There's a kind of market where you don't trade stocks or coins — you trade 

. "Will it rain in Mumbai tomorrow?" "Will this team win the match?" "Will inflation come in above 5%?" You buy a contract that pays a fixed amount if you're right and nothing if you're wrong.

Because the price of each contract is really just the crowd's estimate of a probability, these markets occasionally do something delightful: the numbers stop adding up. And when the implied probabilities don't add up correctly, there's a way to make money no matter what actually happens. Let's build the idea from scratch.

. Take any yes-or-no question. The contract pays out a fixed amount — let's call it 

So the price of a YES contract tells you the market's opinion of the odds:

 the probability, expressed in rupees. That's the one idea everything else rests on.

 each contract pays ₹1 if it comes true and ₹0 if it doesn't, so its price is just the market's estimate of the probability.

And here's the structural fact that makes arbitrage possible: for a yes/no question, 

 of YES or NO must come true. One of them 

2. The foundational trade: buy both sides for less than ₹1

If exactly one of YES and NO is going to pay ₹1, then owning 

 guarantees you receive exactly ₹1 at settlement — no matter how the event turns out. The two contracts together are a ₹1 note with a delay.

So ask the obvious question: what if you can 

. At settlement, one of them pays ₹1. You just turned 97 paise into ₹1 — a locked-in 

, and you never had to guess the outcome. You were completely indifferent to whether it rained, who won, or what inflation did.

 if YES and NO together cost less than ₹1, buy both. One is guaranteed to pay ₹1, so you pocket the difference whatever happens.

This works because the probabilities are mispriced: 55% + 42% = 97%, but the 

 total has to be 100%. The market is implicitly claiming the two outcomes are less than certain to happen — which is impossible. That 3% "missing" is your edge. (The reverse also exists: if YES + NO add up to 

 than ₹1, the platform's mechanics sometimes let you do the opposite trade and sell the overpriced pair.)

The same question often trades on more than one platform at once, and they don't always agree. That disagreement is another way in.

Suppose an event is priced on two different venues:

Buy YES on A (53p) and NO on B (40p). Combined cost: 

 for a guaranteed ₹1 payout. One of your two contracts must win, and you've locked 

 — bigger than the single-platform gap, because two disagreeing crowds give you more room.

 if one platform prices an event higher than another, buy the cheap YES on one and the cheap NO on the other so your total is under ₹1.

Plenty of questions have more than two answers. "Which party forms the government?" "Who wins the tournament?" These markets list several outcomes that are 

Exactly one of those outcomes will resolve YES and pay ₹1. So if you buy YES on 

 and the total cost is under ₹1, you've again guaranteed yourself a ₹1 payout for less than ₹1.

Say a four-way market is priced: 30p, 25p, 22p, 18p. That sums to 

. Buy all four. One of them must win and pay ₹1, so you've locked 

 in a market with several outcomes where one must happen, buy them all. If the whole board costs under ₹1, you're guaranteed a profit.

5. The clever one: markets that contradict each other

This is the type that rewards a sharp, logical mind, because the mispricing isn't in one market — it's in the 

Some questions are logically linked, so their prices 

 obey a rule. When they don't, you've found free money:

 — sets of bets that are mutually inconsistent, letting you construct a guaranteed win. Spotting them is pure logic, not market prediction.

 some questions are linked by logic (if B is true, A must be true). If their prices break that logic, you can trade the inconsistency for a sure profit.

6. The catches — and the one that's unique to this game

Every arbitrage has leaks, but prediction markets have one risk you won't find anywhere else.

 Your "guaranteed ₹1" depends entirely on the event being settled 

. But who decides the answer? A rule, a data source, or a human arbiter — and the wording can be slippery. Markets have resolved in surprising ways because the fine print said something subtly different from what traders assumed, or because the outcome was genuinely ambiguous and went to a dispute. If a market resolves against the side you were relying on — or freezes in dispute for months — your "risk-free" ₹1 doesn't arrive. 

Always read exactly how a market resolves before assuming two contracts add up to ₹1.

 This single risk is why prediction-market arbitrage is far less safe than the clean maths suggests.

 That 3-paise edge looks great until you notice the event resolves in six months. Three percent over six months is about 6% annualised — and your capital is frozen the whole time, unavailable for anything else. The 

 return, not the headline edge, is what matters.

 Each platform takes a cut, and you're often trading two or more legs. On a thin edge, fees alone can wipe out the profit. And prediction markets are usually small — the juicy price is available only for a tiny size, so you can't deploy much before your own buying moves the price against you.

 Doing the two-venue trade means funding accounts on both, dealing with KYC, currency conversion, transfer delays, and the risk that a platform freezes or fails while holding your money.

 the maths says "guaranteed ₹1," but only if the market resolves cleanly. Disputes, fine print, locked-up capital, fees, and thin liquidity are where the safety leaks out.

7. The India reality: this one is mostly off-limits

For an Indian audience, honesty matters more than theory here, because the legal ground has shifted hard and recently.

