Strategies·November 17, 2026·8 min read

In-Game Sports Trading: Buying the Favorite After It Falls Behind

How live win probabilities reprice after a score, how to estimate a comeback, and why fees, delay and fast market makers make the obvious dip buy a losing trade.

in-game tradingsportslive bettingpolymarket
Kashima Stadium — illustrative image
Photo: Wikimedia Commons · CC0Source

The most tempting trade in live sports is the pre-game favorite that just fell behind. Its price drops from 70 cents to 28, the crowd decides the game is slipping away, and the comeback looks cheap. Sometimes it is. Most of the time the new price is closer to right than the old one, and the fee, the spread and the execution delay turn the obvious buy into a losing bet.

Live prices move because win probability moves. Win probability is a function of score, time, possession and the sport's scoring structure. A team down eight with six minutes left in a basketball game is not in the same position as a team down eight with six minutes left in a football game, even though both are down eight with six minutes left. The contract does not pay for the story. It pays for the remaining paths to a win.

Why the price moves so fast

Live models translate the current game state into a probability over the final result. In basketball, the key inputs are the margin, the clock, possession and the expected number of remaining possessions. In football, they are the margin, the clock, field position, timeouts and the two-minute offense. In soccer, they are the goal difference, the minute and the man advantage.

When a favorite falls behind, the price is not the old price minus the deficit. It is a new conditional probability over the remaining game states. A three-point deficit with a full quarter left is a small problem. A three-point deficit with 40 seconds left and no timeouts is close to over. The same score can be worth 40 cents or 4 cents depending on the clock.

The market is also fast. A 2026 study of Polymarket NBA markets reconstructed order-book states from more than 75 million snapshots across 173 games. It found only 7 executable single-market in-game mispricings, with a median duration of 3.6 seconds. The market is not perfect, but the simple version of the trade is competed away in seconds.

Score, time and possession

Three variables do most of the work. The margin sets the size of the hole. The clock sets the number of chances to climb out. Possession sets who gets the next chance and, in basketball, whether the clock stops.

A useful shortcut is to convert the deficit into scoring swings. If each team has about 12 possessions left and a possession is worth roughly one point, an eight-point deficit is roughly four stops-and-scores away from level. That is a real hole but not a fatal one. The same eight-point deficit with four possessions left is nearly over. The price should reflect the number of paths, not the raw margin.

Those numbers are illustrative, not a model. The point is that the relationship between margin and win probability is convex in time. A minute in a tied game is worth more than a minute in a 20-point game, because the trailing team changes strategy: fouls, three-point shooting, onside kicks, an extra attacker, a pulled goalie. The market's live model knows this. A trader who anchors on the pre-game price does not.

The comeback math, with fees

Suppose a team is priced at 28 cents after falling behind, and you think its true win probability is 25%. Buying 500 shares costs $140. The sports taker fee is 500 x 0.05 x 0.28 x 0.72 = $5.04, so the all-in cost is $145.04. If your 25% is right, the expected value is 0.25 x 500 - 145.04 = -$20.04. The price is too high even though your model is only three points away from the market.

Now suppose the price falls to 22 cents. The cost is $110 plus a fee of 500 x 0.05 x 0.22 x 0.78 = $4.29, or $114.29 all-in. The expected value at 25% is $125 - $114.29 = +$10.71, about 9.4% of capital. The break-even probability is 22.86%, not 22%. The fee alone is the difference between a trade and a pass, and it is less than one point wide at this price.

A one-cent spread on 500 shares costs $5, roughly the same as the fee. If you cross the spread and pay the fee, you need more than two points of edge before you break even. That is the whole argument for waiting for a better price instead of buying the first dip.

Why the market is often right

Live markets are not a crowd of casual fans. They are market makers with models, feeds and the ability to cancel. Polymarket publishes a sports WebSocket feed for live game status and scores, and professional desks consume that feed faster than a television broadcast. The person clicking after a touchdown is usually reacting to information that has already been priced.

The same NBA study makes the point quantitatively. Combinatorial inefficiencies were more common than single-market ones, with 290 episodes concentrated in the final minutes of games and a median return of about 101 basis points. But 76.9% of those opportunities were capped at an average executable size of 14.8 shares, and the theoretical middle jackpot was never realized. The edge existed. It was too small to scale at retail size.

So the realistic version of buying the favorite after it falls behind is not a free lunch. It is a claim that the market's live model is too pessimistic about a specific team in a specific state, and the claim has to be large enough to beat fees at a size the book can absorb.

Liquidity and latency

Live books thin out. Market makers widen spreads when the game is in flux, and depth at the touch can be a few hundred shares. A marketable order can walk several cents. Polymarket applies a configured delay to in-play orders in some sports markets. The order waits, cannot be cancelled, and can be rejected if checks fail when the delay expires. That is a real cost when the price is moving.

Pre-game resting orders are cancelled automatically at the official start time, so a strategy that relies on a stale bid surviving kickoff does not work. Tick sizes also vary. Some sports markets use 0.25-cent increments, and minimum order sizes differ by market. Check the book response before sizing, because a price that is valid in one market can be rejected in another.

How to execute a live entry

  1. Write the fair probability before you look at the price. Record the score, the clock and possession in the note.
  2. Compare it with the executable ask, not the midpoint, and add the taker fee at your actual size.
  3. Wait for the first 30 to 60 seconds after a score to pass. The immediate reaction is usually the worst price.
  4. Enter with a limit order at or below your fair value. Never use a market order in a live book.
  5. Size to the depth within one cent. If the visible size is smaller than your order, reduce the order.
  6. Plan the exit before the entry. Use FAK for exits where partial fills are acceptable, because FOK can fail when you need it most.
  7. If the position hedges a pre-game bet, size it to the hedge ratio, not to the same share count.

Risks

  • Model error. Your win probability can be wrong by more than the fee, and the market's can be right.
  • Information lag. An injury, a foul-out or a weather change may not be in your feed.
  • Delay and rejection. An in-play order can be held and then rejected after the price has moved.
  • Rule risk. Overtime, shootouts, shortened games and stat corrections are resolved by the rules, not by the broadcast.
  • Liquidity risk. The exit can be much worse than the entry, especially after a big play.
  • Tilt. The fastest way to lose in live trading is to chase a loss with a bigger position.

Checklist

  • Did I write the fair probability before checking the price?
  • Did I confirm score, clock and possession from a live feed?
  • Did I compute the fee at my actual price and size?
  • Is the depth within one cent enough for my order?
  • Is my order a limit order with a defined maximum?
  • Do I know the resolution rule for overtime and postponement?
  • Is my total exposure per game capped across all contracts?
  • Am I adding to a losing position for a reason other than a new thesis?

Sources

  1. 1.Discover Markets: Sports— Polymarket Documentation
  2. 2.Sports Channel— Polymarket Documentation
  3. 3.Fees— Polymarket Documentation
  4. 4.Risk-Neutral Pricing and Hedging of In-Play Football Bets— arXiv
  5. 5.Arbitrage Analysis in Polymarket NBA Markets— arXiv
  6. 6.Beating the Bookies with Their Own Numbers— arXiv

More from the journal

All articles