Computer predicts final league placings, but how good is the computer?

By Tony Attwood

When I see the words “super computer” used in a football blog I tend to shy away from the report, simply because I have no idea what “super computer” means in terms of anything, let alone football.

I mean, the little laptop I am using to write this while in Australia is about 1000 times more powerful than the computer that was used to land the first men to reach the moon.   So super computer to crunch a few game results and come up with a prediction?  No I suspect any regular computer could do the job.

Anyway, what they are reporting from their so-called super machine is a set of results that lead to a top six of

Wolverhampton come fifth, with Cardiff, Fulham and Huddersfield going down.  Of course if you fancy betting on this you could have a go with

Manchester City
+74 94 >99%
+63 92 >99%
+30 75 70%
+30 75 63%
+24 72 35%
+23 72 32%

Wolverhampton come fifth, with Cardiff, Fulham and Huddersfield going down.  Of course if you fancy betting on this you could have a go with

The problem is of course we don’t have the details of exactly how they came to this conclusion and whether they actually did include the difference in results that Arsenal get at home and away.

And in effect there is little difference between their final table and that which we have now…

Pos Team P W D L F A GD Pts
1 Liverpool 31 23 7 1 70 18 52 76
2 Manchester City 30 24 2 4 79 21 58 74
3 Tottenham Hotspur 30 20 1 9 57 32 25 61
4 Arsenal 30 18 6 6 63 39 24 60
5 Manchester United 30 17 7 6 58 40 18 58
6 Chelsea 30 17 6 7 50 33 17 57

The question arises however as to why Tottenham will end up with more goals scored than Arsenal in the prediction – for it is goals scored that raises them above Arsenal in the “super computer” table.

As things stand Arsenal have scored six more league goals than Tottenham, and we are continuing to score from all over the place.

21 Feb 2019 Arsenal v FC BATE Borisov W 3-0 Europa League
24 Feb 2019 Arsenal v Southampton W 2-0 Premier League
27 Feb 2019 Arsenal v AFC Bournemouth W 5-1 Premier League
02 Mar 2019 Tottenham Hotspur v Arsenal D 1-1 Premier League
07 Mar 2019 Rennes v Arsenal L 3-1 Europa League
10 Mar 2019 Arsenal v Manchester United W 2-0 Premier League
14 Mar 2019 Arsenal v Rennes W 3-0 Europa League

17 goals in the last seven games is a fairly decent return.

Tottenham’s recent results show us ten in the last seven.  And in case the argument is that these were tougher games, if we take out the two Champions League games and replace with the previous league games they still don’t exceed us.

02 Feb 2019 Tottenham Hotspur v Newcastle United W 1-0 Premier League
10 Feb 2019 Tottenham Hotspur v Leicester City W 3-1 Premier League
13 Feb 2019 Tottenham v Borussia Dortmund W 3-0 Champions League
23 Feb 2019 Burnley v Tottenham Hotspur L 2-1 Premier League
27 Feb 2019 Chelsea v Tottenham Hotspur L 2-0 Premier League
02 Mar Tottenham Hotspur v Arsenal D 1-1 Premier League
05 Mar Borussia Dortmund v Tottenham H W 0-1 Champions League
09 Mar Southampton v Tottenham Hotspur L 2-1 Premier League

I think if it does come down to goals scored, we could well win out on that too.

14 Replies to “Computer predicts final league placings, but how good is the computer?”

  1. The so called super cpmputer brought demise to thousands of lives when the newly aquired Boeing 737 crushed and killed all 157 in Ethiopia. I don’t even want to think what will happen when those gadgets get hold of the World.

  2. I guess Tottenham is there above because points are rounded, so it can be like 75,01 against 74,98 with far moge digits in reality
    another predicion from euroclubindex
    1 Manchester City (4089) 92
    2Liverpool (3840)91
    3Tottenham Hotspur (3574)76
    4 Arsenal (3439)75
    5 Manchester United (3671)74
    6Team Chelsea (3570) 72

  3. They may well be taking into account our poor away form hence a poor goals for return ?

    Personally I cannot see us achieving an average 2.5 goals for per game with so many of them away.

    The way our seasons gone I’d fancy our chances much more against top 6 teams at home than that 2nd group away.

  4. What is a supercomputer?

    A supercomputer is something that is significantly beyond the ability of a “typical computer”, usually in terms of the ability to compute some some numerical problem.

