The War of the Oracles: Prediction Markets vsTechnical Analysis | 30-09-2026
The War of the Oracles:
Prediction Markets vs
Technical Analysis
History · Evidence · Head-to-head tests · Live divergences · 18-month outlook
This article asks one question: when the two oracles, prediction markets and technical analysis, disagree about what happens next, which one should a trader believe?
Prediction markets, such as Polymarket and Kalshi, let people bet on a yes-or-no event: “Will the Fed hike in October?” or “Will bitcoin reach $100,000 this year?” The price of the bet is read as a probability. Price action is what the market itself is doing: the moves on the chart, the trend, support and resistance, and the flows behind them. Technical analysts read it to judge where price is likely to go.
Both claim to see the future. We compare them using the research, eight real events, a like-for-like test, and six markets where they disagree today.
Every trader who ever lived wanted the same thing
Amsterdam, 1688. On the stock exchange near the Dam, merchants shout prices for shares in ships that are still months from home. Joseph de la Vega watches them and writes the first book ever published about a stock market. He finds rumour, leverage, short sellers and syndicates, and one obsession that unites them all: to know tomorrow’s price today.
Three and a half centuries later the obsession has not changed, only the machinery. Rice traders in Osaka drew the first candlesticks to read the battle between buyers and sellers. Wall Street brokers printed daily odds on presidential elections decades before the first opinion poll. Today, Kalshi and Polymarket put a live probability on almost anything, from a Fed decision to a ceasefire, while trillions of dollars a day write their own verdict on the charts.
From that one obsession grew two oracles. The first is the prediction market, heir to the wager and the insurance premium. It speaks in plain numbers and answers one question: will it happen? The second is technical analysis, heir to the trading floor and the candlestick. It never states a forecast. It reads what money is actually doing in price action. For four centuries they have competed to see tomorrow first. This is the story of that war, and the answer to the question that matters to anyone with capital at risk: when the oracles disagree, which one should you trust?
Wagers, candles and coffee houses
Amsterdam’s options (1688), John Law’s bubble (1719), Osaka’s rice futures (1730), Wall Street’s election odds (1868–1940). Both tools were born in the same markets, from the same hunger.
The crowd meets the chart
Prediction markets now trade around $50 billion a month. Global currency markets trade $9.6 trillion a day. On 28 September both put October Fed hike odds at about 70%. On bitcoin, oil and gold they disagree.
Tomorrow, already priced
A China trade deal, the US midterms, a Ukraine ceasefire, bitcoin at $150,000, a 2027 recession: the betting markets have already put odds on the next 18 months. Section 10 sets them out and shows where the charts disagree.
The night that frames the question: Brexit, 23 June 2016
At 10pm in London the polls closed on the EU referendum. On Betfair Exchange, the world’s largest betting exchange, Leave was trading at roughly a one-in-ten chance. On the currency screens, sterling was climbing towards $1.50 as traders bet that Britain would stay.
By dawn the results from Sunderland and the Midlands had changed everything. Leave won. Sterling fell to about $1.32 within hours, its biggest one-day fall since floating exchange rates began in the 1970s. The betting favourite had lost, and the currency market had been caught leaning the wrong way. A one-in-ten chance is supposed to happen one time in ten, yet almost nobody was positioned for it.
But not every screen was calm. For weeks the price of insurance on sterling, the cost of options that pay out on a big move, had been climbing to extreme levels. One corner of the market was quietly preparing for the outcome that almost nobody expected.
That night holds the whole contest in miniature. The betting odds gave a probability. The sterling chart showed whether money was accepting or rejecting the risk. The options market showed what it cost to be wrong. Which of these sees the future first, and can a trader combine them into something better than either?
What a prediction market measures, and what price action measures
They look like they answer the same question. They don’t, and most confusion between them starts here.
Prediction market
Asks: Will a specific event happen by a specific date?
Gives you: a probability, e.g. 34 cents = 34% chance.
Who trades: self-selected bettors, mostly small tickets.
Money involved: small next to the underlying markets.
Best at: binary events with clear rules: elections, Fed decisions, deals.
Price action
Asks: nothing. It records what buyers and sellers actually do.
Gives you: a path (trend, momentum, support and resistance), not a probability.
Who trades: everyone: banks, funds, hedgers, central banks, algorithms.
