Computer Says: Norwich to Win League

[Badger 2,405]

Thanks OP. Please keep us updated!

Some interesting stuff on this thread and certainly cause for optimism. Michael Bailey included something similar based on the fact that we had scored 30 points after 16 games and looked at the historic outcomes (see below). It suggested a 68% probability of gaining promotion which is very close to the 65% suggested by FiveThirtyEight.com (thanks BYG).

A two in three chance seems pretty good for me and probably consistent with Ricardo's 80% prediction of the play offs?

 

Michael Bailey's article - at the bottom:

In the last 10 Championship seasons, 28 teams have reached 30 points or more after 16 games – which has included NINE of the 10 title winners, 15 of the 20 teams automatically promoted and FOUR of the 10 play-off winners.

Therefore of those 28 teams, NINE have missed out on promotion – three of which were Middlesbrough: 2011-12, 2012-13 and 2014-15 at the hands of Norwich at Wembley. Norwich also accounted for Cardiff’s failure in 2010-11.

Reading are the only team to win the title having failed to break 30 points after 16 games (21 points, 2011-12), while Sheffield United and Bristol City both reached the early benchmark last season alongside Wolves and Cardiff, who were automatically promoted. United and City finished 10th and 11th respectively."

 

So 19/28 (68%) of the teams reaching 30 points by this stage of the season have gained promotion. IF the past is anything to go by we have a two in three chance of getting promotion and a 1 in 3 chance of not doings so. Evidence for both the optimists and pessimists! 😀

I don't gamble, so I don't know how this translates into betting odds. Perhaps some of my risk-taking fellow supporters could enlighten me?

http://www.edp24.co.uk/sport/norwich-city/sheffield-wednesday-michael-bailey-six-things-learned-buendia-farke-safety-1-5764520

 

[Nuff Said 5,142]

20 hours ago, Petriix said:

The data is 100% factual historical information - it's literally just the results so far this season. The question was "what would happen if all the teams continued to get similar results based on the current position of their past and future opponents?"; it's not really a prediction, just an extrapolation.

I was going to post something similar. These days it is not about the data so much as what you do with it. Financial institutions are spending many millions doing something along these lines with stock market data in the hope of better predicting performance, but no-one has shown (or least publicly said) they can do it accurately in the long term. A one-off correct prediction means nothing (stopped clock...). The issue is recognising what is a cause and ignoring irrelevant correlations. I saw an article this week which showed that the US stock market has always gone up more when the McRib is on sale.

I wonder if FiveThirtyEight .com are guilty of doing that - e.g. overestimating the value of squad worth, and not including less tangible factors. The SPI is apparently based purely on last season's position and the value of the squad - how does that factor in the ‘value’ the coach/manager brings?

I have realised typing this that it actually does agree with your post Swindon - you need to identify what is the relevant data and weight it accordingly.

[Badger 2,405]

25 minutes ago, Nuff Said said:

I was going to post something similar. These days it is not about the data so much as what you do with it. Financial institutions are spending many millions doing something along these lines with stock market data in the hope of better predicting performance, but no-one has shown (or least publicly said) they can do it accurately in the long term. A one-off correct prediction means nothing (stopped clock...). The issue is recognising what is a cause and ignoring irrelevant correlations. I saw an article this week which showed that the US stock market has always gone up more when the McRib is on sale.

I wonder if FiveThirtyEight .com are guilty of doing that - e.g. overestimating the value of squad worth, and not including less tangible factors. The SPI is apparently based purely on last season's position and the value of the squad - how does that factor in the ‘value’ the coach/manager brings?

I have realised typing this that it actually does agree with your post Swindon - you need to identify what is the relevant data and weight it accordingly.

As I understand it Nuff Said, the FiveThirtyEight .com is dynamic. It gets its starting positions from squad worth etc etc but then uses match performance metrics (adjusted goals, shot-based expected goals and non-shot expected goals) once the season has started.

This would seem to me to be reasonable data and is designed to reflect the overall quality of play rather than the simpler metrics of goals and points, which will include an element of randomness and luck.

[Molly Windley 76]

That fivethirtyeight prediction table looks a bit hooky to me..

To finish top with 79 points and second with 76 points would be a season as yet unseen in the Championship.

Figures below are from the 95/96 season to present

Edited November 10, 2018 by Molly Windley

[Badger 2,405]

Molly my guess is that this reflects the closeness of the tables at the moment. The model is dynamic and I suspect that the top points score will increase as the season develops.

Rather than looking at looking at the points score which is even harder to predict than who will be promoted, I preferred to look at the overall chance of promotion - about 65%, which seems reasonable and consistent with other modelling.

[Bethnal Yellow and Green 1,560]

It is worth looking at previous seasons to see how accurate their predictions are. About this time last season they were correct in predicting Wolves and Cardiff coming first and second, but had Fulham well off their eventual 3rd place finish. 

All things like this are of course to be taking with a big pinch of salt, but I’d much rather be in a position where the league table and underlying numbers both suggest positive things, rather than when Norwich were last top of the league under Alex Neil and the numbers screamed that Norwich were due to fall away - which they of course did. 

[It's Character Forming 1,161]

I agree completely with BYAG

 

when we were top under AN the defensive weaknesses were clear and were pointed out at the time on this forum so it was no surprise to me when our form dropped away.

 

at the moment there are really encouraging signs of everything clicking into place the way Farke wants so IMO solid grounds for optimism about this season.

[Petriix 2,838]

Here's the updated more realistic version after yesterday's results. I've slightly tweaked the algorithm based on the fact that teams average closer to 1.37 points per game across the whole season. Interestingly fivethirtyeight.com now have us missing out on the title on goal difference on 77 pts.