How to do prediction market arbitrage with a 3-check gate

Prediction market arbitrage is building a fully hedged YES/NO position where the combined cost is below the fixed $1 settlement payout, or where selling both sides collects more than $1 in proceeds. The hard part of how to do prediction market arbitrage is not the math, it is proving contract equivalence and executing both legs fast enough that fees, slippage, and partial fills do not flip the trade negative.

Key Takeaways

• Binary contracts typically trade between $0.00 and $1.00 and settle at $1.00 for the winning side and $0.00 for the losing side, which anchors the core arbitrage math.
• A screenshot spread is not an edge until it passes a 3-check gate: settlement equivalence, executable depth at your size, and fee-aware net edge.
• Mispricings can disappear within minutes or seconds in liquid markets, so pre-funding and leg-risk sequencing matter as much as pricing.
• Cross-venue arbs like polymarket and kalshi only count when the contracts resolve the same way and access, fees, and transfers do not break execution.

How prediction market arbitrage makes money

A binary contract’s payout is fixed, so the only thing that moves is the entry price. In most prediction markets, a YES share and a NO share trade between $0.00 and $1.00 and settle to $1.00 for the winning side and $0.00 for the losing side. That creates the anchor: if a trader can buy complementary exposure for less than $1.00 all-in, the position settles to $1.00 regardless of outcome.

The screen-level translation is simple. Each market shows a price that traders treat as an implied probability. A YES price of $0.65 is roughly a 65% market-implied chance. Arbitrage ignores whether 65% is “right” and focuses on whether the complementary package is mispriced versus the $1.00 settlement.

Two core structures show up:

1. Buy-the-box: buy YES somewhere and buy NO somewhere else (or on the same venue) when the combined cost is below $1.00 after costs. 2. Sell-the-box: sell YES and sell NO when the combined proceeds are above $1.00 after costs, leaving a theoretical $1.00 liability at settlement.

The thesis that matters for trading prediction markets is that the math is easy and the money is operational. Arbitrage exists because liquidity is fragmented across venues, different user bases react at different speeds, and platform design differences and funding frictions slow convergence. Those frictions are also why “risk-free” setups break when the legs do not fill or the contracts do not actually match.

Arbitrage setups you can actually trade

The cleanest patterns are the ones that map directly to the $1.00 settlement anchor. The first is yes-no sum arbitrage inside a single venue: YES and NO briefly add up to less than $1.00 in a thin book, often during fast repricing. This is the easiest to reason about because settlement rules are identical by definition, but it is also the most competitive because it is mechanically detectable.

The second is cross-venue complementary arbitrage, the one most people mean by polymarket arbitrage or kalshi arbitrage. The setup is “same event, opposite side” across two platforms. It only counts if the contracts are truly complementary, not just similarly titled. Pariflow flags settlement-rule mismatch as a primary failure mode, and that is the right framing. A one-line difference in cutoff time, timezone, resolution source, or void rules can turn a hedge into two separate bets.

The third bucket is logically related markets, where two contracts are linked by real-world constraints. These can be hedgeable, but they are not the same as a complementary YES/NO box because the payout is not guaranteed by construction. They behave more like relative-value trades than pure arbitrage.

The trader angle is speed and microstructure. Benzinga notes that in liquid markets, mispricings can correct within minutes or even seconds, driven by automated traders. That is why “is polymarket-kalshi arb profitable” is usually a question about execution, not about whether the identity holds. If capital is not staged on both venues and fills are not reliable at size, the window is gone before the second leg is even placeable.

Fee-aware math and position sizing

Gross edge is the easy part: Gross edge = 1.00 − (price_1 + price_2). Net edge is what matters: Net edge = Gross edge − (fees + slippage + transfer or funding costs). Pariflow’s template is the right mental model, and it is also where most “free money” screenshots die.

A repeatable gate is to set a personal net-edge floor tied to execution confidence. Pariflow notes some traders use around 1%–2% net as a minimum threshold. The point is not that 1%–2% is magic. The point is that a 0.5%–1% gross edge is usually a no-trade once the venue’s fee schedule, maker taker pricing, and the expected slippage from crossing the spread are priced in.

Position sizing is mostly about matching payouts, not matching dollars spent. For a buy-the-box, the clean structure is one YES share and one NO share on the same event definition. That pair settles to exactly $1.00. The sizing question becomes “how many pairs can be filled at the modeled prices without moving the market.”

Two sizing checks keep people out of trouble:

1. Depth-at-size: the top quote is irrelevant if only a tiny amount is available there. The book needs enough cumulative size at your target price, plus a tolerance band for one or two ticks of drift. 2. Fee sensitivity: if the edge only exists when assuming maker fills but the order will likely execute as taker, the model is wrong. Maker taker matters because a “maker” limit order may not fill at all during a fast move, while a “taker” marketable order fills but pays the spread and often higher fees.