Money involved: trillions a day.
Best at: continuous moves and sudden shocks.
A prediction-market contract pays $1 if the event happens and $0 if it doesn’t. If “bitcoin touches $100,000 before 31 December” trades at 34.5 cents, the market is saying there is a 34.5% chance. The question is narrow, and the answer is a single number.
A chart is different. Nobody asks it a question. It is the combined result of millions of decisions, many of which are not forecasts at all but hedging, rebalancing or forced selling. Technical analysis reads that record for clues: is the trend up, has price broken a key level, is momentum fading? It gives a map of conditions, not a probability.
This is the core trade-off. On a question like “will the Strait of Hormuz reopen by 31 October?”, a prediction market is the only place that gives a direct probability. But that contract traded about $67,000 in a day, while the oil market it is trying to forecast trades billions. The prediction market is clearer, but it is also much easier for one large account to move.
Not all prediction markets are the same
“Prediction market” is a category, not a single market. A 62-cent price on one platform is not the same observation as 62 cents on another, because each has its own traders, rules, fees and way of deciding the result. Before comparing any two probabilities, check where they come from.
| Venue | What it is | How to read its price |
|---|---|---|
| Iowa Electronic Markets (since 1988) | A small-stakes university research exchange, mainly on US elections. | The best long-run accuracy record of any prediction market, but tiny stakes and a narrow set of questions. |
| Betfair Exchange (since 2000) | A UK peer-to-peer betting exchange where users back or lay outcomes. | Odds reflect what users will actually match, before commission, and depend on how much money is waiting on each side. |
| Kalshi (US-regulated since 2020) | Yes/no event contracts regulated by the CFTC, including Fed, economic and market questions. | A 70¢ price is a probability only after you read the exact threshold, time window and settlement source. |
| Polymarket (since 2020) | A crypto-settled global exchange covering politics, crypto, macro and current events. | Its prices follow its own order book and resolution process, so they can differ from Kalshi on the same question. |
Price action has identities too. A CME oil future, a bitcoin spot market and a Nasdaq future are not prediction markets just because traders use them to forecast. They carry leverage, hedging, margin calls and sometimes physical delivery, and those forces move prices as much as opinion does.
What people say vs what money actually does
Traders often say, “People say one thing and do another.” That idea explains most of the difference between the two tools. Think of it as three levels of evidence:
There is a catch. Much of the money behind price action is not trying to predict anything. A pension fund rebalancing at quarter-end, or a gold miner selling future output, moves the price without forecasting it. Price action has far more money behind it, but a smaller share of that money is actually making a forecast. Prediction markets have less money, but almost all of it is trying to predict. Neither is automatically better. It depends on the question.
400 years of pricing tomorrow: where both tools came from
Prediction markets and technical analysis feel like modern inventions. They are not. Both grew from the same root: people disagreeing about the future and putting something at stake. The history explains why each works the way it does today.
Amsterdam, 1602: the future starts trading before it happens
When the Dutch East India Company (VOC) sold shares to the public in 1602, Amsterdam did something new. It made a distant, uncertain venture tradable every day. Ships were months away, cargoes could sink, and wars could wipe out a monopoly, yet the share could change hands today at a price that moved with every rumour.
By 1688, Joseph de la Vega, an Amsterdam merchant, had written Confusion of Confusions, the first book about a stock exchange. He described forwards, calls, puts, short selling, margin trading and groups of traders trying to push prices down. Because a single VOC share was so expensive, traders invented small contracts that let them bet on its direction without owning it. That was an early version of the small yes/no bet that Kalshi and Polymarket sell today.
“I really thought that I was at the construction of the Tower of Babel when I heard the confusion of tongues and the mixture of languages on the stock exchange.” — Joseph de la Vega, Amsterdam, 1688
Both traditions are already visible here. One asks, “What is this outcome worth if it happens?” That is the prediction-market question. The other asks, “What does the sequence of trades tell us about what comes next?” That is the technical analyst’s question.
Paris, 1719: when a rising price becomes its own proof
France was drowning in debt. John Law, a Scottish financier and gambler, offered a solution: a bank that issued paper money and a company, the Mississippi Company, with trading rights over the vast and largely unexplored Louisiana territory. Investors could buy shares with cash or with unwanted government bonds. Rumours of gold in Louisiana did the rest.
