2b.3 Calibration for in-play OddsCompiling

OddsCompiler Platform is a unique tool for pre-game and in-play oddcompiling, but you need a careful study to find the best sample assumptions in order to calculate probabilities and odds.

Pre-game odds compiling is much easier to deal with and it is better to focus on data sample from last season.

When you deal with in-play oddcompiling it is necessary to deal with data from more than one previous seasons.

In the examples below we make an in-play depth analysis for Arsenal vs Chelsea assuming:

  1. The score is 0 – 1

  2. Elapsed time 23min

 

We will investigate the differences in the probability results by increasing the data sample:

In all options we assume games from Arsenal as Home and games from Chelsea as Away.


 

Option 1:

Data Sample seasons 2014-2015

Sample 2: 5 games

Probability for Home to win : 20.00%

Probability for Away to win : 80.00%

Probability for the Draw : 0.00%

Probability for the Under 2.5 : 40.00%


 

 

Option 2:

Data Sample seasons 2014-2015, 2013-2014

Sample 2: 10 games

Probability for Home to win : 20.00%

Probability for Away to win : 60.00%

Probability for the Draw : 20.00%

Probability for the Under 2.5 : 30.00%


 

 

Option 3:

Data Sample seasons 2014-2015, 2013-2014, 2012-2013, 2011-2012

Sample 2: 20 games

Probability for Home to win : 25.00%

Probability for Away to win : 60.00%

Probability for the Draw : 15.00%

Probability for the Under 2.5 : 20.00%

 

Option 4:

Data Sample seasons 2014-2015, 2013-2014, 2012-2013, 2011-2012, 2010-2011

Sample 2: 22 games

Probability for Home to win : 22.73%

Probability for Away to win : 63.64%

Probability for the Draw : 13.64%

Probability for the Under 2.5 : 18.18%

 


When you deal with in-play oddscompiling you need data.

Looking at the results option 1 which assumes only one season of previous data, may be appropriate for pre-game oddscompiling, but not for in-play oddscompiling. In in-play oddscompiling you need to increase data samples, with more seasons to make calculated probabilities more reliable.