The Cash-Out Edge: The Math Behind Selling Event Contracts Early Instead of Holding to the Final Out

A sportsbook bet has one exit: the final out. An event contract has an exit on every pitch.

That single structural difference is worth more than most bettors realize, and it is the strongest argument for moving your baseball action from DraftKings and FanDuel over to Kalshi, Polymarket, Robinhood Prediction Markets, or Coinbase. Not because the prices are better on day one (though they usually are — see MLB event contracts vs sportsbooks), but because you get paid for being right early instead of only for being right last.

This article builds the case with numbers: how often early leads actually survive, what those survival rates imply about fair mid-game prices, and the exact conditions under which selling beats holding.

The data: how often early MLB leads get blown

We pulled Retrosheet game logs for the 2015–2025 regular seasons (2020 excluded), 24,294 games with complete inning-by-inning line scores, and measured how often the team leading at the end of an inning went on to win the game.

Lead after 4 innings	Games	Leader wins	Lead is blown
1 run	6,716	66.3%	33.7%
2 runs	4,805	77.2%	22.8%
3 runs	3,133	86.5%	13.5%
4 runs	1,982	92.5%	7.5%
5+ runs	3,102	96.4%	3.6%

And the same cut at the end of the 6th and 7th:

Lead	After 6 innings	After 7 innings
1 run	72.0% hold	77.8% hold
2 runs	84.6% hold	88.6% hold
3 runs	91.8% hold	95.0% hold
4 runs	94.8% hold	97.5% hold
5+ runs	98.8% hold	99.3% hold

Two things jump out.

One: a 3-0 lead in the 4th is not a win. It is an 86.5% win — roughly one in seven of those games flips. That is not a rounding error; it is a coin that lands wrong 423 times per decade.

Two: the price of certainty is steep and it is paid late. Going from the end of the 4th to the end of the 7th only buys a 3-run leader 8.5 points of win probability (86.5% → 95.0%). You are risking 13.5% of your equity for three innings to earn 8.5 points of comfort. On a contract you can sell, that trade is optional. On a sportsbook ticket, it isn't.

The F5 version of the same problem

First-five bettors think they've solved this by shortening the exposure. They haven't — they've just moved the cliff earlier.

Taking every game where a team led at the end of the 3rd inning, and grading the F5 result:

Lead after 3 innings	Games	F5 win	F5 push	F5 loss
1 run	7,163	68.2%	14.0%	17.8%
2 runs	4,608	82.3%	8.4%	9.3%
3 runs	2,760	91.7%	3.7%	4.6%
4+ runs	3,858	97.2%	1.6%	1.1%

And after the 4th, with one half-inning of exposure left on each side:

Lead after 4 innings	F5 win	F5 push	F5 loss
1 run	79.2%	10.9%	9.9%
2 runs	90.3%	5.3%	4.4%
3 runs	95.3%	2.8%	1.9%

A one-run F5 lead through three innings is a 68% proposition with a 14% push branch — worse than most people's mental model of "we're winning." And for the record: a team leading after five innings wins the full game only 82.8% of the time (17,099 of 20,661). The F5 bet and the full-game bet are far less correlated than the bet slips suggest.

Turning survival rates into a sell decision

Here is the actual math. Let:

• p = your true win probability right now (from the survival tables plus game state)
• b = the current bid on your contract, in cents
• c = your entry price, in cents

Holding to settlement has expected value 100p cents per contract. Selling right now has expected value b cents, immediately, with zero variance.

Sell when b ≥ 100p. Hold when b < 100p. That's the whole rule.

Everything else is estimating p honestly and finding out what b actually is.

Worked example: the 3-0 lead in the 4th

You bought YES on the home team at 52¢ pre-game. They go up 3-0 through 4. The Kalshi book is now bid 88¢.

• True p from the table above: 86.5% — call it 87% with a decent starting pitcher still in.
• Fair value: 87¢. Market bid: 88¢.
• Sell. You are getting paid a penny above fair to eliminate a 13% chance of losing 52¢.

Realized: +36¢ on a 52¢ basis, a 69% return in four innings, with capital freed up for the next game. Hold instead, and your expected outcome is 87¢ — one cent worse — with a 13.5% chance of a total loss.

The same spot when holding is right

Same 3-0 lead, but the book is bid 80¢ because liquidity is thin and the only resting order is a market maker being cautious.

• Fair value 87¢, bid 80¢. You'd be paying a 7¢ haircut, roughly an 8% rake.
• Hold — or sell half. Never donate 7¢ to a thin book just to feel safe.

