WARNING: If you don't like stats, stats models, or examination of future success based on historical precedence, this thread is not for you.
I think it is clear now that we will finish bottom 4. There is still doubt as to what our final position will be, with every position from 18 to 15 a live possibility (unlike we finish any higher).
The question I have is, what does this mean in terms of how long we must wait until our next premiership? Hawthorn is a very successful club and we bat well above the average. I'm not going to consider our history, instead, I will review the history of outcomes to set our expectations around average outcomes and then we can tack on a 'Hawthorn is ******* awesome' factor on at the end
I'll be making some comparisons between the daft era and pre-draft era. I have excluded data prior to 1950 as too many changes in league size, and even the particular teams involved. I will be focussing in on the probability of winning a premiership within ten years following a season in which a team finished a particular place on the ladder.
I have used a logistic regression model using final ladder position, a dummy for draft era (taking the value of 1 for draft era and 0 for pre-draft era) and a count of teams which competed in the league for the year related to the particular record. I also include an interaction term between the draft dummy and the final ladder position to allow the model to be more flexible. I trialled several other variables (mostly lags of final ladder position) and it didn't improve the predictive power of the model so I applied the principal of parsimony and ditched them. The model predicts whether or not a team will win a premiership within the next ten years around 68% of the time based on those variables. That is good enough for this purpose.
Lastly, due to the complex and dynamic structural changes to the league over time, I have chosen to report on average probability. This is calculated by first estimating the probability of of an individual team winning a premiership within ten years and then taking the mean of all teams that finished in that position the same position on the ladder and grouping them by draft and pre-draft eras. I also do the same thing except using a reverse ladder so I can examine the effect of finishing last regardless of league size.
Now to the fun part
Top 4 v Bottom 4
According to the model, pre-draft era top 4 teams had around 58% chance of winning a premiership in the following ten years. This is higher than draft era top 4 teams, which had around a 44% of winning a premiership in the following ten years.
Pre-draft bottom 4 teams had substantially less probability of winning a premiership in the following ten years (around 25% chance) as top 4 teams. This contrasts with draft era bottom 4 teams, which had around a 20% chance of winning a premiership in the following ten years.
On face value, neither stat is an endorsement of the draft system! But the reality is it has more to do with the increase in the number of teams that coincided with the draft.
By Ladder Position
As per the chart below, the probability of winning a premiership in the next ten years decreases approximately linearly with an increase in ladder position. The slope is steeper for pre-draft era, which leads to it crossing over in probability at around 9th. No team has won after finishing 17th of lower but the model still estimates a positive probability.
Notably, this model is good at predicting actual outcomes but when I attempted to use it to simulate an 18 team competition for pre and post draft, it struggled, overstating the total probability of winning a premiership within ten years.
I re-trained model using the entire dataset back to 1897 so I could get a wider range of final ladder positions. I then created the chart below assuming the league had 18 teams before and after the implementation of the draft. The draft line is probably still overstating the probability of winning a premiership within ten years (technically, the area under each curve should be the same). If I was able to appropriately adjust the draft era probability curve lower to equal the area under the pre-draft curve, my guess is it crosses the line around position 4-5. So essentially, everyone bar the top 4 was made better off under the draft system, in terms of winning a premiership in the ten years following finishing top 4.
Chances of Winning A Flag in the draft era after finishing bottom four?
As you can see, there is little difference in the probability of winning a flag in any given year following bottom four (based on historical frequency, not modelling) but the best chance is sooner rather than later for finishing 3rd last of 4th last while it is in year 9 and 10 after finishing 2nd last that has historically seen the best chance of winning a flag. Finishing last never rises above a 5% chance for a given year.
What does it all mean?
I think this analysis demonstrates that the draft really does give teams a greater shot at winning a flag in the following ten years than the old zone system. Only the top 5 had a 20% chance or better (under the assumption of 18 team competition) of winning a premiership in the next 10 years. The second model suggests that the top 9 have a 20% chance or better. While this probability is likely too high, the flatter curve shows that the probability is more evenly distributed across the ladder. Should we finish 17th, we would have around a 10% chance of winning a premiership within 10 years, though we would be the first ten to finish 17th to do so.
The analysis also shows that there is no shortcut to success, on average. Your probability doesn't increase by being lower down the ladder. It literally decreases! This is not to make any judgement on the draft or talent but is probably reflective of the fact that teams that lower just have less to build on and so only a small chance of winning a premiership in the following ten years. Finishing last seems to have a particular negative effect on your chances of winning a flag in the following ten years. This is likely the case because if you Arne't good enough to come second last your list is irredeemable with little to build on...In short, if we can't hold off North (or finish above another club) then we are likely in for a world of pain unless our hawthorn awesome factor can get us up the ladder! Go hawks!
PS I plan to re-run this model an include variables for clubs that received priority draft picks. But I need to put that dataset together as I unfortunately lost the one I had...
