Updated August 26, 2026 · 16 min read · by Eric Lindquist
Bitcoin set the low of its August on the first afternoon of the month, at $62,209.81. Two days later it came back to the same price and stopped at $62,210.14. Thirty-three cents separated the two bottoms, and nothing since has come close to either. By the small hours of August 25, Bitcoin had run all the way to $81,265.30, its highest print since May.
Kalshi lists eight contracts on how low Bitcoin gets this month, starting at "below $60,000" and stepping down in $2,500 increments to "below $42,500." Not one has been touched. Each step is called a rung, which makes the whole set a ladder. All eight rungs now show the same thing: a seller asking one cent, the smallest price the exchange can quote, and no buyer offering anything at all. On an exchange, the ask is what a seller wants and the bid is what a buyer will pay; a contract with an ask and no bid is one person's price, and the market has not agreed to it. That distinction runs through everything below.
We asked eight AI models to price the whole ladder without letting any of them see what the market was charging. Every one of them came back under a single percent, which is roughly where the exchange is, and then they spent a whole revision round arguing about something else entirely: what this contract actually measures. Five cut their numbers by as much as twentyfold on the strength of that argument. One held its ground and named the reason. Chasing down which side was right meant going to the exchange's own paperwork, and what is in there is worth more than the trade.
New to this corner of the market? Our plain-English guide to how Bitcoin price contracts work on Kalshi covers the mechanics before a single number below means anything.
Free: The Weekly PM Market Brief — the 8-model panel's graded record, the week's biggest market-vs-model gaps, and what's spiking next. One email, Sundays.
Prices on this page refresh during active news waves; every quote below is stamped "as of August 26, 2026" so you can see when it was last read.
These are market prices and model estimates, not predictions of fact and not financial advice. Kalshi is a CFTC-regulated event-contract exchange (18+; availability varies by state).
What Bitcoin Actually Did In August
August opened soft. Bitcoin spent the first eighteen days in a band running roughly $62,200 to $65,400, printed its two lows on the 1st and the 3rd, and failed four separate times to hold above $65,000. Through the 18th, the month read as a continuation of a summer that had already produced the cheapest Bitcoin of the year.
Then the tape turned. From the August 18 close near $64,700, Bitcoin ran to $81,265 by the early hours of the 25th, about 26% in a week. Reporting at the time attributes it to three things arriving together: a U.S. Treasury announcement doubling its bond buybacks, $1.92 billion of weekly inflows into spot Bitcoin ETFs, their biggest week in ten months (Bloomberg), and more than $4 billion of short positions liquidated as price climbed through them (CryptoBriefing). Bloomberg reported on August 26 that the move had stalled around $80,000 with traders waiting on inflation data.
How that rally was built matters more than how big it was, and all eight models landed on the same reading of it without conferring. A rally powered by short liquidations is one where the leverage got flushed on the way up. The classic Bitcoin unwind runs the other way: leveraged longs pile into a run and then blow out of it together. Whatever risk sits under this price, that particular fuse is not lit. CoinDesk's read on August 25 reached the same place from the other direction, describing a move driven by short covering with no fresh leverage behind it, and by open interest, the count of futures positions still on the books, collapsing as it ran.
Set against the rest of 2026, August's floor was not even close to the cheapest Bitcoin of the year. January bottomed at $75,644 after topping out at $97,964 on the 14th, still the high-water mark for the year. February came within a dollar of a round number, low tick $60,001. March through May never traded below $64,900. June bottomed at $58,000.00, to the cent, and July went lower still, setting the 2026 floor at $57,718. August's $62,210 is the fourth-lowest month of the year, and it happened before most people were paying attention.
July is the month worth holding onto, with one wrinkle. Kalshi's months run on New York time, so the July 1 print that set the calendar-year low actually landed inside the JUNE contract. Measured on the exchange's own July window, Bitcoin's low was $58,243, and Kalshi's July ladder settled with "below $60,000" paying out and "below $57,500" expiring worthless. That pair of results is about to do some work.
