisdictions.

The other side filed too. Smart Approaches to Marijuana and the states of Idaho, Indiana and Nebraska wrote that "The Government now proposes to reschedule marijuana not based on the facts, but rather based on new and different standards—standards that are not found in the Controlled Substances Act." The National Drug and Alcohol Screening Association put its objection in commercial terms: "If the transfer of marijuana to Schedule III is finalized, the authority of the Federal government to test for it will end." The Tennessee Bureau of Investigation, an impaired-driving group and two physicians filed alongside them; the impaired-driving group and physician Kenneth Finn argued the DEA "failed to carry its burden" on medical value.

Then came the paperwork that appears to have moved the price. On August 20 Judge Julius ordered corrections to the transcript, 294 of them by one count, some of which flipped meaning ("there's adulterants" became "there's no adulterants"). The corrected version, all 2,533 pages, was published August 25. The final decision belongs to DEA Administrator Terrance Cole; the judge's recommendation is advisory.

Julius's own words in the record are the only forward guidance anyone has: "I will thoroughly consider and evaluate all the evidence presented at this hearing." No date, no range, no hint.

The docket this replaced is also part of the record, and it is why the panel prices stalls rather than sprints. In January 2025, DEA Chief Administrative Law Judge John J. Mulrooney II cancelled the hearing that was eight days from starting and suspended the case, after participants alleged improper private communications between DEA leadership and rescheduling opponents, and that the DEA Administrator had hand-picked the witness list. The case then sat frozen for more than a year. The April order is being fought as well: petitions from Republican state attorneys general, a drug-testing trade group and a cannabis pharmaceutical company were consolidated in the D.C. Circuit. Nebraska, Indiana and Louisiana filed together on May 22; Louisiana's attorney general asked to be dismissed a week later, leaving two states. The Justice Department told the court those petitioners "come nowhere near satisfying the demanding standard for that extraordinary relief" and invoke "pocketbook interests" served by keeping marijuana in Schedule I. Briefing on the stay motion closed July 17; as of the newest reporting we could find, dated August 1, both motions were still pending, and the April order has been in effect since April 28.

The Clock Between Here And A Payout

Everything on the board is a bet on how the next chain of steps runs. There are five links, and only one has a number attached.

Step	Who	Time allowed
Recommended Decision	Chief ALJ Derek Julius	"As soon as practicable." No deadline
Exceptions	The parties	20 days after receiving the report
Certify The Record	Judge Julius to the Administrator	No sooner than 25 days after serving the report
Final Order	DEA Administrator Terrance Cole	No deadline
Legal Effect	Publication in the Federal Register	Publication alone does not pay this contract; the April order took effect the day it printed, and nothing requires the next one to

Two things in that table do most of the work. The only mandatory floor is the 25-day certification wait, which means the calendar does not forbid a payout before January 1. What the calendar does is hand two undated, discretionary steps to two people, a judge who just took delivery of a 2,533-page record and an Administrator who has never been bound to a date on this file. The one I keep coming back to is the last row, because April proved this Attorney General can collapse signature, filing, publication and legal effect into six days when he wants to. A fast Julius plus an April-style effective date is the only path to the 2027 rung paying, and it needs both.

Hottest Prediction Markets Right Now

• 2028 Democratic presidential nominee · $197M traded
• 2027 Pro Football Champion · $73M traded
• 2028 U.S. Presidential Election winner? · $63M traded
• Pro Baseball Champion · $62M traded
• 2028 Republican presidential nominee · $60M traded

Every market above links to our full AI model verdict; browse them all on the OddsShopper prediction markets hub, and see how every settled call actually scored on the full graded scoreboard.

The Board

Seven models were shown the settlement rules, the Federal Register record and the dated procedural history above, and were not shown the price. Each priced all three deadlines from one view of the question rather than rung by rung. Then each read the others' reasoning anonymously and was invited to revise. The middle answer of the seven is the panel number below.

Deadline	What has to be true by then	Market	Panel
Before Jan 1, 2027	A broad rule in legal effect, about four months from today	37.5¢	11%
Before Jan 1, 2028	Same, with all of calendar 2027 to get there	68¢	60%
Before Jan 20, 2029	Same, by the end of the current presidential term	77¢	74%

Market column is the midpoint of the bid and the ask at 8:45 PM ET on August 28, 2026: 36 bid and 39 ask on the 2027 rung, 67 and 69 on 2028, 75 and 79 on 2029. All three are quoted two-sided. The 2027 rung carries about 699,000 contracts of lifetime volume with roughly 220,000 still open; 2028 and 2029 carry about 92,000 and 59,000 lifetime, with 53,000 and 31,000 open. Panel column is the seat median after the revision round.

