The calculation starts by taking the line-up for a given match, then looking back over the 30 preceding matches and counting how often a given player appeared in a given spot in the starting line-up over that period.
For example, over a given period Richie McCaw might have started 25 of 30 matches in the 7 shirt and Dan Carter 20 in the 10 shirt. If they started in those shirts in the next match, they’d contribute 25 and 20 points respectively to that match’s raw similarity score.
Repeat this for each player/position and you get a total out of 450 (15 positions x 30 possible starts over the comparison period); I divided this raw score by 450 to get a score expressed as a % of the maximum.
I then took a rolling average (over 30 games) to iron out any match-to-match variability and highlight the underlying trend.
So I only have the lineups back as far as 2020 for Ireland, I might go back further in future but I didn't have the data readily available, this does mean that the further back I go on the 30 game rolling average the less accurate it is
So these are the rolling average scores from the start of 2023 onwards
0.5147483695
0.5040817028
0.4891928139
0.4868965176
0.4816372584
0.473266888
0.4763779991
0.47327747
0.4696478404
0.4698453712
0.4753268527
0.4808016002
0.4829497483
0.4832973267
0.4878793373
0.4983978558
0.5035552632
0.5064397948
0.5100447331
0.5116626668
0.5161441483
I had collected the team sheets manually previously as there was no good consistent source for all lineups for Ireland
Ah very interesting! Not far off peak All Black levels then - and presumably a bit more cohesion developed at lower levels as so many playing for Leinster together. Thanks for sharing! Will let you know if I ever track down full data for other countries.
But then again, Erasmus and the Springboks seem to be getting rewarded for a policy of regular selection gambles and changes. I'd be curious to compare the recent Springbok consistency score to the NZ one.
I’d like to do that analysis too! Unfortunately other countries don’t make line-up data as readily available as New Zealand do (at least to my knowledge — do pass on if you know where to find it for SA).
Can you share how you calculated the similarity score please?
I have all of the Ireland lineups from nearly the last 20 years and I'd love to see how Ireland compares
Sure! I’d be really interested in that too.
The calculation starts by taking the line-up for a given match, then looking back over the 30 preceding matches and counting how often a given player appeared in a given spot in the starting line-up over that period.
For example, over a given period Richie McCaw might have started 25 of 30 matches in the 7 shirt and Dan Carter 20 in the 10 shirt. If they started in those shirts in the next match, they’d contribute 25 and 20 points respectively to that match’s raw similarity score.
Repeat this for each player/position and you get a total out of 450 (15 positions x 30 possible starts over the comparison period); I divided this raw score by 450 to get a score expressed as a % of the maximum.
I then took a rolling average (over 30 games) to iron out any match-to-match variability and highlight the underlying trend.
Hope that makes sense!
My R code is also here if that’s helpful: https://github.com/plottheball/ptb-newsletter/blob/main/newsletter/2024/69-abs/ptb-69-final.R
So I only have the lineups back as far as 2020 for Ireland, I might go back further in future but I didn't have the data readily available, this does mean that the further back I go on the 30 game rolling average the less accurate it is
So these are the rolling average scores from the start of 2023 onwards
0.5147483695
0.5040817028
0.4891928139
0.4868965176
0.4816372584
0.473266888
0.4763779991
0.47327747
0.4696478404
0.4698453712
0.4753268527
0.4808016002
0.4829497483
0.4832973267
0.4878793373
0.4983978558
0.5035552632
0.5064397948
0.5100447331
0.5116626668
0.5161441483
I had collected the team sheets manually previously as there was no good consistent source for all lineups for Ireland
Ah very interesting! Not far off peak All Black levels then - and presumably a bit more cohesion developed at lower levels as so many playing for Leinster together. Thanks for sharing! Will let you know if I ever track down full data for other countries.
Brilliant. Thank you very much
But then again, Erasmus and the Springboks seem to be getting rewarded for a policy of regular selection gambles and changes. I'd be curious to compare the recent Springbok consistency score to the NZ one.
I’d like to do that analysis too! Unfortunately other countries don’t make line-up data as readily available as New Zealand do (at least to my knowledge — do pass on if you know where to find it for SA).
Did write about the Springboks earlier in the year too actually: https://www.plottheball.com/p/the-springboks-have-won-consecutive