Risky business: The perils of setting Premier League odds

If, as a famous saying goes, the thoughts of a young man turn to love in springtime, then at this point in the summer – at the start of the football season – it’s all about what bookmakers are going to lay.

For trading teams, there are a multitude of factors to take into account – the significant one being transfers.

New signings can generate a lot of excitement among fans in the same way that losing a key player can be a big blow. Tottenham have drifted by around 7% to finish in the EPL top four next season, largely on the back of Harry Kane heading to Bayern Munich.

While the transfer spending firepower of both the Saudi Arabia and MLS leagues and the subsequent means of offloading expensive unwanted players is good news for some Premier League clubs, it also adds a lot more volatility into which players are likely to be leaving before the transfer window closes. 

Making moves

At the time of writing, Mo Salah may be on the move to Al-Ittihad; Liverpool could look to replace him with a loan of Mbappe and PSG might have attempted to cover this by pinching Kane before Bayern came calling.

It’s an unlikely scenario, but it is an example of the type of transfer that can now happen and would cause multiple price moves in both short- and long-term markets – all moves that trading teams need to take into account.

Aside from transfers, early season form also causes a lot of uncertainty. Wrexham drifted by around 10% to win League Two on the back of a single opening-day defeat and it isn’t unusual for the “stalking horses” of the fallow summer period to become dead donkeys once the season is in full swing.

However, the last game a team plays contributes only a tiny part of their overall “rating” in models that use exponential smoothing – possibly less than 1%. But the impact a defeat can have on the price for that team in the market usually far exceeds this. 

Leveraging the true price

The art of bookmaking isn’t in laying the “true” price. It is in understanding what the true price is and then laying a price to your customers that is as far away from that as to be advantageous to the operator.

If your customers believe the probability of something occurring is 20% when you consider it to be only 10%, then you consequently have that “margin” in your favour. 

The difficulty for operators is in knowing what the “true price” actually is. How do you price up teams when there are a near infinite number of variables? For example, if last year’s home form was exceptional what is that worth in a new season (hint – not very much)?

How do you rate the strength of a manager? Does it matter how good your players are come matchday if you get a bad referee? Is one defeat a blip or a signal?

Reading the signals

In today’s industry, signals are hard to come by. Operators have outsourced large parts of their trading operations to supply chains that have no experience of constructing prices without a market to copy from.

The result is that most prices are blindly following market moves, in a belief that “staying in line” means staying away from bad business. It doesn’t matter if you don’t know what the price to lay should be on an outcome, if your business model is built around restricting access to customers that show they know more about betting than your supply chain does. 

The only long-term sensible answer for operators is to find a way to buck this trend.

Price optimisation

There are two key changes that operators can make to optimise their pricing. The first approach is to utilise “ratings” models predicated on deeper and wider arrays of data.

As a result, when a transfer happens in the market or games take place, robust models that are built to take more variables into account can immediately be updated and give an understand of what the true reflection of the outcome is.

Operators who switch to using models built in this way, or to suppliers that have them, can therefore optimally position themselves around the prevailing market noise and maximise trading margin.

Analysing the data

The second method is for operators to use the data implicit in their customers’ bets.

The growth of AI means that there is machine-learning technology that can distil and react to any “information” contained in the bets that customers place. Even very small numbers of bets can enable users to optimise prices.

Combining this with the sort of sensitive ratings models outlined above, plus the oversight of expert traders, will enable operators to begin to generate industry-leading, or “alpha”, returns.

So, back to Harry Kane. Is he really worth 7% over a season? Bookmakers out there are still trying to price that in for this week. At 10star we’re confident we know. And we’re just glad none of us are Spurs fans.