Read as one sentence, the market says: 37.5 percent it happens by New Year's, 30.5 more points across 2027, nine more points in the 13 months after that, and 23 percent it never happens under this administration. The panel says: 11 percent by New Year's, 49 points across 2027, 14 in the final stretch, 26 percent never.

Those two curves agree on the destination and disagree about the calendar. On the long rung the gap is three points, a rounding error on a four-cent spread. On the near rung it is 26.5 points. The models put the modal month somewhere in spring 2027. The market put its money on the next 18 weeks.

Every Seat's Number

Every model appears to the others as a randomly assigned letter in the revision round, never by name, so nobody defers to a brand. The letters in the quotes below are theirs.

Seat	Before 2027	Before 2028	Before Jan 20, 2029	Revised?
Claude Fable	9%	46%	65%	later rungs
Claude Opus	11%	60%	74%	yes
Claude Sonnet	6%	48%	68%	yes
GPT	12%	60%	76%	yes
Gemini	10%	78%	88%	yes
Kimi	20%	57%	74%	yes
Grok	14%	67%	81%	near rung
Seat Median	11%	60%	74%
Market	37.5¢	68¢	77¢

Every seat on this panel is graded against real market settlements — records to date: Claude Fable 84% on 807 graded calls · Claude Opus 84% on 877 graded calls · Claude Sonnet 82% on 852 graded calls · GPT 84% on 6,460 graded calls · Gemini 86% on 5,151 graded calls · Kimi 82% on 2,576 graded calls · Grok 84% on 4,427 graded calls. Recomputed daily; the full scoreboard is public.

These are model estimates, not predictions of fact and not financial advice. Seven of the panel's nine seats ran this board; two were unavailable when it ran and are excluded rather than filled in. The blend is the seat median, not the average. Every number in this piece gets graded in public once the market settles, and you can check the panel's homework on the full graded scoreboard.

More live boards from the same panel: will the U.S. confirm aliens exist before 2027, the same deadline shape on another federal act, quoted 5.3 to 5.8¢ · government shutdown on October 1, the nearest federal-calendar contract on the exchange, at 5 to 6¢ · Trump third term odds, where the family-nominee-for-2028 contract sits at 6.7 to 9.1¢. Prices fetched August 28, 2026.

Even the highest seat on the near rung, Kimi at 20 percent, sits 17 points under the market. The lowest, Claude Sonnet at 6, would price the same contract near six cents. Nobody on the panel got within 15 points of what buyers paid on Friday.

Where The Panel Changed Its Mind

All seven seats moved after reading the others, and the near rung is where the arguments actually collided. The first-round spread ran from 5 to 23 percent; after the revision round it closed to 6 to 20, and the median came down from 15 to 11.

The two seats that had started highest both cut. Grok went from 23 to 14:

"Forecasters F and E correctly note that Julius's undated recommendation on a 2,533-page record plus the mandatory ~45-day post-report clock leave almost no room for an effective broad rule before Jan 1 2027; my 23% overweighted April-style same-day compression and conflicted with my own Mar–May 2027 modal, so I cut only that rung." — Grok

GPT went from 19 to 12, and named the settlement language as the reason:

"Forecaster E changed my mind modestly on the near rung by emphasizing that the settlement requires legal effect, not merely a signed or published order, so the ALJ recommendation, exceptions, certification, Administrator order, and effective-date window all have to fit into four months." — GPT

One seat moved the other way, and its reasoning is the sharpest case for the buyers. Claude Opus had called a 2026 payout close to impossible on procedural grounds, then raised its number from 6 to 11:

"I called the post-report clock a physical bar, but only ~45 days of it is mandatory, leaving four months of slack from today — the real constraint is Julius's discretionary timing, not arithmetic, so the 2026 hazard is small-but-real rather than forbidden, and April's same-day-effective order proves the executive can compress everything downstream." — Claude Opus

[Editor's note: the regulations set the mandatory floor at 25 days from service of the report to certification, with the 20-day exceptions window running inside it. The direction of the point stands for both Opus and Grok above; the mandatory stretch is shorter than the 45 days they cite.]

