Editorial: Smoke and Mirrors - the Betting Exchange Shell Game

4 October 2002

Three weeks ago I wrote in IGN about the flaws in the bet exchange business model. Now it's time to look at how the exchanges are wallpapering over the gaps, how they have everyone fooled and why it's a very good business if you can get away with it.

A fortnight ago came news that sports betting software developer Orbis is launching a P2P betting system called "BetX." The new software is available to Orbis licensees as an add-in module to Orbis' OpenBet system or as a standalone system.

Orbis' systems don't come cheap, so where is the money going to come from to justify buying those BetX systems?

Betfair recently announced two new senior executive positions and said its staff numbers now exceed 100. How does it pay for all those people?

Is it smoke and mirrors?

Toss a Coin!

In the classic binary event of tossing a coin, if two people were to bet $10 each on the outcome of the toss a coin, one would lose $10 and one would win $10. Put another way, one would win 50 percent of the total stakes $20.

Simple. But if the coin was to be tossed 100 times with each time the two players betting $10 on the outcome, what would happen?

Without going to deep into probability theory, the most likely occurrence would be nil profit/loss, so each would win about 50 times. The total betting turnover would be $2,000, with both players having their original $10 in their pockets.

A betting exchange charges a commission on the punters' net win because it takes its commission (say 4 percent) every time the coin flips--but only from the winner, who thinks, "Wow, how easy was that? I made $9.60."

Using this example, though, every time the coin is tossed the "friend" in the middle holding the cash takes 40 cents. In 100 tosses of the coin the "friend" pockets $40. The exchange says, "How easy was that?" The two players both say, "That was fun, I only lost $20."

The Money-Go-Round

Traditional wagering regulation is predicated on two main features: (1) All betting is with licensed wagering operators and (2) the wagering operator will strive to make a profit in the offering of wagers to consumers. Combining these factors means the consumer is in effect input taxed. It is also relevant to note that taxation isn't purely a revenue raising device; it also is used as (in fact, some regulators would say the primary purpose of tax is) a means of blunting demand (consumption) and as a revenue offset against any negative affects on the community.

Betting exchanges, both by facilitating the betting between individuals and by turning the second proposition on its head (that the wagering operator doesn't make a margin in the odds offered), reach a utopian state where the money wheel just keeps turning. They get to clip the ticket every time the wheel turns--at a time when the punter is least concerned with a deduction from his winnings. As we will see below, the negligible retention rate has spawned a massive increase in wagering turnover.

But this method of business operation is diametrically opposed to the traditional wagering model, where the bookmaker seeks to maximize punter expenditure for given turnover.

The 'Happy' Punter - What You Don't Know Can't Hurt You

In Bet Unfair!, I put forward the proposition that the punter is getting too good a deal. On more detailed analysis, he mightn't be getting as good a deal as I thought, but then the punter, enthralled with betting in a "perfect market" of approximately 100 percent, mightn't be aware of the deal he is getting either.

In the U.K. context, where punters were used to having 9 percent deducted from their winnings (having had to beat the bookies' over-round margin first), paying 2 percent to 5 percent of winnings from a bet at significantly better odds is, well, a gift horse.

The Bookmaker Myth

It is a trap to even think like a bookmaker, or a regulator of bookmakers, when looking at betting exchanges. Betting exchanges make a big issue of the fact that they aren't a bookmaker--that there is no middle man--at least not one that arbitrages the market between persons of opposing views. But betting exchanges are definitely in the middle.

First, it is critical to understand that all betting exchange transactions, be they for (bets) or against (laying) a runner, are in fact all bets. Escrowed funds, therefore, represent that sum total of bets, or what bookmakers would call "turnover."

On this point (turnover) it is worthwhile pausing to understand the terminology used.

Exchanges use the terms "lay" and "available to lay" somewhat loosely; they display the bet that the layer will accept not the amount that a layer is prepared to risk (so for a bet at 9.00, or 8/1, a layer may have £50 available), but this means a liability of £400, which means the layer is simply a punter backing (in the case of racing) all other runners at 1/8, and his stake is £400 (to win £50).