The Contract Says Two Different Things
Kalshi's headline rule for each rung is a straightforward touch test:
If the price of BTC after issuance and through 11:59 PM ET on Aug 31, 2026 is ever below $60000.00, then the market resolves to Yes.
Read the fine print underneath it and a different machine appears:
At each minute throughout the market duration, a settlement value is calculated using every minute-by-minute CF BRTI price for BTC. For each of these individual minute-by-minute checks, the top 20% and bottom 20% of the cumulative dataset are removed before calculating the average of the remaining values.
The binding contract terms, which live in a PDF most traders never open, say the same thing in the exchange's legal register: spot price is measured "by taking a simple average of the values for each minute after Issuance and prior to [the deadline], ignoring the top 20% and the bottom 20% of values."
CF BRTI is the CF Bitcoin Real-Time Index, a reference price built from several exchanges so no single venue's glitch can set it. On the most natural reading of that paragraph, the number compared to each strike is a running average of every minute since August 1, taken after the cheapest fifth and the dearest fifth of those minutes are thrown out. One bad minute barely moves it. One bad afternoon moves it a little. Hold onto the phrase "most natural reading," because it is doing work.
A trimmed average also has a property that is easy to miss. Because the bottom fifth of the data gets discarded, new lows count for how MANY of them there are, not how low they go. Once the minutes left in the month fall below that 20% band, fresh crash prints land inside the discarded fifth and stop moving the average at all.
We rebuilt the series to see where that stands. Using Coinbase one-minute prints as a stand-in for the CF BRTI tape, here is Kalshi's trimmed average for every minute of August, as of August 26, 2026:
- It sits at $65,273 right now, roughly $13,153 below spot, because it still carries three weeks of $62,000-to-$65,000 Bitcoin.
- Its lowest reading all month was $62,925, on August 2. That is $2,925 above the shallowest rung, and it is as close as this route ever came to paying out.
- 7,803 of the window's 44,629 minutes are left, which is 17.5% of the data. The bottom trim discards 20%.
Push that to its limit. If Bitcoin printed one dollar for every one of the 7,803 minutes remaining in August, the final trimmed average would come to $64,220, still $4,220 above the shallowest strike, because every one of those catastrophe minutes would land inside the discarded bottom fifth. Dragging the average under $60,000 would take roughly 11,463 minutes at a dollar, or 8.0 days. August has 5.4 left.
Before leaning on any of that, we checked it against a month the exchange has already paid out. Running the identical reconstruction across July's contract window, all 44,629 minutes of it, puts the low of the trimmed average at $58,695. Kalshi settled July with "below $60,000" YES and "below $57,500" NO, so the value had to sit between those two strikes. The reconstruction lands $1,195 inside that window.
Run the same arithmetic rung by rung and each strike has a moment after which no amount of downside could drag a month-long average to it before expiration. Hold that thought, because the next section asks whether a month-long average is what this contract measures at all. The deepest strikes went quiet first, during the squeeze itself:
| Rung | Last minute the trimmed-average route could still have reached it |
|---|---|
| Below $60,000 | Aug 24, 4:33 PM ET |
| Below $57,500 | Aug 23, 11:23 PM ET |
| Below $55,000 | Aug 23, 6:12 AM ET |
| Below $52,500 | Aug 22, 12:59 PM ET |
| Below $50,000 | Aug 21, 7:45 PM ET |
| Below $47,500 | Aug 21, 2:29 AM ET |
| Below $45,000 | Aug 20, 9:10 AM ET |
| Below $42,500 | Aug 19, 3:50 PM ET |
Reconstructed from Coinbase one-minute prints, so read it as the arithmetic of the trimmed average rather than a determination by the exchange.
Here is the catch, and it is the reason this section has a plural in its heading. That table closes one door. Whether there is a second door is a question the exchange's own documents answer both ways.