Shares worth about 500 livres in January 1719 reached roughly 10,000 livres by December, a rise of about 1,900% in under a year. Then the story broke. By September 1721 they were back near 500.
Law’s scheme was not a prediction market, but it teaches the most important warning for both tools. A price can feed its own forecast. A prediction-market quote that is rising attracts buyers because it is rising. A chart breakout attracts momentum traders because it is breaking out. In both cases the visible price becomes part of the story. A rising number is evidence of demand, not proof of being right.
Osaka, 1700s: the candlestick is born
In eighteenth-century Japan, rice was money. It paid taxes and samurai salaries. At the Dojima Rice Exchange in Osaka, officially recognised in 1730, merchants traded contracts for rice that had not yet been harvested. That was one of the world’s first organised futures markets.
The trader most associated with this world is Munehisa Homma, whose name is linked to the candlestick chart. The full legend is hard to verify, but the problem he solved is clear. A single closing price hid too much of the battle. Traders needed to know where rice opened, how high and low it travelled, and whether buyers or sellers won by the close.
One candle, two answers
A prediction-market contract on “does price touch $100?” settles Yes here. A technical trader sees a rejection: price reached $100, failed to hold it and closed near the lows. Both are right under their own rules. This is the difference between an outcome and a path, and it goes back to Dojima.
America, 1868–1940: elections become market prices
Long before opinion polls, Americans bet openly on presidential elections. Brokers on Wall Street called the odds, and newspapers printed them daily. Economists Paul Rhode and Koleman Strumpf studied these markets from 1868 to 1940. They found the betting odds forecast elections remarkably well, despite limited information and active attempts to rig them. The markets faded only when scientific polling arrived and other forms of gambling became easier.
Their study carries a second lesson that still applies. A market could correctly name the favourite while offering a bad price. A bettor who bought the favourite too late could still lose money. The forecast could be right and the trade wrong.
From Charles Dow to the yes/no contract
In 1865 the Chicago Board of Trade standardised futures contracts, so strangers could trade the same obligation without negotiating every term. Around 1900, Charles Dow’s editorials in The Wall Street Journal described primary trends and secondary reactions, the foundation of what became Dow Theory and modern technical analysis. Analysts later added volume, momentum and volatility. None of these tools created certainty, but they created a vocabulary for the path.
In 2008, US exchanges listed the first standardised binary options: contracts that pay a fixed amount if a price is above a level at expiry, and nothing otherwise. They joined a prediction market’s yes/no structure with a technical analyst’s price threshold. The retail online versions that followed were often abused, prompting US regulators (the CFTC and SEC) to warn about platforms that refused withdrawals and rigged payouts. It was a reminder that a simple-looking yes/no price can hide a poor deal.
What decades of research say about each
Prediction markets: accurate on average, biased on long shots
The best evidence comes from the Iowa Electronic Markets, a small-stakes university exchange. Researchers Berg, Nelson and Rietz compared its prices with 964 election polls from 1988 to 2004. The market was closer to the result 74% of the time.
There are two known weaknesses. First, the favourite–longshot bias: unlikely outcomes are priced too high, just as at a racetrack, especially when the result is months away (Page and Clemen). Second, single large accounts can move the price. On Intrade in 2012, one trader supplied about a third of all money bet on Mitt Romney (Rothschild and Sethi). In 2024 a French trader known as “Théo” bet around $30 million on Trump through Polymarket and reportedly won about $85 million.
Technical analysis: a real but modest edge, strongest in trend-following
Brock, Lakonishok and LeBaron (1992) found that simple moving-average and breakout rules had forecasting power on the Dow from 1897 to 1986. Later work by Sullivan, Timmermann and White showed that edge faded once the rules became widely known. Lo, Mamaysky and Wang (2000) found that classic chart patterns carry some useful information. A review by Park and Irwin found that 56 of 95 modern studies reported positive results, with caveats about costs and data-mining.
The strongest evidence is for trend-following. Moskowitz, Ooi and Pedersen found that trends persist across 58 futures and currency markets, and Hurst, Ooi and Pedersen found trend-following was profitable on average in every decade since 1880. In practice, around 90% of London FX dealers use technical analysis for short-term decisions (Taylor and Allen).