I think it is clear now that we will finish bottom 4. There is still doubt as to what our final position will be, with every position from 18 to 15 a live possibility (unlike we finish any higher).
The question I have is, what does this mean in terms of how long we must wait until our next premiership? Hawthorn is a very successful club and we bat well above the average. I'm not going to consider our history, instead, I will review the history of outcomes to set our expectations around average outcomes and then we can tack on a 'Hawthorn is ******* awesome' factor on at the end
I'll be making some comparisons between the daft era and pre-draft era. I have excluded data prior to 1950 as too many changes in league size, and even the particular teams involved. I will be focussing in on the probability of winning a premiership within ten years following a season in which a team finished a particular place on the ladder.
I have used a logistic regression model using final ladder position, a dummy for draft era (taking the value of 1 for draft era and 0 for pre-draft era) and a count of teams which competed in the league for the year related to the particular record. I also include an interaction term between the draft dummy and the final ladder position to allow the model to be more flexible. I trialled several other variables (mostly lags of final ladder position) and it didn't improve the predictive power of the model so I applied the principal of parsimony and ditched them. The model predicts whether or not a team will win a premiership within the next ten years around 68% of the time based on those variables. That is good enough for this purpose.
Lastly, due to the complex and dynamic structural changes to the league over time, I have chosen to report on average probability. This is calculated by first estimating the probability of of an individual team winning a premiership within ten years and then taking the mean of all teams that finished in that position the same position on the ladder and grouping them by draft and pre-draft eras. I also do the same thing except using a reverse ladder so I can examine the effect of finishing last regardless of league size.
Now to the fun part
Top 4 v Bottom 4
According to the model, pre-draft era top 4 teams had around 58% chance of winning a premiership in the following ten years. This is higher than draft era top 4 teams, which had around a 44% of winning a premiership in the following ten years.
Pre-draft bottom 4 teams had substantially less probability of winning a premiership in the following ten years (around 25% chance) as top 4 teams. This contrasts with draft era bottom 4 teams, which had around a 20% chance of winning a premiership in the following ten years.
On face value, neither stat is an endorsement of the draft system! But the reality is it has more to do with the increase in the number of teams that coincided with the draft.
By Ladder Position
As per the chart below, the probability of winning a premiership in the next ten years decreases approximately linearly with an increase in ladder position. The slope is steeper for pre-draft era, which leads to it crossing over in probability at around 9th. No team has won after finishing 17th of lower but the model still estimates a positive probability.
Notably, this model is good at predicting actual outcomes but when I attempted to use it to simulate an 18 team competition for pre and post draft, it struggled, overstating the total probability of winning a premiership within ten years.
I re-trained model using the entire dataset back to 1897 so I could get a wider range of final ladder positions. I then created the chart below assuming the league had 18 teams before and after the implementation of the draft. The draft line is probably still overstating the probability of winning a premiership within ten years (technically, the area under each curve should be the same). If I was able to appropriately adjust the draft era probability curve lower to equal the area under the pre-draft curve, my guess is it crosses the line around position 4-5. So essentially, everyone bar the top 4 was made better off under the draft system, in terms of winning a premiership in the ten years following finishing top 4.
Chances of Winning A Flag in the draft era after finishing bottom four?
As you can see, there is little difference in the probability of winning a flag in any given year following bottom four (based on historical frequency, not modelling) but the best chance is sooner rather than later for finishing 3rd last of 4th last while it is in year 9 and 10 after finishing 2nd last that has historically seen the best chance of winning a flag. Finishing last never rises above a 5% chance for a given year.
What does it all mean?
I think this analysis demonstrates that the draft really does give teams a greater shot at winning a flag in the following ten years than the old zone system. Only the top 5 had a 20% chance or better (under the assumption of 18 team competition) of winning a premiership in the next 10 years. The second model suggests that the top 9 have a 20% chance or better. While this probability is likely too high, the flatter curve shows that the probability is more evenly distributed across the ladder. Should we finish 17th, we would have around a 10% chance of winning a premiership within 10 years, though we would be the first ten to finish 17th to do so.
The analysis also shows that there is no shortcut to success, on average. Your probability doesn't increase by being lower down the ladder. It literally decreases! This is not to make any judgement on the draft or talent but is probably reflective of the fact that teams that lower just have less to build on and so only a small chance of winning a premiership in the following ten years. Finishing last seems to have a particular negative effect on your chances of winning a flag in the following ten years. This is likely the case because if you Arne't good enough to come second last your list is irredeemable with little to build on...In short, if we can't hold off North (or finish above another club) then we are likely in for a world of pain unless our hawthorn awesome factor can get us up the ladder! Go hawks!
PS I plan to re-run this model an include variables for clubs that received priority draft picks. But I need to put that dataset together as I unfortunately lost the one I had...
Last edited: