statsman74
Club Legend
THE BACKGROUND
As part of the BigFooty data loss, fortunately for some and unfortunately for the others involved, we lost the two threads in relation to the suggestion by poster Starz that the % method of separating teams on equal points be replaced by +/- of points for and points against, because, in Starz words, it was “confusing” and “annoying”.
As expected a tide of opinion solidly opposing Starz’ suggestion was made, with many of the pro-% crew building, sustaining and supporting a credible argument for the retention of % and dismissing +/-, whereas the core of the pro +/- crew survived on “you’re wrong, we’re right” and were left with its only advantage being that it is simpler.
In these posts I had mathematically proven that there is a correlation between total points scored in a game and margin, and therefore, a +/- system would be biased towards good teams playing at high scoring venues and biased against poor teams playing at the same venues.
The percentage system shows no such correlation.
What this means is that the +/- approach is a flawed system as it has inbuilt uncontrollable biases, whereas % does not.
Late last week the opinion poll was running 38-13 in favour of %.
WHERE WERE WE?
Starz found an unlikely ally in Frodo, who provided an adult approach to a juvenile one. After reading my posts Frodo suggested that he could tell what I was saying, and wasn’t 100% sure it was correct.
THE CHALLENGE
Frodo’s challenge was that I needed to demonstrate one real example of where a change from percentage to +/- would result in ladder position changing and that the change would be wrong because of the biases mentioned previously.
Starz’ contribution to this was suggesting it was time for me to “put up or shut up”. (Nothwithstanding that he had done absolutely nothing to attach any credibility to his argument or debunk any arguments I had made.)
THE CHALLENGE – TAKEN
Once I started digging and running the original correlations, I was always going to look further - mainly because I am naturally curious, but also because I wanted to reinforce my original calculations with data from different years.
My fear was that I wouldn’t find enough “real” situations to look into – after all, I would need to find two or more teams on level points, and their percentage to be very close. From there I’d need the lower ranked team to be a substantially higher scoring team and the venues they played at to be significantly different to create the bias.
Nevertheless, I trusted my calculations and knew “the truth would set me free”.
THE EXAMPLE
The year was 1998. Amongst a group of mid-table teams that ended the regular season on losing streaks were the 7th placed West Coast Eagles and the 8th placed Essendon Bombers.
(In the Finals Carey, Stevens & Bell drove the 1st placed Kangaroos to a victory over the Bombers while Chris Grant and Simon Minton-Connell kicked 9 between them in a romp by the 2nd placed Bulldogs over the Eagles.)
Both teams won 12 games but the Eagles had a percentage of 109.42 and the Bombers was just 108.64. However the Bombers were a higher scoring team, and in fact they had 2250 points for and 2071 against, while the Eagles had 1940 points for and 1773 against.
THE CHANGE
If the % method was ditched in favour of +/-, Essendon would move up to 7th (+179) and West Coast would fall to 8th (+167).
However;
Essendon played 16 games at the high scoring MCG, 3 games at the lower scoring Waverley, and their final 3 games at Princes Park (high), AAMI (low) and Subiaco (low).
West Coast played 9 games at the lower scoring Subiaco, 3 at the WACA (high), 3 at the MCG (high), 2 at AAMI (low) and other games at Skilled (low), Gabba (low), SCG (high), Waverley (low) and Victoria Park (insufficient data to be certain).
After adjusting for these venues back to a standard baseline, the adjusted for and against for Essendon is 2209/2026 (109.03% or +183) and for West Coast is 1989/1796 (110.75% or +193)
WHAT DOES THIS MEAN?
Using +/-, Essendon move up to 7th, with WCE back to 8th.
But after removing the bias of venue, this is unjustified, as WCE should have been in 7th (+193 to +183).
Notably, after removing the bias of venue, nothing changes under the % method. WCE remains in 7th in both cases. (as discussed and analysed previously, % is unbiased in this regard)
CONCLUSION
This is one example, and in the end I didn’t need to look too far to find it. In this case, perhaps 7th/8th or 8th/7th makes no difference. But what about if it happens between important spots like 1st/2nd, 2nd/3rd, 4th/5th, 8th/9th or 15th/16th? Or maybe it still does matter back in 1998, as some teams match up better with others.
Yes, +/- is simpler.
But it isn’t fair.
FINAL WORDS
The argument was made by Starz. The argument was debunked by me and many others. The challenge was made by Frodo and Starz. The challenge was accepted - I’ve made the proof, and therefore “put up”.
Its now time for Starz (and the +/- system) to shut up.