If a punter (backer) matches this bet by betting £50, the exchange will report matched bets of £100, however, the amount escrowed in punters' accounts will be £450.

Bookmakers, however, do not record their liability to pay out a bet as a counter-bet.

Given both betting exchanges and bookmakers are wagering operators, there is a desire find common ground to compare them. The Betting Duty Reform legislation has perpetuated the myth by the requirement that "bet exchange[s]... must determine before the outcome of the bet is known which of the individuals is the bet-taker, and which is the bettor. At the end of each duty accounting period you must calculate your duty payment based on the net stake receipts of all bet-takers you act as bet-broker for."

That is, the betting exchange pretends it is a bookmaker, by law! We will return to the GPT issue shortly.

This has given rise to the practice of halving the trading volume (matched bets) to supposedly produce an apples-for-apples comparison with bookmakers. True, it calculates the backer stakes, but that's where the comparison ends.

Everything Is Not as It Appears

A bookmaker looking at betting builds a book by accepting various wagers on all runners such that (hopefully) in aggregate, no matter the outcome, he has more stakes received than his liability for each separate outcome; he uses the stakes received to pay out on that liability.

But a betting exchange, well, it has "other" money to pay out on those liabilities: the other punter's money. It is simply illogical (from a wagering operator's perspective) to ignore those funds staked by the punters betting against an outcome, even if it is bracketing all other runners in a single bet, in determining turnover.

The only logical assessment of real turnover is total escrowed funds. That definition then allows comparison with bookmakers, noting, however, that the betting exchange will have a zero gross wagering margin because of the perfect risk management system.

To highlight bracketing, Australian bookmakers are permitted to offer field against the favorite. Without jumping ahead of ourselves, we will see below in Table 2 where the layers of the short (weighted average) 1.56 Camlet are in fact getting odds 2.77 (>even money) about the field as backers.

So who is the layer?

Neither. All bets are made with the bookmaker and all bets constitute stakes received for tax and levies.

To use a simpler sports betting analogy, in a finals (ie. there must be a result) match between Leeds and Arsenal, if a bookmaker only offered punters the opportunity to back Arsenal or lay Arsenal--never mind that laying Arsenal is a bet on (backing) Leeds--how would the bookmaker account for turnover? On the basis if the backers of the Arsenal only? Or including the layers of the Arsenal because they are backing Leeds? Obviously it's both, but this is not how betting exchanges account for wagering on the distilled racing sub-event, "a runner against all others."

Advanced betting exchange systems operate on this principle; they create depth and liquidity by creating internal system bets. So when a punter says he wants to back the Arsenal £50 at evens, the system offers a system bet laying Leeds for £50 at evens, why do this? When a would-be Leeds backer looks in the betting exchange, he will find the £50 (on offer) and back it, triggering the first punter's bet (request) on Arsenal. In this case, who is the layer? While neither punter is technically the layer--having both backed their respective teams--the HMCE rules say the first user is deemed to be the bet taker for the purposes of the exchange, calculating its GWM and GPT liability.

Again, this defies logic. Both punters are backers making bets, and the wagering operator is, in effect, writing two bets where it is the layer for both, but for a nil gross wagering margin (although it is possible to build a margin into both system bets and bet requests if required).

It gets even better. In a betting exchange, each transaction is independent of all other transactions, and as each transaction is binary, there is one winner and one losing party. As we have seen with the coin-toss example, this makes for potentially enormous margins for the friend--the middle man--that matches the person of opposing views and holds money escrowed in trust.

This is mitigated somewhat, as exchanges have made the concession (in their own interest to create depth and liquidity) to aggregate the total wins and total losses on specific event to determine what they call "net winnings on a market."