Gemini did something unusual: it cut the near rung and raised both far rungs at once, from 15/75/85 to 10/78/88, after abandoning a court stay as its main worry:

"Forecaster G successfully argued that obtaining a pre-effective judicial stay is extraordinary relief, and Forecaster F noted that current litigation is running alongside rather than tolling effective dates. This convinced me I was overestimating the risk of a court blocking the rule before it takes legal effect." — Gemini

And Claude Sonnet held its near-rung number at single digits while conceding the far rungs, with the clearest statement of why the panel is where it is:

"The higher A/B/D/G near-term numbers rest on confidence about executive-branch speed, not a new procedural fact." — Claude Sonnet

That is the whole disagreement in one sentence. The buyers are betting on speed. The panel is pricing the absence of a clock.

What Would Change The Panel's Mind

Four events, each of them checkable rather than a mood, and each with a direction.

1. Judge Julius files his recommended decision. The one step with no deadline. Two seats named it the bottleneck outright and a third put it first; the other four pointed past it to Administrator Cole's order or a court stay, but every seat that named the update that would move it most named this one. A report in September or October keeps the 2027 rung alive; one that slips past Thanksgiving leaves almost nothing, because the 25-day certification floor behind it cannot be shortened by the parties who want speed. A recommendation against Schedule III would hit every rung at once.
2. The D.C. Circuit rules on the pending stay. The motion has been fully briefed since July 17. A stay of the April order would be the first sign courts will interrupt this docket rather than run alongside it, and it would cut the 2028 and 2029 rungs. A denial confirms what four seats moved toward in the revision round, and lifts them.
3. Administrator Cole signals an effective date. April's order took effect the day it printed. If the broad rule is written the same way, the last link in the chain costs days instead of a month, which is worth several points on the near rung by itself.
4. Administrator Cole tips his hand on the recommendation. The judge advises; Cole decides, and nothing binds him to adopt what Julius sends up. Any public signal that he intends to adopt the recommendation lifts every rung; a signal that he will reject or rework it hits the back half of the ladder hardest, because the front half already assumes a near-miracle of speed.

Now	Corrected transcript published August 25, 2026. Judge Julius is writing his recommended decision, with no deadline set
After The Report	20 days for exceptions; the record is certified to the Administrator no sooner than 25 days after service
Then	Administrator Cole issues a final order. It has to take legal effect before any rung pays
Rung Deadlines	January 1, 2027 · January 1, 2028 · January 20, 2029, each closing early if the event happens first
Also Live	The consolidated D.C. Circuit challenge to the April order, briefed and awaiting a ruling

This page gets re-scored when the story moves, and the next thing that moves it is Judge Julius filing his report, which carries no date. Prices and panel numbers are as of 8:45 PM ET, August 28, 2026.

More on this:

• Michigan Governor Odds: Seven Raters, Not One Says Safe
• Crypto 2026 Odds: Can Any Of These Ten Finish The Year Green?
• Michigan House Races 2026 Odds: The Nominee Who Quit And Won
• UFC Heavyweight Title Odds: A Champion Who Hasn't Defended
• Bradley Barcola Transfer Odds: The \u00a3120M Liverpool Deal Priced Overnight

The Bottom Line

Friday's buyers are not wrong about the shape of the story. The agency that runs the schedules has asked, in writing, for the change a presidential order demands, the hearing is over, the record is closed, and the only rescheduling this administration has produced so far took legal effect the day it was published. If you believe that pattern repeats, the near rung is where the reward is, and the 2029 rung is already priced close to where seven models put it, while 2028 sits eight points above them.

The panel's answer is that the pattern ran through the one office in the building that nobody can hurry. Judge Julius has a record the length of three novels, a standard of "as soon as practicable," and a docket whose last judge froze the case for a year. The models put around one chance in nine on all of that resolving, being certified, being ordered and taking legal effect inside 126 days. The market put nearly four in ten.

Both numbers get graded on the same day, and it is the one date on this whole page that is fixed: January 1, 2027.

These are model estimates, not predictions of fact and not financial advice. Event contracts carry real risk of loss, availability varies by state, and you must be 18+ to trade. Do your own work at whatever price you are actually offered.

Hero illustration: OddsShopper, in the house collage style. Donald Trump photo by Shealeah Craighead, licensed Public domain; photos cropped, toned, and composited.