Two different scoreboards
This is why no study can simply declare one tool “more accurate”. The Iowa study measured how close forecasts came to vote shares. The Brock study measured whether trading rules earned more than chance. The targets are different, so the headline numbers cannot be compared directly.
| Prediction market | Technical analysis | |
|---|---|---|
| What it forecasts | Whether a defined event happens. | How price is moving towards or away from that event. |
| How it is scored | Calibration: do 30% events happen about 30% of the time? Measured with the Brier score. | Profit: hit rate, average win vs average loss, drawdown and trading costs. |
| Common trap | Treating 80% as certain. | Treating a high hit rate as success: a 75% win rate can still lose money if one loss wipes out a year of gains. |
A fair test needs a common target. For example, “Will Brent close at least 8% lower within ten trading days?” The prediction market supplies its probability. The chart signal is converted into a probability by checking how often the same signal hit that target in the past. Then both are scored on the same events. Section 06 runs two versions of this test.
Eight events: what the prediction market said, what price action said, and who was right
For each event we record the prediction-market reading, the price-action reading, what happened, and a verdict. Use the filters to group them.
Four lessons from the scorecard
1. When both agree, they have usually been right (2004, 2024).
2. When both agree and are wrong, the options market often gives the warning. Before Brexit, spot sterling rallied, but the price of sterling insurance was flashing red.
3. Price action wins on sudden shocks. There was no bet on SVB, and gold did not wait for the betting odds to catch up.
4. Getting the event right does not mean getting the trade right. The bitcoin ETF was approved and the price still fell 20%. Trump won in 2016 and stocks rallied anyway.
Lesson 4 is where traders get hurt most. A prediction market answers “will it happen?” A trader needs to know “is it already in the price, and what will the market do next?”
Head-to-head: two tests of prediction markets against price action
Test 1: do the prediction-market odds match the options market?
To compare the two fairly, both need to give a probability. Price action can do that through the options market. Option prices tell you how far traders expect an asset to move (its “implied volatility”). From that you can calculate the chance of bitcoin touching any price before a deadline. This is still price action, just translated into odds by professional traders with large money at stake.
Bitcoin is the ideal test. Polymarket runs a contract on how high and how low bitcoin will go in 2026, with $71 million traded. On 28 September, bitcoin options expiring in December implied 37.3% volatility, with bitcoin near $83,476. The chart compares the two sets of odds.
| Bitcoin touches… | Polymarket | Options market | Ratio | What it means |
|---|---|---|---|---|
| $90,000 | 67.0% | 68.9% | 0.97× | They agree |
| $100,000 | 34.5% | 33.7% | 1.02× | They agree |
| $120,000 | 9.5% | 5.4% | 1.8× | Bettors overpay |
| $150,000 | 3.5% | 0.2% | ~19× | Lottery-ticket pricing |
| $70,000 | 33.5% | 34.9% | 0.96× | They agree |
| $60,000 | 14.0% | 7.9% | 1.8× | Bettors overpay |
Finding 1: near today’s price, Polymarket is not making a forecast. It is copying volatility. Its 34.5% for $100,000 is almost exactly what the options market gives with no view on direction. The bullish technical signals (record ETF inflows, the September golden cross) are not in either number. If you trust those signals, neither market is paying you for them yet.
Finding 2: on far-away targets, bettors overpay for long shots, exactly as the racetrack research predicts. A 3.5% price for $150,000 tells you more about how people like to gamble than about bitcoin.
Test 2: can the two be combined? A worked backtest from 2024
The first test compared probabilities. This one asks the question traders care about most: does combining a prediction market with a chart signal make money? It uses a real, resolved Polymarket contract, “Will Bitcoin hit $100k in 2024?” This contract resolved Yes when bitcoin crossed $100,000 in December 2024. It is paired with bitcoin perpetual futures on Binance.
| Rule | How it works |
|---|---|
| Technical only | Buy only when price is above its 20-day average, the 20-day is above the 50-day, and price breaks its previous 20-day high. |
| Prediction market only | Hold bitcoin whenever Polymarket’s “Yes” price is 50¢ or more. |
| Combined | Buy only when both agree. Position size grows with the Polymarket probability. |
| Rule | Total return | Worst drawdown | Days in the market | Win rate when active |
|---|---|---|---|---|
| Technical only | +14.32% | −4.18% | 4.8% | 58.3% |
| Prediction market only | +2.82% | −6.98% | 11.2% | 53.6% |
| Combined | +1.04% | −0.60% | 1.2% | 66.7% |
What it shows. The prediction market was right: bitcoin did hit $100,000 in 2024. Yet holding bitcoin whenever the odds were above 50% was the weakest strategy, with the worst drawdown. A correct forecast about the destination did not make every day on the road a good day to be long. The chart rule earned the most, because it only entered when the trend was confirmed. The combined rule earned less but cut the worst drawdown to just 0.6% and won two trades in three. It worked as a filter, not an oracle.