Reading A, the one the arithmetic above models: the trim runs across the whole month's accumulated minutes, so the value being tested is a single running average and late lows get thrown out with the rest of the cheap fifth. The market page's phrase "the cumulative dataset" is what points here. Under Reading A the table above is the whole story.
Reading B: the trim runs WITHIN each minute, throwing out that minute's outlier quotes to produce one robust price for that minute, and the contract fires the first time any single minute's price prints below the strike. The binding terms' phrase "the top 20% and the bottom 20% of values for each minute" points here. So does Kalshi's own name for the series in its product metadata, "BTC one touch minimum," which is a strange label for a month-long average. So does the early-close condition every one of these markets carries, which the sibling annual contract describes in terms that name the raw index: "If the BRTI crosses the threshold at any point during the measurement period (triggering early expiration), the market immediately resolves." Under Reading B this ladder is an ordinary crash bet with five days on the clock.
The only evidence here with money behind it, and it is circumstantial rather than interpretive, is what traders are paying on the annual board. Kalshi's 2026 contract on the same question prices "below $55,000" at 30¢ bid, 31¢ ask. Under a pure Reading A that price is a stretch: the 2026 trimmed average since that market opened in February runs about $68,400, and pulling a 200-day average down to $55,000 inside four months would take a collapse far deeper and longer than a three-in-ten price implies. Under Reading B, "Bitcoin prints below $55,000 once before New Year" at 31¢ is an unremarkable quote. That is circumstantial and it is not proof. Traders can be wrong about a rule, or they may be pricing a genuinely prolonged collapse that would drag even a year-long average down. But a three-in-ten price is hard to square with Reading A. We cannot settle this from public documents and we are not going to pretend otherwise; what we can say is that the money is behaving as though Reading B is the live one.
The rest of Kalshi's Bitcoin shelf shows where the money went. Four contracts on the same asset, one pinned at the floor and three carrying real two-way prices:
| Contract | Price now |
|---|---|
| Aug Low < $60,000 | 1¢ |
| Aug High > $82,500 | 29¢ |
| 2026 Low < $55,000 | 30¢ |
| 2026 High > $100,000 | 26¢ |
Kalshi bids, fetched August 26, 2026. The August floor line is an ask with no bid behind it; the other three are two-sided quotes. Same asset, same exchange, same afternoon.
Two honest caveats before the board. We are reading Coinbase prints rather than the CF BRTI index itself, and the two are close cousins rather than the same series. And Kalshi is the only authority on how its own contracts settle; everything above is our reconstruction with its method attached.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $194M traded
- 2027 Pro Football Champion · $72M traded
- 2028 U.S. Presidential Election winner? · $62M traded
- Pro Baseball Champion · $61M traded
- 2028 Republican presidential nominee · $60M traded
- Bitcoin price at the end of 2026 · $33M 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
The panel priced blind. No seat saw a Kalshi quote, and no seat knew what any other seat had said until the revision round. Verdicts generated August 26, 2026; prices as of August 26, 2026.
| Rung | Move needed from here | Market | Panel blend |
|---|---|---|---|
| Below $60,000 | -23.5% | 0 Bid / 1¢ Ask | 0.42% (six-seat revision round: 0.14%) |
| Below $57,500 | -26.7% | 0 Bid / 1¢ Ask | 0.27% (six-seat revision round: 0.082%) |
| Below $55,000 | -29.9% | 0 Bid / 1¢ Ask | 0.17% (six-seat revision round: 0.048%) |
| Below $52,500 | -33.1% | 0 Bid / 1¢ Ask | 0.11% (six-seat revision round: 0.029%) |
| Below $50,000 | -36.2% | 0 Bid / 1¢ Ask | 0.060% (six-seat revision round: 0.018%) |
| Below $47,500 | -39.4% | 0 Bid / 1¢ Ask | 0.032% (six-seat revision round: 0.011%) |
| Below $45,000 | -42.6% | 0 Bid / 1¢ Ask | 0.017% (six-seat revision round: 0.0068%) |
| Below $42,500 | -45.8% | 0 Bid / 1¢ Ask | 0.0090% (six-seat revision round: 0.0040%) |
None of the eight contracts has a bid. All eight sit one-sided at the exchange's one-cent minimum increment, so "1¢" here means "somewhere between nothing and one percent" rather than a real quote, and a round trip costs the entire position. Lifetime volume across the rungs runs $6,557 to $430,507, with 200,489 contracts still open on the shallowest one. The panel-blend figures are model estimates for informational purposes, not predictions of fact and not financial advice.