Table 1: Betfair, Yarmouth Race 1, 17/9/02 - Losing Strong Favorite

Betfair Matched Trades Betfair Avg. Odds (decimal) Betfair backers' stakes Betfair layers' liabilities Total Escrowed Stakes or Real Turnover Max. "Net Winnings" (if winner) Maximum Gross Wagering Margin Betfair Commission @3.5% of Max. Net Winnings Margin on .5 Matched Trades (70,741) Margin on Total 'Turnover' (547,938)
Morning After 176,410  2.40 88,205  123,487  211,692  150,464  30.9% 5,266  4.6% 1.08%
Zingari 11,765  7.20 5,883  36,472  42,354  145,771  30.0% 5,102  4.4% 1.05%
Foolhardy 20,958  8.00 10,479  73,353  83,832  178,056  36.6% 6,232  5.4% 1.28%
Daisy Do 11,682  9.00 5,841  46,728  52,569  156,069  32.1% 5,462  4.7% 1.12%
Any Camp 5,900  10.00 2,950  26,550  29,500  138,782  28.5% 4,857  4.2% 1.00%
Non Ultra 1,292  21.00 646  12,920  13,566  127,456  26.2% 4,461  3.9% 0.92%
Burmese Princess 1,217  21.00 609  12,170  12,779  126,743  26.0% 4,436  3.9% 0.91%
My Girl Pearl 453  71.00 227  15,855  16,082  130,810  26.9% 4,578  4.0% 0.94%
Fine Frenzy 453  71.00 227  15,855  16,082  130,810  26.9% 4,578  4.0% 0.94%
Doubleuceone 233  70.00 117  8,039  8,155  123,104  25.3% 4,309  3.7% 0.89%
Total/Expected Value by P(n) 230,363  102.94% 115,182  371,428  486,610  149,572 30.7% 5,235  4.5% 1.08%

Table 1 attempts to clear some of the smoke.

First, the real turnover, £486,610, is a multiple (2.1x) of the matched-bets number, £230,363, published by Betfair and is 4.2 times the size of the bookmakers turnover assessment (half-matched-bets, e.£115,182).

Second, the maximum net winnings for all runners greater than the half-matched bets. I say "maximum," as the calculation assumes complete independence of all bets and that no punter was betting against himself by backing multiple runners or laying some.

Patently there is a lot of the type of activity by virtue of the Betfair commission scale where you have to do the best part of £500,000 turnover to get the best rate, although with this example, 76.7 percent of the backers' stakes was basically a field-against-favorite bet so (unless punters got cold feet and bet the reciprocal bet or arbitrated their position in that bet) the excess couldn't have diluted the net winnings by more than 20 percent.

Third, it highlights that irrespective of the winner (which was Zingari), the Betfair commission is of the order of £5,000 or 1 percent of real turnover with a gross wagering margin (net winnings as a percentage of real turnover) on the race of approximately 30 percent. The expected value of these key numbers, based on the probability (the odds) were:

  • Commission: £5,235;
  • % Margin: 1.08%

Using a "traditional" bookmaking approach (that is looking only at the backers' stakes), the gross wagering margin was 126.6 percent (for the winner) and varied between 106.9 percent and 154.6 percent. Betfair's retention rate (margin on half-matched-bets) was 4.4 percent (despite being commission being calculated at a weighted rate of only 3.5 percent) against the expected 4.5 percent.

Page 2

The views expressed in this editorial do not necessarily reflect the views of Interactive Gaming News or the views of the Australian Bookmakers' Association, which has not examined the U.K. issues in any detail.




Tim Ryan is a graduate of Agricultural Science, though he is probably best described as an innovator. He has worked in consumer online information and transaction systems since their inception and was the Australian pioneer of television data broadcast and is still the Consultant General Manager of the Seven Network’s teletext and datacast services – the principal provider of real time wagering information in Australia. Tim was an innovator of integrated wagering information and transaction systems. He has also consulted to a number of wagering system developers on the internationalisation of their systems. Tim became a NSW licensed bookmaker early in 2000 for regulatory reasons associated with the development of wagering systems. Despite not being a ‘working’ bookmaker Tim was appointed to the Australian Registered Bookmakers Advisory Council, Australia’s peak body representing all its 800 bookmakers. In this capacity he was instrumental in lobbying the independent Senators and the Government to ensure the exclusion of wagering from the Interactive Gambling Act.