This is exactly the test to run against a sportsbook cash-out offer, and sportsbooks fail it constantly. A book offering cash-out on a 3-0 4th-inning position typically prices you at 78–82¢ of fair value on the dollar. The 8–15% haircut is the fee for an exit that event contracts give you at the spread.

Why the sell-early strategy compounds

The reason this beats hold-to-settlement over a season isn't that each individual sale is a big edge — the example above netted one cent. It's three structural effects.

1. Capital velocity. Money returned in the 4th inning can be deployed on the 7:05 slate. Money locked in a sportsbook ticket cannot. If you turn your bankroll over 1.6× per night instead of 1.0×, a 2% edge per position becomes 3.2% per night of bankroll.

2. Variance reduction improves your Kelly size. Kelly-optimal stake scales with edge divided by variance. Systematically stripping the last 13% of variance out of your winners lets you size up on entries with the same risk of ruin. Two bettors with identical picks and identical bankrolls do not earn the same amount if one of them can safely bet 30% more per position.

3. You get paid twice for the same read. Your pre-game edge earns the move from 52¢ to 87¢. That's the part your model is actually good at — pricing the game before first pitch. Innings 5 through 9 are a different game with different skills (bullpen leverage, pinch-hit sequencing, umpire fatigue) and most bettors have no edge there at all. Holding through them is unpaid exposure.

When not to sell

Selling is not automatically correct. Hold when:

• The bid is materially below fair value — thin books on Polymarket niche markets and any sportsbook cash-out offer usually are.
• Your edge is still live. If you bought because the market misprices a bullpen and that bullpen hasn't pitched yet, the thesis hasn't paid out. Selling now realizes only the price move, not the edge.
• Fees eat the trade. A round trip on Kalshi is small but not zero — see Kalshi fees explained. A 1¢ theoretical edge is not a trade after fees.
• You're up on a 5+ run lead. At 96.4% (after 4) and 98.8% (after 6), the remaining spread between the bid and 100¢ is too small to be worth the transaction cost. Just take settlement.

Using Heat scores as the sell trigger

Survival tables give you the baseline p. Live market movement tells you when the baseline is wrong.

The Heat score on the Live Signals board measures how hard, how fast, and across how many venues a price is moving. In a mid-game context we use it three ways:

• Heat confirming your side — price moving your way faster than the game state alone justifies. Hold; the market is finding something and you'll get a better bid.
• Heat reversing against your side — money coming the other way while you're still nominally ahead. This is the strongest sell signal on the board, because it usually front-runs a bullpen or injury update by several minutes.
• Heat collapse to flat — the move is fully priced. Fair value and market value have converged, which is precisely when selling costs you nothing.

Layer the ½-life estimate on top: when a move's half-life expires, the market has finished repricing, and every tick after that is noise you're holding for free.

A practical in-game framework

Position is in profit and the game is past the 4th?
├── Estimate p from the survival table + current game state
├── Read the live bid (Kalshi / Polymarket / Robinhood / Coinbase)
├── b ≥ 100p            → SELL 100%
├── b within 2¢ of 100p → SELL 2/3, let 1/3 ride to settlement
├── b < 100p by 5¢+     → HOLD, set a Heat-reversal alert
└── Heat reversing      → SELL regardless of the haircut

And the venue-selection layer, briefly:

Situation	Best venue
Mid-game exit on a marquee game	Kalshi (deepest game-day book)
Niche / milestone market	Polymarket
Fast exit, capital already parked	Robinhood or Coinbase
Full-game ML you cannot sell	Sportsbook — hedge on an exchange instead of cashing out

The bottom line

A 3-0 lead in the 4th inning is worth 86.5 cents on the dollar, not 100. A one-run F5 lead through three is worth 68 with a 14% push. Those aren't opinions, they're eleven seasons of Retrosheet.

Sportsbooks let you convert that equity to cash only at a haircut they choose. Event contracts let you convert it at the market price, on your schedule, as many times a night as you want. Over 162 days, the bettor who sells at fair value and redeploys will beat the identical bettor who holds every ticket to the final out — with the same picks.

Source: Retrosheet game logs, 2015–2025 regular seasons excluding 2020 (24,294 games with complete line scores). Cross-check against the Baseball Savant Game Strategy Explorer and Tango's win-expectancy tables.

What to read next

• Take-profit & cash-out strategy
• MLB event contracts vs sportsbooks
• Kalshi vs Polymarket
• Kalshi fees explained
• How Heat scores work