The panel's first-round middle answer on the shallowest rung is 0.42%, and the exchange cannot quote below a penny, so the two are not really arguing. What the panel adds is resolution. Every rung from "below $60,000" to "below $42,500" wears the same 1¢ tag even though the models separate the top from the bottom by a factor of 47. The deep end is where that matters: "below $42,500" needs 46% in five days, a move Bitcoin has made 11 times in 4,044 five-day windows since 2015, and in none of the 1,687 of those windows that sit after 2022.
Every Seat's Number
| Model | < $60k | < $57k | < $55k | < $52k | < $50k | < $47k | < $45k | < $42k | 5th-pct low |
|---|---|---|---|---|---|---|---|---|---|
| Fable | 0.50% | 0.30% | 0.18% | 0.11% | 0.060% | 0.035% | 0.020% | 0.012% | $69,500 |
| Opus | 0.35% | 0.24% | 0.17% | 0.11% | 0.074% | 0.048% | 0.030% | 0.019% | $69,500 |
| Sonnet | 0.53% | 0.39% | 0.29% | 0.21% | 0.15% | 0.11% | 0.084% | 0.062% | $70,000 |
| ChatGPT | 1.10% | 0.55% | 0.27% | 0.13% | 0.060% | 0.028% | 0.013% | 0.0060% | $69,000 |
| Gemini | 0.22% | 0.11% | 0.050% | 0.020% | 0.0090% | 0.0040% | 0.0020% | 0.0010% | $71,500 |
| Kimi | 1.00% | 0.60% | 0.36% | 0.22% | 0.13% | 0.080% | 0.050% | 0.030% | $67,000 |
| GLM | 0.030% | 0.020% | 0.015% | 0.010% | 0.0070% | 0.0050% | 0.0030% | 0.0020% | $69,000 |
| DeepSeek | 0.050% | 0.028% | 0.017% | 0.011% | 0.0070% | 0.0050% | 0.0036% | 0.0026% | $69,600 |
| Panel Median | 0.42% | 0.27% | 0.17% | 0.11% | 0.060% | 0.032% | 0.017% | 0.0090% | $69,500 |
Model estimates for informational purposes. Not predictions of fact and not financial advice. The 5th-percentile low is each seat's answer to a different question: how cheap does Bitcoin get in a bad-but-not-crazy version of the next five days.
Every seat on this panel is graded against real market settlements — records to date: Gemini 86% on 4,952 graded calls · Kimi 83% on 2,534 graded calls · GLM 81% on 2,544 graded calls · DeepSeek 80% on 2,583 graded calls. Recomputed daily; the full scoreboard is public.
Every number in this piece gets graded in public once the market settles, and the running tally lives on the full graded scoreboard.
More live boards from the same panel: August's ceiling question is still two-sided, with "above $82,500" bid 29¢ on the monthly high board; the year-long version of that same question has "above $100,000" at 26¢ on Bitcoin's 2026 price board. Prices fetched August 26, 2026.
Where The Panel Changed Its Mind
After the first round, each model was shown the other seven ladders and the reasoning behind them, stripped of names, and invited to revise or hold. Six of the eight seats ran the revision round. Every one of them revised something, five of them cut the top of their ladder hard, and all five cut it for the same reason.
