BookieBashing have published Combination Bets on its Bet Tracker for some time.

An example of a Combination type of bet is “RequestABet – Bordeaux v Marseille – Marseille to score 2+ goals, 5+ Marseille corners & 20+ Bordeaux booking pts 11/2”.

On the 18th January 2020 we began recording the p/l of all of the +EV ComboBets we publish. These are: Requestabets (Sky), WhatOddsPaddy / OddsOnThat (Paddy Power+BFSB), YourOdds (William Hill), #PickYourPunt (Betfred), #GetAPrice (Ladbrokes) and #YourCall (Coral).

Results of each individual bet can be found here.

These bets have a similar structure; they combine various targets of goals, corners and cards in a match.

The purpose of recording the bets is not to prove the viability of a Strategy. The analysis of the viability of a strategy should only be done by an assessment of the logic and the mathematics that underpin the logic. The logic behind our strategy is “If we can regularly beat the Exchange lay price / last price matched and/or the mid-points of the Buy-Sell line at the Spread Betting sites then we will make a profit in the long term”.

The purpose of recording the bets was three-fold:

- To provide a database of the types of bets we are betting on, with typical back odds and fair odds.
- To understand typical winning and losing runs over a four week sample size.
- To provide a counter-argument to the narrative of three common misconceptions “Everyone is losing money on ComboBets”, “the bookmakers margins are so large it’s impossible to make money on them” and “All ComboBets are -EV”.

Our argument at BookieBashing is that it is possible to make money from these bets due to differing opinions on the xG, xCorners and xCards in a game, especially as lines change over the day.

### Strategy: Betting to win every +EV bet to win £200

We published 322 bets at an average of 81 bets per week. Profit was returned on each of the four weeks when betting each bet to a consistent liability of £200 per bet.

### p/l by week

### p/l by day

### Estimating overall EV: the difficulty of multiple inter-related bets on a single outcome

It is complex to calculate an Expected Value figure for the overall p/l based on the published bets due to the inter-dependency of multiple bets being placed on the same outcome of a single game. When a line moves, it tends to make several related bets +EV at the same time. For example, if smart money comes in for corners then multiple bets relating to corners may all become instantly +EV. These will all get published on the tracker, however the overall EV for the set of bets may change based on a number of factors. The sum of the EV of the individual bets may under-estimate the overall expectation.

### Variance

Assuming each bet is independent (despite the fact we know many are related), we can plot the variance of the average EV (109%) over 348 bets at average odds 9.32 using the variance calculator on the Spreadsheets page.

The graphs shows that the final p/l of 23% finished higher than mean expectation. In fact the probability of finishing negative (< 0% ROI) is approximately similar to finishing >23% ROI. The next 322 bets could provide a negative return and yet the viability of the strategy and the logic would not change.

### Benchmarking of the Spread and Exchange data

On any single market, an argument can be made whether to use the lay price, the last price matched or the back price for a reasonable estimate of the probability of an event using the exchange. Equally an argument can be made whether to use the Buy line, the Sell line or the mid-point line – or indeed which spread site to use when the lines differ. Whilst the accuracy of an estimation can be enhanced on any individual bet, we provide a consistent approach for two reasons:

- The process needs to be efficient.
- Any optimism in one line will be balanced by pessimism in another line over a large number of bets over a long time-scale and the overall net affect of pessimism/optimism will converge to zero.

The approach we use in our strategy to benchmark Goals, Corners and Cards is:

- Exchange price: Use Last Price Matched if that price is still between the available Back and Lay otherwise use the Lay Price.
- Spread price: We use the midpoint of the Buy/Sell line and SportingIndex.