What happens to both tools when a crisis hits
On a normal day, both tools work as described. In a crisis, the rules change. Order books empty, prices jump over levels without trading at them, and the last price on the screen may be one that nobody could actually trade at in size.
The clearest example is 20 April 2020. The May contract for US crude oil (WTI) traded as low as −$40.32 a barrel and settled at −$37.63. Oil everywhere had not become worthless. The expiring contract required delivery in Cushing, Oklahoma, storage there was almost full, and holders who could not take delivery had to pay others to take it. No trend indicator could have forecast a price below zero, and no prediction market was asking the question. The decisive signal was the contract’s plumbing: expiry, storage and who was forced to sell.
Price action in a crisis
Still the fastest signal, but harder to read. Candles stretch, indicators calculate from prices that could not be traded in size, and “oversold” only describes the past. It does not promise a bounce.
Prediction markets in a crisis
Prices can spike, freeze or look strangely cheap because a few traders want out at any price. A contract can even settle correctly for the wrong reason, on a brief spike or an odd closing print.
| Normal market | Crisis | What it means for traders |
|---|---|---|
| Deep order book, tight spreads | Spreads widen, depth disappears, prices gap | The chart is still informative but harder to trade and easier to misread. |
| Indicators update smoothly | Moving averages and RSI lag the break | Indicators describe the shock after it starts rather than predicting it. |
| Traders can hedge or exit | Forced selling and margin calls dominate | Prices reflect who must trade, not what anyone believes. |
| Betting odds fairly stable | Odds spike, freeze or go stale | Check the spread and whether you could actually exit before trusting the number. |
The lesson is not that either tool fails. It is that in a crisis, the thing being predicted changes. Markets stop forecasting fundamentals and start forecasting who will be forced to trade next. At those moments, the most valuable information is structural: expiry dates, margin, storage and how the contract settles.
When to trust the prediction market, and when to trust price action
Prediction markets tell you what people expect. Price action tells you what money is doing. Options tell you what it costs to be wrong.
Six markets where prediction markets and price action disagree right now
Data from 28–29 September 2026. Each case sets the prediction-market odds against the price action, applies the guide above, and gives a view on which side may prove right. These are research views, not trade instructions. Odds on small contracts can go stale quickly, so re-check before acting.
Prediction markets expect no deal. The oil futures curve expects relief.
- Hormuz back to normal traffic by 31 Oct: 8.5% (Polymarket, 25 Sep)
- US–Iran final nuclear deal by 31 Dec: 15% ($18.9m traded, 28 Sep)
- Saudi East–West pipeline restart by 31 Oct: 75%
- Brent up about 75% this year
- November trades more than $7 above December, up from under $1 a month ago
- Trump rejected Iran’s reopening proposal on 28 Sep; Brent rose again
The futures curve slopes sharply down, with later months much cheaper than today. That usually means the market expects supply to return. But bettors give a diplomatic deal very little chance this year, and Iranian officials have reportedly been pessimistic about a deal before the US midterms in November.
Both can be true. A steep curve often reflects how scarce oil is today rather than a forecast. It can also reflect oil arriving by other routes: the Saudi pipeline already moves about 3.5 million barrels a day, and bettors give a full restart 75%. So the curve may be pricing workarounds, not peace.
Here they fully agree, and that has its own risk
- Kalshi ~69%, Polymarket 68.5% for a 25bp hike on 28 Oct
- On 25 Sep Kalshi was at 66% while futures priced ~55%
- 91% odds of at least one more hike before 2027
- 10-year yield 5.26%, a 19-year high; 2-year near 4.87%
- Gold −3.2% on 28 Sep; dollar firmer
- Bond volatility rising; stock volatility (VIX 16) calm
This is the control case: both sides agree almost exactly, and bonds, gold and the dollar are all moving as a hike would imply. Notably, the prediction market moved first. Kalshi was above futures pricing on 25 September, and futures caught up by the 28th.
Rule four applies. Bond volatility is rising, but stock volatility is low. If the next inflation or jobs data come in soft, both could be wrong together, and stocks are not paying much to insure against that.
Prediction markets say 35% by 2027. Stocks say no. Bonds are worried.
- US recession in 2026: 8.5% (Polymarket, 29 Sep)
- US recession by end-2027: 35% (24 Sep)
- Q2 GDP slowed to 1.5%; Q3 data due 29 Oct
- S&P 500 at 7,683.69 vs record close 7,798.99 (13 Aug)
- But 10-year yield at a 19-year high; 30-year above 5.3%
- Oil up 75% this year, the Fed hiking, mortgages at 7.03%
First, the fine print: the 2026 bet has only three months left and recessions are declared late, so its low number mostly reflects the calendar. The meaningful figure is 35% by end-2027.
Price action is split here. Stocks are calm and near their highs. Bonds are not: long-term yields are at a 19-year high while the Fed raises rates into an oil shock. That combination came before the recessions of 1973, 1980, 1990 and 2008. On 28 September, more than 65% of US stocks fell.
Money is flowing in. Prediction markets are not pricing it.
- Touch $90k: 67% · $100k: 34.5% · $70k: 33.5%
- Options market: 68.9% · 33.7% · 34.9%
- In other words: no directional view at all
- $2.4bn into spot bitcoin ETFs in the week to 25 Sep, the most since Oct 2025
- Golden cross (50-day above 200-day) on 11 Sep
- But rejected at $87,270 on 21 Sep
Bettors and the options market agree with each other, and both price bitcoin as a coin toss. The flows and trend are bullish. Against that, a 19-year high in bond yields is a real headwind for an asset that pays no interest, and price has already failed once near $87,000.
The chart has broken down. Prediction markets still expect a rebound.
- Kalshi: gold at or above $4,300 at year-end 46%; $4,400 29%; $4,500 20%
- A neutral options-style estimate for $4,300 is about 35%
- On 1 Sep bettors gave $5,000 by year-end slightly better than even odds
- Record close $5,419.83 on 28 Jan; now down ~23%
- 2026 gains erased; seven-week low; −3.2% on 28 Sep alone
- Rising yields: 10-year at 5.26%, Fed hiking
At $4,400 and $4,500 the betting odds are close to neutral. At $4,300 bettors are about ten points more optimistic than a neutral estimate. That may be real dip-buying, or it may be a stale price that has not caught up with Monday’s 3% fall. The bettors’ record this month is poor: they still saw $5,000 as a coin toss when the chart had already turned down.
Prediction markets were confident in August. The chart has stalled since.
- Kalshi: 66.9% for the S&P hitting 8,000 in 2026 (6 Aug)
- Year-end brackets now: 7,800–7,999 at 20%, 8,000–8,199 at 18%
- The index needs +4.1% from 7,683.69
- No new closing high since 7,798.99 on 13 Aug
- Still above its 50- and 200-day averages
- Rising yields pressing on valuations
A neutral volatility estimate puts the chance of touching 8,000 by year-end at about 62%, close to the August betting odds. Bettors’ year-end brackets cluster around 7,800–8,199, which implies a modest year-end rally. The chart shows an uptrend that is stalling as yields rise.
| Market | Prediction market says | Price action says | Agree or disagree? | Likely right |
|---|---|---|---|---|
| Oil / Hormuz | No deal this year | Futures curve expects relief | Disagree | Prediction market |
| Fed October | 69% hike | 70.3% hike | Agree | Both |
| Recession | 35% by 2027 | Stocks calm, bonds stressed | Disagree | Prediction market + bonds |
| Bitcoin | No direction | Bullish flows, capped at $87k | Disagree | Price action, if $87.3k breaks |
| Gold | Expects a rebound | Downtrend | Disagree | Price action |
| S&P 8,000 | Modest rally | Uptrend, stalling | Unclear | Wait for trigger |
The next 18 months, as priced by prediction markets
Prediction markets have already put odds on the events that will shape markets until early 2028. Read together, they describe a clear picture: interest rates stay high, the US economy runs hot, trade peace with China arrives, peace in Ukraine comes slowly, and the Middle East stays unresolved for longer than the oil curve assumes.