    In a sense, every GPU in most people’s PC’s could be considered a supercomputer. There are only very limited sets of problems these devices can attack (such as bitcoin “mining”), but they do so significantly faster than a general purpose CPU can do the same computations.

    Most people today have PC’s with 2 CPU cores. And usually a single GPU. I believe the large end of the Top-500 Supercomputer is around 100,000,000 GFLOPs. The GPUs people have in their PCs are typically in the 1,000 to 10,000 GFLOPs range. So the top (general purpose) supercomputer is about 30,000 times more capable than the highly specialised GPU in a PC.

    IBM Summit is the current leader of the Top-500 Supercomputers, and it has 9,216 general purpose CPUs, each with 22 cores (202,752 cores) and 27,648 high end GPUs.

    I live in a part of Canada which is about the size of Germany, except there is only 2-3 hundred thousand people living here. All the weather researchers in Canada are more concerned about Toronto (and southern Ontario), Montreal, Vancouver and other places, than they are about this huge farming region. So, I am going to try and use the supercomputer calculations of Global Circulation Models “downscaled” to this area (The Peace Country). I have 6 computers with 44 cores and 5 fair GPUs. I expect this is enough computer power to at least approach the question. In a sense, I have a supercomputer for this region.

  5. I approached this problem last week using a very non supercomputer brain and came to similar numbers with us finishing on 74 points, one behind Spurs on 75. Gord, if I rember rightly has both us and Spurs finishing on 75 points in third and fourth places.

    What none of us and certainly supercomputers are incapable of calculating are any possible machinations by the PGMO.

  6. The PGMOL are a big worry.

    I hardly dare think about the effort they will put in to assure Spurs and United attain a Champions League spot.

    When you look at the shite Spurs have already got away with what with head butts, stamps, studs up challenges etc. heaven forbid what will happen now.

    As for United only a clean break of the leg will be considered for a red, and even then of course it could be overturned if he ‘didn’t mean it’ .

    I’m not hopeful to be honest.

  7. manure were on the best run of form before we beat them and I cant see that changing. I think they will beat chelski and maybe even city so they are my favourites for third.
    The spuds run in could depend on their game against liverpool, confidence wise if they lose that game they are continuing their bad form and may just slip away. As for chelski they blow hot and cold but may concentrate on the europa league.

  8. Why use a supercomputer?

    As of the end of Game Day 30, there were 300 games played. The results of a game are determined by the number of goals scored by each team. The probability distribution applicable for the scoring of goals is (very nearly) a Poisson distribution.

    What I would start with, is for each game, I would draw 10,000 random games In other words, we would have 10,000 leagues running in parallel. If Team A scored 2 and Team B scored 1 in the real league, we would draw 10,000 Poisson deviates where the population average for Team A was 2 and 10,000 Poisson deviates where the population average for Team B was 1. And do this for all 300 games in any one of these 10,000 Leagues. People are expecting whole numbers, so I would find the median points acquired, median goals for, median goals against for each team. And then treat those medians as the “true” results.

    Then, for the games remaining, the supercomputer could attempt to predict the probability of winning, drawing and losing each of their remaining games, similarly to what I am doing.

  9. Rescheduled games (from

    Arsenal v Crystal Palace
    Sunday, April 21 (originally Saturday, April 20)
    Kick Off: 4pm


    Wolves v Arsenal
    Wednesday, April 24 (originally Saturday, March 16)
    Kick Off: 7.45pm

    The TV people are expected to wait and see how many foreign Arsenal fans sign up to make the trip for either game. If lots of foreign people sign up, the TV people will then demand the game be rescheduled so as to cause a problem for the foreigners.


    Wenger @32, below Someone &klopp? The disrespect. I’m a WOB, but even I won’t have him so low and below those guys in a list of all time greatest coaches. I won’t even have Zidane above him, not yet. In his short career he definitely achieved more than Wenger, but the career has been too short to make a generalization on. However, I’d put van gaal above Wenger in that list

  11. Tomp

    how many trophies has Klopp won compared to Leprof Wenger?that list is a lot of bs

  12. The key to Wengers success is the percentile increase in value of the club over the period of his managership. There isn’t any other club at the level Arsenal grew.

    The trophies & titles are cherries on the cake but the value of the club is the cake.

Comments are closed.