The argument that did it came from the seat that had priced the whole ladder lowest in round one. Its case was arithmetic, not judgment:
The haircut is effectively total, roughly 99% versus a naive spot-low model, because the remaining ~7,800 minutes (17.3% of the cumulative dataset) all fall within the 20% bottom-trim band, so even a crash to $40,000 for the entire remaining window would be entirely trimmed away. — GLM
Circulated anonymously, that reasoning cut through the panel. ChatGPT dropped its shallowest rung from 1.10% to 0.080%, Gemini from 0.22% to 0.010%, Sonnet from 0.53% to 0.060%. What is striking is that none of them adopted it wholesale. Each named the piece it could not verify:
My round-1 haircut (~45%, based on "needs a sustained move") was the wrong kind of adjustment. It treated this as a magnitude question when it's actually a denominator-arithmetic question. I didn't go as low as F/G's literal near-zero because I can't fully verify "cumulative dataset" means an expanding since-issuance time-series trim rather than some other construction. — Sonnet
I do not go to F's 0.03% because I hold ~25-30% probability that the methodology is implemented or interpreted less literally (per-minute or rolling-window trimming), in which case peer-style sustained-crash probabilities of 0.3-0.5% apply. — Kimi
Then there is the seat that held. Fable moved its deep rungs on a separate point about decay shape and left the top of its ladder exactly where it was, and it is the only seat that put the other reading of the contract on the record:
Forecaster F's whole-month cumulative-trim argument was the strongest against me, but I reject it: the clause "the trimmed mean associated with any single minute," the early-close behavior of crossed markets, and July's $60k rung settling YES while August opened at $62.7k all indicate a per-minute trimmed touch trigger, not a month-average that late-window prints cannot move. — Fable
[Editor's note: the panel could not see Kalshi's contract-terms document or the sibling annual market's rules, both of which we pulled after the revision round. Those sources do not settle the argument, but they keep Fable's side of it alive: the terms trim "for each minute" and the early-close language names the raw index, which is Reading B almost word for word. Read the market page's "cumulative dataset" literally instead and the majority is right. GLM's figure of 17.3% for the remaining share of the window is quoted as written; our own count of the minutes puts it at 17.5%, which does not change its argument.]
Here is the whole revision round on the shallowest rung, which is where the argument had the most room to bite:
| Model | Round 1 | After reading the others | Change |
|---|---|---|---|
| Fable | 0.50% | 0.50% | held |
| Opus | 0.35% | 0.30% | 0.86x |
| Sonnet | 0.53% | 0.060% | 0.11x |
| ChatGPT | 1.10% | 0.080% | 0.07x |
| Gemini | 0.22% | 0.010% | 0.05x |
| Kimi | 1.00% | 0.20% | 0.20x |
| Panel Median | 0.52% | 0.14% | 0.27x |
Model estimates for informational purposes. Not predictions of fact and not financial advice. Two of the eight seats did not run the revision round, so the medians in this table cover the 6 that did and will not match the eight-seat blend above.
Which leaves an unusual scoreboard. The panel's disagreement about Bitcoin was trivial, a few tenths of a percent on a market quoted in whole cents. Its disagreement about what the contract measures was the whole ballgame, and it is still open. So read the two numbers as answers to two different questions. Under Reading B the blind first-round 0.42% on the shallowest rung is the estimate, and the question is simply how likely a crash is. Under Reading A, our own arithmetic puts August at effectively nothing, and what is left over is the chance we have misread the rule or mis-rebuilt the index. The revision round's 0.14% is roughly what you get by mixing the two, which is what five of the six seats said they were doing. We lead with 0.42% because it is the number that does not depend on our own reconstruction being right. Either way, both are a rounding error next to the plain arithmetic: Bitcoin needs 23.5% in five days, which it has managed in 0.89% of the 1,687 five-day windows since 2022, and in none of the 71 of those that followed a week as strong as this one.