These are the crowd’s odds, not Capital Street forecasts. Long-dated contracts are also where the research finds prediction markets least reliable. The favourite–longshot bias is strongest far from the resolution date, and many of these contracts are thin. Treat them as a map of what the market already expects, and so of where the surprises would be.
What the priced future means for traders
What comes next: AI, more contracts, and the end of the either-or debate
For four centuries the raw material has not changed: price, volume, news, rumour and the record of other people reacting. What is changing is the reader. A large language model can read a central bank statement, a chart, an options market and a prediction-market rulebook in seconds, and turn them into a list of conditions and a proposed trade.
That makes AI less a new oracle than a fast translator between the two worlds. It can turn a contract that says “above 91.49 at settlement” into a chart checklist: what price must do, by when, and what would make the current odds stale. But early research is cautious. A 2026 study of five AI models found promising simulated results alongside persistent number errors and weak performance in sideways markets. A 2026 review of 77 AI-trading studies found that few reported trading costs or tested on truly unseen data. There is also a new risk: if many AI agents read the same pattern the same way, they could all trade in the same direction at once.
More, narrower contracts
Markets are splitting broad questions into thresholds, ranges, closes and touches. That can reveal the whole distribution of expectations, but also thin books and false precision.
Rules become data
Exchange data feeds now publish settlement rules, timestamps and order books. For the first time, event odds and price paths can be compared systematically rather than by screenshot.
The future is not one tool replacing the other. It is a stack. A Kalshi contract prices whether the Fed hikes. Rate futures price the path of rates. Options price the cost of protection. And the currency chart shows whether money is accepting the new story. The traders who do best will be the ones who know which layer answers which question.
So which is the better predictor: prediction markets or price action?
Neither wins everywhere, and knowing which one to use for which question is the edge.
Prediction markets are better on clear, scheduled yes-or-no events. They beat the polls in 2024, and they moved ahead of futures on this month’s Fed odds.
Price action is better on sudden shocks and continuous moves where nobody has written a bet. It saw SVB first, it called the 2023 stock recovery, and it has led gold all September.
On price targets, the bitcoin test shows prediction markets mostly repeating what the options market already knows, while overpaying for long shots.
And on “people say one thing and do another”: betting beats talking, and large institutional money can see what small bettors miss. But much of that large money is hedging, not predicting, so a chart needs careful reading.
Four hundred years after Amsterdam, the lesson is unchanged. Prediction markets descend from the wager: they price the destination. Price action descends from the trading floor and the candlestick: it shows the road. The trader’s job is not to pick one. It is to let them challenge each other, and to check whether they are looking at the same journey.
When prediction markets and price action agree, listen. When they disagree, check which one is answering your actual question. When they agree and insurance is cheap, check your exits.
Tomorrow will always be priced before it arrives. But the price of tomorrow is never the same as the path taken to reach it.
Sources and notes
- Kalshi News, “Fed October rate hike odds spike to 66%”, 25 Sep 2026; DeFi Rate Fed decision odds tracker (Kalshi ~69%, Polymarket 68.5%); USAGOLD daily report, 28 Sep 2026 (FedWatch 70.3%, gold $4,148.69).
- TheStreet, “Stock Market Today”, 28 Sep 2026 (index closes, 10-year yield, Trump rejects Iran proposal); Saxo Options Brief, 29 Sep 2026 (10-year 5.255%, VIX 16.07, MOVE 101.82); Tenbrief, S&P 500 record close 7,798.99 on 13 Aug 2026; Advisor Perspectives S&P snapshot, 18 Sep 2026.
- FinanceFeeds, “Brent rose 3.4% as Hormuz deal talk hit”, 26 Sep 2026 (Polymarket Hormuz and pipeline odds); Rigzone, “Oil Edges Higher as Supply Tightens”, 28 Sep 2026 (Brent curve, prompt spread, Saudi pipeline); Benzinga, Trump–Iran sanctions relief and Polymarket deal odds, 28 Sep 2026.
- CryptoSlate Polymarket data, “What price will Bitcoin hit in 2026”, 28 Sep 2026; bit.com research, “Bitcoin Stalls at $83,000”, 28 Sep 2026 (DVOL, December ATM IV); Benzinga and The Block on ETF inflows, 26 Sep 2026; Yahoo Finance crypto prices, 29 Sep 2026.
