Certified Legendary Thread Squiggle 2017

Remove this Banner Ad

Final Siren said:
That's a valid assumption if the models are better predictors than the bookies.

The real question is why there's a marked difference between the bookies and the models! There are quite a few R1 games like this. On the site, you can see that Punters (which are really bookies) are a lot more optimistic about Essendon's chances against the Hawks, for example, and more pessimistic about North's vs West Coast.

What's going on is information asymmetry: Punters know something the models don't. None of the models (I think) are sensitive to off-season personnel changes, so they don't know that Essendon is regaining a bunch of players as well as, perhaps, some self-belief and passion, nor do they know that North cut a bunch of veterans while West Coast acquired a player who's so good he won a Brownlow in the off-season.

So the Punters have an advantage over the models here, and I reckon they're probably more right. It's hard to beat the bookies at the best of times, and this isn't the best of times: it's been six months since the models last tasted fresh data, while human beings have had that time to think about just nine games. That's an advantage-humans situation. In a few rounds time, it should balance out, as we start to struggle to make intuitive sense of the mix of results, while the computers can go about processing it and forming objective conclusions.
Click to expand...

I wonder if you can increase your data points to include players. That way you can take into account personnel changes in the offseason. I think you could even factor in expected output from a draft position plus expect improvement in the first 5 year and expected loss of output after 28. Perhaps an injury or time out of the game could lead to a drop in expected output.

Furthermore, are you putting too much weight on past performances from a team perspective and that is why there is a mild downward trend in your prediction analysis (season average). If so, perhaps you need to weight past performances with a time code rather than "last game played" code to give more weight to early season games. This is obviously something you care about as you have an article about early season games being a great predictor. However, I suppose taking a lot of historical evidence into account says a lot about getting through the whole season, which is a big part of finals.

All that said, can I congratulate you on all of the visualisations you use. My favourite one is the results of the survey for season finishing position. Smoothing out the distribution of predictions is a fantastic idea. This is exactly the sort of data visualisation we need to support our democracy in such a data driven world. Perhaps, if you were able to gather data each week on that it would make for a wonderful evolution visualisation.

I also think you could combine the ability of crowd intelligence with actual game results to somehow feed into your data model.

I hope your modelling is picked up by a news organisation so the can fund you to do this full time, as well as giving you computer resources to crunch more data and people to help gather more data for you. Keep it up bro!
 

Log in to remove this ad.

fishardansin said:
I wonder if you can increase your data points to include players. That way you can take into account personnel changes in the offseason. I think you could even factor in expected output from a draft position plus expect improvement in the first 5 year and expected loss of output after 28. Perhaps an injury or time out of the game could lead to a drop in expected output.
Click to expand...
You would need a lot better data to map off-season changes, and much of that data is not publicly available. How do you weight expected output of a number 1 draft pick versus that of fourth round pick? How do you plot that over the season? For example, Andrew McGrath had a good debut, and will get better as he gets more experience, but will probably tail off in the final rounds of the season (and will probably at some point be rested).

Some of these values are found in things like Supercoach scores and Champion Data rankings, but they are very carefully guarded. AFL Player rankings might be able to do it, but you'd still have to crunch the historical data of all high and low draft picks.

I'm sure there are people at footy clubs who do just this, with a view to how they can plan their seasons and multi-year campaigns and rebuilds.
 
i got caught up in the cubs last year and have become quite hooked on MLB as a result.

interesting article here about season simulations, which marries up quite well with similar things the squiggle does (eg, the 'tower of power').

i think baseball is probably the most statistically rich sport, so possibly easier to do with MLB than the AFL, but anyway, the article could be of interest to some on here even who arent into baseball:

http://www.espn.com.au/mlb/story/_/id/19037525/the-65-win-cubs-tales-mlb-2017-alternative-lines

i particularly liked this:

"They don't play the game to figure out who is best. They play to find out who won. It's a subtle, but crucial difference. They play the game because it's not fun to win the game on paper, and it's not fun to watch a team win the game on paper. Simulations can tell you who won it on paper, but we all know the fun really started Sunday, when absolutely nobody is allowed to tell the Reds they can't win more games than the Cubs. Realistically, they probably won't, but mathematically, they just might."
 
Love to see a chart comparing ages & games experience. There's a phenomenal gap between the Swans experienced 9 players and their next 13 tonight. I blogged about it, but can't keep linking it here - check it out on the Swans board.

Final Siren what do you think? Would be cool to have an age/experience squiggle :D
 
Yay Squiggle has us 10th! Moving up in the world
u5OYeGe.gif
 
SterlingArcher said:
Keep beating sides like the Eagles and doing what is necessary over the easier sides and 9th finals is a monty for Richmond. Especially with Rance returning to form today, and Martin continuing on.
Click to expand...
Fingers crossed Geelong have a bad year. We'll be able to meet our 9th quota with their first round pick
 

(Log in to remove this ad.)

I'm surprised the Swans are still projected to finish 6th at the end of the H&A. I wonder what might happen if Sydney loses to the Eagles (listed as a 50% Sydney win chance by the Squiggle) and GWS (a 44% Sydney win chance). I guess it may depend on the size of any potential losses, rather than just the win or loss, as that has a flow-on effect to the win % for the rest of the projected games in the season.

Final Siren, given the large personnel change at Essendon from 2016 to 2017, do you think the Squiggle currently underrates them considerably? If so, how many weeks do you think it will take for that effect to fall away?
 
igon value said:
I think the Squiggle has Essendon correct. Their wins so far aren't that inspiring.
Click to expand...
I still think they're a little under-rated. The Squiggle currently has them finishing equal second last with 7 wins.

I think, despite yesterday's result, Essendon will finish higher than Carlton, Gold Coast, Fremantle, Brisbane and, if they keep playing the way they are currently, Hawthorn.
 
WhiskeyTangoFoxtrot said:
I still think they're a little under-rated. The Squiggle currently has them finishing equal second last with 7 wins.

I think, despite yesterday's result, Essendon will finish higher than Carlton, Gold Coast, Fremantle, Brisbane and, if they keep playing the way they are currently, Hawthorn.
Click to expand...
Has them all at the right end of the ladder though!
 
You must log in or register to reply here.

Similar threads

Always Ballin
Carlton 2022 = Melbourne 2017
2
Replies
46
Views
1K
Gysberts2Bate
Gysberts2Bate
CASEYS9420
Opinion Richmonds 2017-20 Dynasty would have beaten every other Premiership side of the 2010's
10 11 12 13 14
Replies
344
Views
10K
Nebular
Nebular

Remove this Banner Ad

Back
Top