Thread by @AntoineJWMartin

I created a betting system for football betting using expected goals (xG) data.

And I am also sharing it for free.

Here is a simple 5-step approach to predicting the upcoming Premier League matches.

[🧵 thread]

Step 1.

👉Create a Match Rating

You can create a match rating by subtracting the away team’s xG difference (xG - xG allowed) from the home team’s xGD over the course of their 3 most recent games.

Let's illustrate the idea using the 2021/22 Brentford-Leicester City game 👇

Step 1. [Example]

- Brentford entered the game with 4.7 xG scored and 4.1 xG allowed, or an xGD of 0.6 (4.7 - 4.1)
- Leicester entered the game with 7.4 xG scored and 3.4xG allowed, or an xGD of 4 (7.4 - 3.4)

👉 The #BRELEI match rating was therefore -3.4 (0.6 - 4).

Step 2.

👉Calculate the historical full-time results

Using 4 seasons of historical data, you can calculate how many games with a match rating of -3.4 finished with a home win, a draw or with an away win.

Let's continue with the #BRELEI illustration 👇

Step 2. [Example]

Since 2017/18, there have been a total of 110 Premier League games with a pre-game match rating of -3 (we rounded the match rating to the nearest whole number).

- The home teams won in 35 games
- A draw occured in 28 games
- The away teams won in 47 games.

Step 3.

👉Calculate each full-time result’s probability of occurring

You can calculate the share of each FTR by dividing the respective number of outcomes from Step 2 by the total number of games available for that match rating.

Let's calculate it for the #BRELEI game 👇

Step 3. [Example]

Using the data from Step 2, you can easily calculate that:

- the home wins represent 31.8% (35/110),
- the draws represent 25.5% (28/110) and
- the away wins 42.7% (47/110) of all available matches.

👉These will be your theoretical probabilities.

Step 4.

👉Calculate the fair odds

You can calculate the fair odds for each full-time result by simply inverting the relative shares calculated in Step 3.

Let's illustrate the idea using our favourite #BRELEI match 👇

Step 4. [Example]

By inverting the probabilities from Step 3, you can easily calculate the following fair odds:

- 3.14 for a home win (1/0.318)
- 3.93 for a draw (1/0.255)
- 2.34 for an away win (1/0.427)

👉These will be your points of reference when placing your bets.

Step 5.

👉Identifying value

You can compare the fair odds from Step 4 with the odds offered by the bookmaker and identify the outcome offering the highest value.

Let's continue with the #BRELEI illustration 👇

Step 5. [Example]

By comparing Pinnacle's prices with our fair set of odds, you can easily conclude that:

- PInnacle's away win odds offer a positive expected value of 9% ((2.55-2.34)/2.34).

👉You should look to place the away win bet.

Ok, but how reliable is this model? 👇

If you were to bet £100 on all positive-value predictions, then over the previous #PremierLeague season:

👉 You would have won £2927 based on 301 bets
👉 You would have correctly predicted 28.9% of your bets for a ROR of 9.7%.

Data: @12Xpert @StatsBomb via @fbref

@12Xpert @StatsBomb @fbref That's a wrap!

If you enjoyed this thread:

1. Follow me @AntoineJWMartin for more of these
2. RT the tweet below to share this thread with your audience

https://twitter.com/AntoineJWMartin/status/1554385763822981121