- Kalshi “Gold price at year end” market (retrieved 29 Sep 2026); Mining.com and Yahoo Finance, 1 Sep 2026 (gold record $5,419.83, Polymarket gold odds, Fed hike odds after Jackson Hole).
- CryptoSlate, Polymarket US recession 2026 (29 Sep 2026); PredictionNews, recession by end-2027 at 35% (24 Sep 2026).
- Kalshi 8,000 odds via CNBC / Bloomingbit, 6 Aug 2026; Kalshi S&P year-end range market (retrieved 29 Sep 2026).
- BIS Triennial Central Bank Survey, April 2025 ($9.6tn/day FX turnover); Pew Research Center, prediction-market volumes, 23 Sep 2026.
- Berg, Nelson & Rietz (2008), “Prediction market accuracy in the long run”, International Journal of Forecasting; Page & Clemen (2013), Economic Journal; Rothschild & Sethi (2016), “Trading strategies and market microstructure: evidence from a prediction market”, Journal of Prediction Markets; Snowberg, Wolfers & Zitzewitz (2007), “Partisan impacts on the economy”, Quarterly Journal of Economics.
- Brock, Lakonishok & LeBaron (1992), Journal of Finance; Sullivan, Timmermann & White (1999), Journal of Finance; Lo, Mamaysky & Wang (2000), Journal of Finance; Park & Irwin (2007), Journal of Economic Surveys; Moskowitz, Ooi & Pedersen (2012), Journal of Financial Economics; Hurst, Ooi & Pedersen (2017), “A Century of Evidence on Trend-Following Investing”; Taylor & Allen (1992), Journal of International Money and Finance.
- Joseph de la Vega, Confusion de Confusiones (Amsterdam, 1688); Beursgeschiedenis, “The story: over 400 years” (VOC shares, Amsterdam exchange, 1720 crisis); Finance Watch, “Lessons from history: not so innovative financial innovations”.
- Mississippi Department of Archives and History, “John Law and the Mississippi Bubble: 1718–1720” (share-price levels and collapse).
- Paul W. Rhode & Koleman S. Strumpf (2004), “Historical Presidential Betting Markets”, Journal of Economic Perspectives.
- Folger Shakespeare Library, The Merchant of Venice (literary context only).
- US SEC, CBOE binary-options approval order (2008); CFTC/SEC Investor Alert, “Binary Options and Fraud”; Investor.gov, “Binary Options”.
- US EIA (2020), “Low liquidity and limited available storage pushed WTI crude oil futures prices below zero”; CFTC Interim Report on NYMEX WTI trading around 20 April 2020.
- Polymarket, “Will Bitcoin hit $100k in 2024?” market rules and CLOB price history; Binance USDⓈ-M futures kline data (backtest inputs from the Capital Street research record).
- “Evaluating Large Language Models for Technical Market Analysis”, arXiv (2026); “Agentic Trading: When LLM Agents Meet Financial Markets”, arXiv (2026).
- Betfair Exchange, Kalshi and Polymarket venue documentation (back/lay mechanics, event-contract rules, resolution processes).
- DeFi Rate, 2026 midterm odds (Kalshi/Polymarket average, 29 Sep 2026); 2028tracker.com, presidential odds average (28 Sep 2026).
- TradingView / 99Bitcoins, Polymarket US–China tariff agreement odds (24 Sep 2026); PredictionNews, Polymarket Russia–Ukraine ceasefire odds (6–7 Sep 2026).
- Kalshi, “State of the economy at the end of 2026” and “Fed decision in July 2027” markets (retrieved 30 Sep 2026); DeFi Rate Fed tracker (January 2027 and December 2026 meeting odds).
- CryptoSlate, Polymarket “When will Bitcoin hit $150k?” (late Sep 2026).
Method notes. Options-implied probabilities use a driftless log-normal model with flat volatility; touch probability = 2·N(−|ln(K/S)| / σ√T). Historical figures for 2016–2024 events are rounded from widely reported market data. Verdicts are editorial judgements.
Important. This article is educational market commentary, not personalised investment advice. Prediction-market quotes on small contracts can be stale or thin; verify contract rules, settlement terms, liquidity and current prices before acting. Trading leveraged products carries a high risk of loss.