Promotion and Regulation of Online Gaming Act (PROGA) in August 2025

, and its rules took effect on 1 May 2026 with a blanket ban on "online money games." Crucially, the domestic "opinion trading" apps that worked exactly like prediction markets were swept in: platforms like Probo and similar services let users wager on outcomes ranging from sports to elections, and regulators treated that as chance-based gambling. Probo shut its real-money operations in India after the government moved to ban real-money online games, following enforcement action that included a large freeze of its funds.

Offshore platforms aren't a clean escape either. Despite the ban, US-based venues like Polymarket and Kalshi kept allowing Indian users to sign up, which is precisely why authorities have been examining how to intervene, including against VPN-based access. Under most readings of Indian law — the Public Gambling Act of 1867 plus state rules — prediction markets have never been clearly classified, so without an explicit exemption they fall under the gambling prohibition. And even the crypto route carries its own tax: India applies a flat 30% tax on crypto gains plus a 1% TDS on transactions, which would hit any stablecoin used to fund such a platform.

The practical takeaway: as of mid-2026, running real-money prediction-market arbitrage from India is not a legal, accessible activity — the domestic platforms are gone and the offshore ones are prohibited. So treat this strategy as something to 

 — a beautiful piece of market logic — rather than something to deploy. (Play-money and forecasting platforms, which involve no stakes, are a different matter and a perfectly good place to study the mechanics.)

 real-money prediction markets are effectively banned for Indian users right now. This is one to learn from, not to trade.

The logic is simple and satisfying to code: add up what a guaranteed ₹1 costs, subtract fees, and flag anything you can buy for less. The discipline, as always, is in subtracting 

def complete_set_edge(yes_price, no_price, fee_pct=2.0):
    """
    Buy YES + NO. Exactly one pays 1 at settlement.
    Profit per 1 of payout, after fees. Positive = arbitrage.
    Prices are in rupees (0.55 = 55 paise).
    """
    cost = yes_price + no_price
    fees = cost * fee_pct / 100
    return round(1.0 - cost - fees, 4)

def whole_board_edge(yes_prices, fee_pct=2.0):
    """
    Mutually-exclusive, exhaustive outcomes (exactly one pays 1).
    Buy YES on all of them; profit if the board costs < 1 after fees.
    """
    cost = sum(yes_prices)
    fees = cost * fee_pct / 100
    return round(1.0 - cost - fees, 4)

# 1) single market, YES + NO mispriced low
print("YES+NO  :", complete_set_edge(0.55, 0.42))      # cheap pair -> arb
print("YES+NO  :", complete_set_edge(0.55, 0.46))      # too pricey -> loss

# 2) same event across two platforms (buy cheap YES on A, cheap NO on B)
print("2-venue :", complete_set_edge(0.53, 0.40))      # 93 paise total

# 3) four-outcome 'whole board'
print("board   :", whole_board_edge([0.30, 0.25, 0.22, 0.18]))


YES+NO  : 0.0106
YES+NO  : -0.0302
2-venue : 0.0514
board   : 0.031


The four lines tell the whole story: the first pair is a real (if tiny) edge; the second looks similar but the fees tip it into a 

; the two-venue version is fatter because two disagreeing crowds give more room; and the whole-board check generalises the same idea to many outcomes. A production version would also pull live order-book depth (so it knows the 

 price for your size), and — most importantly — read each market's resolution rules, because that fine print is what decides whether the "guaranteed ₹1" is actually guaranteed.

 the scanner just checks whether a sure ₹1 is on sale for less, after fees. The judgement the code can't fully automate is whether the market will resolve the way you think.

9. Why it's worth understanding even if you never trade it

Strip away the trading and prediction markets are a remarkable invention: a machine for turning scattered opinions into a single number — a live, money-backed probability. The arbitrage we've described is the cleanup crew that keeps those numbers honest, forcing the implied probabilities to add up to 100% and stay logically consistent. When you see a prediction market quote "73% chance," you're reading a price that arbitrageurs have pressured into coherence. Understanding how that pressure works tells you how much to trust the number — which is valuable whether you're forecasting an election, a policy, or a match, and regardless of whether you ever place a single trade.

Prediction-market contracts pay ₹1 if an event happens and ₹0 if it doesn't, so their prices are just implied probabilities. Because exactly one outcome must come true, you can sometimes buy a guaranteed ₹1 for less than ₹1 — by buying both sides of a yes/no market, by combining cheap halves across two platforms, by buying every outcome of a multi-way market, or by exploiting two markets whose prices contradict each other logically. The unique danger is resolution risk: the "guaranteed" payout depends on the market settling cleanly, and fine print or disputes can break that. And for Indian users specifically, real-money prediction markets are effectively banned as of 2026 — so this is a strategy to admire for its logic, not one to run.

Educational only, not investment advice, and not a suggestion to access any prohibited platform. Prediction-market and event-contract activity is restricted or illegal in many jurisdictions, including India; know your local law before doing anything.