What Would Change The Panel's Mind
Five things bear on these numbers before the window shuts. Two are still ahead on the calendar, one has already happened, and two are conditions rather than events.
Jackson Hole, August 27 to 29. The Kansas City Fed's symposium runs those three days on the theme "Financial Innovation: Implications for Payments and Policy" (Federal Reserve Bank of Kansas City), which puts digital assets closer to the agenda than they have ever been. A hawkish surprise would tend to push these numbers up, and anything reassuring would tend to push them further down, though a room full of traders has surprised everyone before.
Warsh's first keynote, Friday morning. Kevin Warsh took the oath as Fed chair on May 22 (Federal Reserve Board) and delivers his first Jackson Hole address on the morning of August 28. With the symposium's theme pointed at payments and financial innovation, a new chair's opening statement of intent is the largest scheduled source of surprise left in the window.
The inflation print, already behind us. The data the Bloomberg piece said traders were waiting on landed at 8:30 a.m. ET on August 26: July personal income and outlays, with the PCE price index up 0.2% on the month and 3.7% over the year, 3.3% excluding food and energy (Bureau of Economic Analysis). That removes one of the two macro events people had stacked into this week, and it means the remaining calendar risk sits almost entirely on Friday.
The ETF bid reversing. The rally's most-cited support was $1.92 billion of weekly spot-ETF inflows. A week of outflows near that size would remove the buyer that carried price through $70,000, and daily flow prints are public, which makes it the cleanest measurable falsifier of the panel's read.
Spot demand failing to replace the squeeze. Bitget Research's chief analyst Ryan Lee framed the next phase as a test of whether ordinary buying can replace the squeeze as the thing holding price up (crypto.news). If it does not, a drift back toward August's low is the natural path. Even then the shape is wrong for these contracts: a full round trip to $62,210 is about $16,217 of downside from here, and the shallowest rung needs $18,426.
When This Settles, And What Comes Next
| Settlement | 11:59 PM ET, August 31, 2026 |
| Time Remaining | 5.4 days (7,803 minutes of a 44,629-minute window) |
| Catalysts Before Then | Jackson Hole, Aug 27 to 29 · Warsh's first keynote as chair, Aug 28 · month-end ETF flow prints. July PCE already released, Aug 26 |
| Early Close | A rung that meets its criterion closes and expires immediately, so a settled rung will not wait for August 31 |
| Next Cycle | Kalshi lists a fresh floor ladder for September within hours of expiry |
| As Of | August 26, 2026 |
This page is the standing home for this market rather than a one-off, and it gets re-scored when the story moves. When September's ladder opens, this URL becomes September's board and August drops into the track record. The graded result of every number above goes onto the scoreboard either way.
- How High Will Bitcoin Go In August 2026? Kalshi Vs. AI
- Where Does Bitcoin End 2026? Kalshi's Price-Band Odds
- Bitcoin Price Prediction 2026: 130 Days To Beat Jan.
- Bitcoin 250K By 2027? Kalshi's Longshot Just Woke Up
- How Bitcoin Price Contracts Work On Kalshi
The Bottom Line
August's floor was set in the first three days of the month, at a level Bitcoin touched twice and left thirty-three cents apart, and the 26% squeeze that followed has been the market agreeing ever since. Eight contracts sit at the exchange minimum with nobody bidding. Our panel's first-round answer on the shallowest of them is 0.42%, its revised answer is 0.14%, and the gap between those two numbers is not about Bitcoin at all.
That gap is the useful part. A market quoted in whole cents cannot tell you the difference between a one-in-235 shot and a one-in-714 one, and neither can a price chart. The difference lived in a settlement clause and in a PDF, and one model on this panel would not stop reading it. The August question that still has two sides to it is the ceiling, and the ceiling board is where that one is being argued.