This is also where prediction market arbitrage bots show up. Automation is not required to understand the math, but it is the reason many small edges are not available to manual traders for long.

Execution workflow to reduce leg risk

Leg risk is the trade. One side fills, the other side moves, and the hedge becomes a directional position at the worst moment. The execution workflow below is built to reduce that failure mode and to force a fast “pass” when the setup is not real.

1. Prove contract equivalence before looking at price. Match the exact event wording, cutoff timestamp and timezone, resolution source, and void or cancellation rules. 2. Pre-fund both venues you plan to use. Transfer delays are part of the edge math because the window often closes while funds are in flight. 3. Check executable depth at your intended size on both legs. Look at cumulative size available at the price you need, not just the best bid or ask. 4. Compute net edge with conservative assumptions. Treat the fills as taker unless there is a clear reason they will execute as maker, and include expected slippage. 5. Hit the thin leg first with a strict limit. The thin book is the one that will gap away. If it does not fill quickly at your price, the trade is already deteriorating. 6. Hedge immediately on the deeper leg. The goal is to neutralize exposure fast, even if the second fill is a tick worse than the model. 7. Verify post-trade exposure and average fills. Confirm both legs are fully filled and the payout is balanced across outcomes. If one leg is partial, the position is not arbitrage. 8. Apply an abort rule if the second leg drifts beyond your edge. If the hedge price moves enough that net edge goes negative, the workflow needs a predefined response rather than improvisation.

This is the operational spine of news trading prediction markets. The “news” part creates the dislocation, but the P&L comes from whether the legs were executable at the modeled prices.

Settlement, platform, and access pitfalls

The most expensive mistake is assuming “same headline = same contract.” Pariflow and the practitioner discussion on Reddit both point to settlement mismatch as the hidden killer. Two markets can look identical and still resolve differently because one uses a different data source, a different cutoff time, or a different void policy when the event becomes ambiguous.

Cross-platform access is another constraint that turns theoretical arbs into non-trades. Benzinga flags regulatory access differences as a practical barrier, including that Polymarket does not serve U.S. customers while Kalshi operates under CFTC oversight. That matters because cross-venue arbitrage requires both legs to be executable by the same trader, with capital available on both venues.

Fees erase more trades than bad math. Benzinga gives an example that Kalshi fees can average around 1.2% but vary by product and conditions. Even if the exact schedule changes, the lesson is stable: small gross edges are fragile. A trade that looks like 1% on the screen can be negative after two sets of fees and a couple cents of slippage.

Platform and operational risk is not theoretical. Outages, withdrawal delays, and order matching quirks can strand a trader with one leg. That is why the 3-check gate starts with settlement equivalence and ends with executable depth and net edge. If any check fails, the correct response is to pass instantly, not to negotiate with the spreadsheet.

A beginner process for repeatable edges

A workable retail approach is specialization, not scanning everything. The Reddit practitioner heuristic is to look for relatively large positions, around $20K–$50K, in markets with around $100K total volume where a single participant can move price by multiple cents. That is a microstructure observation: thin books can be pushed off equilibrium by one motivated trader, creating temporary dislocations that are large enough to survive fees.

A beginner process that compounds skill looks like this:

1. Pick 2–3 categories and learn their settlement rules cold. This is where most “arbitrage” becomes real, because contract language and resolution sources vary by category. 2. Build a simple log that records the contract wording, cutoff time, your modeled net edge, your actual average fills, and the realized costs. 3. Start with small size and treat the first month as execution research. The goal is to measure slippage and fill rates by venue and time window. 4. Use a hard net-edge floor and raise it when execution is uncertain. Pariflow’s 1%–2% net threshold framing is a useful starting point for discipline. 5. Decide early whether automation is needed. Prediction market arbitrage bots and ai bots prediction markets tools are why liquid markets correct quickly, and manual traders need to operate where speed is less dominant.

This sits next to how to find good polymarket trades, not as a replacement for it. Arbitrage is one lane inside trading prediction markets, and it rewards process more than opinions.

The Take

I’ve watched people treat “YES + NO below $1” like a coupon code, then get clipped on the only two things that matter: the contracts weren’t actually equivalent, and the second leg didn’t fill at the screenshot price. The ugliest version is a Polymarket vs Kalshi lookalike where one contract’s cutoff is different by a timezone or the void rules diverge. That is not arbitrage, it is two separate settlement bets wearing the same headline.

The habit that keeps this sane is a hard 3-check gate every time: settlement equivalence first, executable depth at your size second, and fee-aware net edge third. If the thin leg can’t be hit immediately, or the net edge doesn’t clear a real floor after taker-style costs, I’ve learned to pass without negotiating with the market.

Sources

• Pariflow
• Benzinga
• reddit.com