How odds move and why favorites lose

Two things baffle every first-time toteboard watcher: the odds won't sit still, and the horse "everyone knew would win" keeps not winning. Both have the same explanation, and it's arithmetic, not conspiracy. This is the deep-dive.

If you've read how parimutuel betting works, you know the one-sentence version: in a parimutuel pool nobody quotes you a price — the pool is the price. This guide takes that sentence and squeezes everything out of it: why the board drifts, what the odds secretly say about probability, why the favorite losing is the system behaving normally, and why the numbers lurch one last time just as the gate opens.

Odds are a ratio, not an opinion

Start with a win pool of 1,000 points, of which 400 sit on the #4 horse. If #4 wins, the whole 1,000-point pool is divided among the people who staked that 400 — every point staked returns 1,000 ÷ 400 = 2.5 points. That's the original point back plus 1.5 profit, so the board shows 3-2. Nobody chose that number. It's not a prediction, not an expert's view, not the house's offer. It is the current ratio of two sums, updated every time anyone makes a pick.

Why the board drifts

Now let 200 more points land on #4 before the pool closes. The pool grows to 1,200; #4's share grows to 600. The payout per point falls to 1,200 ÷ 600 = 2.0 — even money. Notice what happened: other people agreeing with you made your own pick less valuable. That's the whole mechanism of drifting odds. Money arriving on a horse shortens its price; money arriving elsewhere lengthens it; and since money never stops arriving until the close, the board can't sit still — it isn't supposed to.

One rule follows that surprises everyone: you get the closing odds, not the odds you saw when you picked. Everyone who backed #4 gets 2.0 per point in the example above — including the early bird who picked when the board said 3-2. Parimutuel pools settle at the final ratio, full stop. This is why real bettors watch the board to the last second, and why your party's pool gets loud right before it closes.

The board is a probability poll

Flip the arithmetic around and the odds become a survey of what the crowd believes. If a horse's backers hold a fraction s of the pool, its payout is 1 ÷ s per point, which is odds of (1 ÷ s − 1) to 1. So board odds translate straight into the crowd's implied win probability: a horse at odds X-1 holds 1 ÷ (X + 1) of the pool. Even money means the crowd staked half the pool on it. 2-1 means a third. 9-1 means a tenth. Read a toteboard this way and it stops being a wall of numbers and becomes a live opinion poll with the sample size of everyone in the room.

In a points pool the poll is perfectly calibrated to itself: the implied probabilities sum to exactly 100%. Take a three-horse pool of 100 points split 50 / 30 / 20. The payouts per point are 2.0, 3.33, and 5.0 — board odds of 1-1, roughly 2.3-1, and 4-1 — and the implied probabilities are 50% + 30% + 20% = 100% on the nose. The pool always pays out exactly what went in, to the point. (Real tracks break this tidy identity; we'll get there.)

Why favorites lose: the part everyone gets wrong

Here is the fact that rescues every deflated party guest: "favorite" means most likely single outcome, not likely outcome. A favorite at 2-1 carries an implied probability of one in three. If the crowd has it priced about right, that horse should lose two of every three runnings — and the crowd still wasn't wrong, any more than a weather forecast of 33% rain is wrong on a dry day. The favorite is simply the least-unlikely runner in a race where every runner is unlikely.

Big fields make this brutal. In a twenty-horse Kentucky Derby, even a standout favorite rarely trades shorter than 2-1 — an implied one-in-three — because nineteen rivals, traffic trouble, and a mile and a quarter add up to a mountain of ways for the best horse to lose. So across any run of races, the parade of beaten favorites isn't evidence the crowd is bad at this. It's what correctly-priced uncertainty looks like, over and over, in public.

What would actually be strange is favorites winning most of the time. For that, the favorite would have to routinely carry more than half the pool — and the moment a crowd piles that deep onto one horse, the payout shrivels toward nothing, which chases the pool toward other horses and lengthens the favorite right back out. The parimutuel mechanism is self-correcting: it structurally resists ever making any horse a heavy odds-on certainty unless the crowd's conviction is overwhelming.

The favorite-longshot bias

One wrinkle in the poll deserves naming, because it's one of the most consistently documented patterns in the economics of betting markets: crowds tend to over-back longshots and under-back favorites, relative to how often each actually wins. The usual explanation is that a 30-1 shot is a lottery ticket — the story of the huge payout is worth more to most people than the arithmetic says it should be — while backing a short favorite feels like a lot of risk for a little thrill. You will see the same psychology at your own party: hand twenty guests a toteboard and somebody always empties their stash onto the silliest name on the board at 25-1. The implication cuts both ways. The favorite is usually a slightly better deal than its price suggests, and the extreme longshot usually a slightly worse one — but the longshot backer is buying entertainment, and at a points party, entertainment is the actual currency.

Why the odds jump at the bell

At real tracks the board famously lurches in the final minute — a 5-2 shot snaps to 8-5 as the gate opens. Two boring reasons: serious money waits until the last moment (why reveal your opinion early, and why accept more odds-drift than you must?), and the tote system recalculates in cycles, so the last flush of picks lands in one visible thump. Your living-room pool reproduces this in miniature. Guests hover, watch each other, and pile in at the deadline — partly strategy, partly theater. As host, embrace it: announce a hard close, count it down out loud, and let the final board be a small dramatic reveal. The mechanics of closing a pool cleanly are in the derby party guide.

What takeout does (and why your party math is cleaner)

One honest asterisk. At a licensed track, a slice of every pool — the takeout, commonly somewhere in the teens or twenties percent, varying by state and by pool type — comes off the top for the track, purses, and taxes before winners divide the rest. Rerun our first example with a 17% takeout: the 1,000-point pool nets to 830, so #4's backers get 830 ÷ 400 = 2.075 per point instead of 2.5. Do that to every horse and the implied probabilities no longer sum to 100% — they sum to about 120%, because the crowd as a whole is dividing less than it staked. That gap is the price of the racetrack's lights and litter crew. A PartyMutuel points pool has no takeout: the implied probabilities sum to exactly 100%, every pool pays out every point staked, and the whole system balances on a napkin. It is, in this one narrow sense, mathematically purer than Churchill Downs.

Using all this at your party

Three practical takeaways. Split when unsure: a stake divided across two horses is a perfectly legal way to say "one of these two" — real handicappers do it constantly. Hunt the neglected middle: the favorite-longshot bias suggests the value at most parties hides in the unglamorous second and third choices nobody's excited about. Never mock the favorite-backer or the longshot-lover: the first is playing the percentages, the second is buying the room's best story, and a good pool needs both. The full vocabulary for arguing about it is in the glossary.