As part of the BigFooty data loss, fortunately for some and unfortunately for the others involved, we lost the two threads in relation to the suggestion by poster Starz that the % method of separating teams on equal points be replaced by +/- of points for and points against, because, in Starz words, it was “confusing” and “annoying”.
As expected a tide of opinion solidly opposing Starz’ suggestion was made, with many of the pro-% crew building, sustaining and supporting a credible argument for the retention of % and dismissing +/-, whereas the core of the pro +/- crew survived on “you’re wrong, we’re right” and were left with its only advantage being that it is simpler.
In these posts I had mathematically proven that there is a correlation between total points scored in a game and margin, and therefore, a +/- system would be biased towards good teams playing at high scoring venues and biased against poor teams playing at the same venues.
The percentage system shows no such correlation.
What this means is that the +/- approach is a flawed system as it has inbuilt uncontrollable biases, whereas % does not.
Late last week the opinion poll was running 38-13 in favour of %.
WHERE WERE WE?
Starz found an unlikely ally in Frodo, who provided an adult approach to a juvenile one. After reading my posts Frodo suggested that he could tell what I was saying, and wasn’t 100% sure it was correct.
THE CHALLENGE
Frodo’s challenge was that I needed to demonstrate one real example of where a change from percentage to +/- would result in ladder position changing and that the change would be wrong because of the biases mentioned previously.
Starz’ contribution to this was suggesting it was time for me to “put up or shut up”. (Nothwithstanding that he had done absolutely nothing to attach any credibility to his argument or debunk any arguments I had made.)
THE CHALLENGE – TAKEN
Once I started digging and running the original correlations, I was always going to look further - mainly because I am naturally curious, but also because I wanted to reinforce my original calculations with data from different years.
My fear was that I wouldn’t find enough “real” situations to look into – after all, I would need to find two or more teams on level points, and their percentage to be very close. From there I’d need the lower ranked team to be a substantially higher scoring team and the venues they played at to be significantly different to create the bias.
Nevertheless, I trusted my calculations and knew “the truth would set me free”.
THE EXAMPLE
The year was 1998. Amongst a group of mid-table teams that ended the regular season on losing streaks were the 7th placed West Coast Eagles and the 8th placed Essendon Bombers.
(In the Finals Carey, Stevens & Bell drove the 1st placed Kangaroos to a victory over the Bombers while Chris Grant and Simon Minton-Connell kicked 9 between them in a romp by the 2nd placed Bulldogs over the Eagles.)
Both teams won 12 games but the Eagles had a percentage of 109.42 and the Bombers was just 108.64. However the Bombers were a higher scoring team, and in fact they had 2250 points for and 2071 against, while the Eagles had 1940 points for and 1773 against.
THE CHANGE
If the % method was ditched in favour of +/-, Essendon would move up to 7th (+179) and West Coast would fall to 8th (+167).
However;
Essendon played 16 games at the high scoring MCG, 3 games at the lower scoring Waverley, and their final 3 games at Princes Park (high), AAMI (low) and Subiaco (low).
West Coast played 9 games at the lower scoring Subiaco, 3 at the WACA (high), 3 at the MCG (high), 2 at AAMI (low) and other games at Skilled (low), Gabba (low), SCG (high), Waverley (low) and Victoria Park (insufficient data to be certain).
After adjusting for these venues back to a standard baseline, the adjusted for and against for Essendon is 2209/2026 (109.03% or +183) and for West Coast is 1989/1796 (110.75% or +193)
WHAT DOES THIS MEAN?
Using +/-, Essendon move up to 7th, with WCE back to 8th.
But after removing the bias of venue, this is unjustified, as WCE should have been in 7th (+193 to +183).
Notably, after removing the bias of venue, nothing changes under the % method. WCE remains in 7th in both cases. (as discussed and analysed previously, % is unbiased in this regard)
CONCLUSION
This is one example, and in the end I didn’t need to look too far to find it. In this case, perhaps 7th/8th or 8th/7th makes no difference. But what about if it happens between important spots like 1st/2nd, 2nd/3rd, 4th/5th, 8th/9th or 15th/16th? Or maybe it still does matter back in 1998, as some teams match up better with others.
Yes, +/- is simpler.
But it isn’t fair.
FINAL WORDS
The argument was made by Starz. The argument was debunked by me and many others. The challenge was made by Frodo and Starz. The challenge was accepted - I’ve made the proof, and therefore “put up”.
Its now time for Starz (and the +/- system) to shut up.
