What TAB actually charges

A bookmaker's board is not a set of predictions. It is a pricing structure, and it is built so that the house is paid out of every dollar that passes through it, whoever wins, whatever happens. That charge is real, it is large, and it is not written anywhere a customer can see.

New Zealand has one licensed bookmaker, so there is nothing to compare it against and no reason for it to publish the number. I measured it instead: 146,956 sports prices and 2,537 races.

The charge is inside the price

Every price implies a probability: 2.00 implies 50%, 5.00 implies 20%. Add up the implied probabilities of every outcome in a market and a fair book totals exactly 100%. Real books total more, and the excess is the charge.

Take a head-to-head priced 1.42 and 2.90. That is 70.4% and 34.5%, a total of 104.9% for a match with only 100% of outcomes in it. The extra 4.9% belongs to the operator, charged on everything staked regardless of the result.

Two consequences follow that customers are rarely told. It is taken from everything you stake, not from what you lose, so anyone who churns the same $100 twenty times has paid it twenty times. And it is not uniform. It varies by a factor of three depending on what you are betting, which is the whole reason it is worth measuring rather than assuming.

Here is the structure on one scale, cheapest at the top.

The sport, filtered by code

Sport is the competitive end, and the reason is not complicated: TAB's sports prices are quoted against every bookmaker on earth, so they cannot drift far without looking absurd. Bets with two possible results, a straight winner, a line, a total, cluster tightly near 7.0% across every sport measured.

The one that breaks the pattern is football with the draw in play. Adding a third outcome takes the same operator to 10.3%, half as much again, on the most-bet market in the most-bet sport. A third result is a third place to put the charge.

Pick a sport below to see its spread and the real markets behind it, cheapest to dearest.

The sports capture is a single full-board snapshot taken on 2026-06-14 and has not been regenerated since, margins move slowly, but it is a snapshot and is labelled as one. The market-by-market write-up is in the sports margin study.

And then there is the racing

Everything above is the part of the board that has competition on it. This is the part that does not. Across 2,537 races, the median fixed-odds book adds up to 123.9% , 3.4 times a sports head-to-head, from the same operator, on the same app.

Nothing about a horse makes it intrinsically harder to price than a rugby league match. The difference is that the sports book is quoted against the world and the racing book is quoted against nobody.

The tabs below regroup the same races three ways, and the table underneath is real races you can open. Watch the implied chances accumulate past 100% and the number stops being an abstraction.

This is not a handful of dear races pulling an average around

A median can hide a long tail, so here is the whole distribution. Nearly half of all races sit in a single five-point band, and the entire shape sits high. There is no cheap corner of the racing book to move to.

TAB's fixed price is worse than TAB's own pool

The same operator offers two ways to back the same horse, and they are not priced the same. The fixed-odds book runs at 23.9%; the tote pool on those same races runs at 18.6%.

The tote figure is unusual in that it is not an estimate. A pool pays out whatever is left after the deduction is taken, so the gap is the deduction, and nothing has to be inferred. It comes to 15.7%, which is roughly the published rate. That is the useful part: the method lands on a number that can be checked against a figure TAB does disclose, which is the best evidence available that the figure beside it is right too.

So the comparison holds. On the same race, at the same moment, the fixed price you are offered first is the dearer of the two prices the same bookmaker is willing to give you.

What a percentage looks like as money

Percentages are easy to shrug off. A season is not. Set your own spending below and the arithmetic will do the rest. This is what the house takes before anyone has been right or wrong about anything.

How this was measured

1. One race, one number. The total is worked out per race and summarised with a median, never pooled across runners. Racing books widen with field size, so a runner-weighted average would let a 20-horse field count twenty times and drag the headline toward whatever shape of race happened to be on.
2. Scratchings are excluded before the book is summed. The margin that matters is the one on the field that actually ran; counting a withdrawn runner inflates it with a price nobody could take.
3. Bets with two outcomes and bets with three are never mixed. Pricing a third result makes a book wider no matter who is doing it, so blending the two would inflate the sports figure racing is compared against, which would flatter our own comparison.
4. A handful of sports markets are misfiled, and rather than quietly leave them there: the June capture put seven cricket markets under football. They are hidden from the tables above, but the medians still include them, because the sports half has no generator to re-run. Measured on the sample, taking them out moves the two-outcome figure from 7.14% to 7.17% and the three-outcome one from 10.27% to 10.29%, inside the rounding either is is quoted at.
5. Partial books are dropped. A race is only counted when every runner that started carried a price. A book missing two runners understates the margin, and quietly.
6. The two halves have different dates, and we would rather say so than round it off. The racing figures regenerate from a rolling archive and cover 2026-09-03 to 2026-09-18. The sports capture is a single full-board snapshot taken on 2026-06-14; margins move slowly, but it is a snapshot and is labelled as one.

Why publish it

Partly because nobody else will. A single-operator market has no price comparison in it by definition, and a number that never gets quoted never gets argued with. Every figure here comes off a public board and anyone with the patience could repeat the whole exercise.

But mostly because the margin is the bar. Knowing it does not make anyone a winner, and this is the part it would be easy to overclaim, so we will not. A 24% book is not evidence that racing is rigged, and a 7% book is not an invitation to bet more sport. It is the headwind, and it says how much better than the board you have to be before you are even level.

That is the whole reason this project exists. Beating a bar that size by picking winners is not realistic; the only route over it is to bet the individual prices that are wrong, which means having something honest to compare them to. We price every TAB market against the sharpest books in the world, send the ones where TAB has it wrong, and grade every one of them in public, won or lost. You can read how that works, or start with what positive expected value actually means.