Experimental 361

[Look Out There Are Llamas]

If there are minimal player movements then the 30 game things makes perfect sense. But, we've had a transfer window and squads will be quite different.

 

[shoddycollins]

I was reading a website the other day about using a ratings system to calculate win/draw/loss percentages. Apparently, based on historical results, if two teams with exactly the same rating are playing each other, the possibility of the home team winning is around 46.5%. Thus you should be looking at 47% and higher for 'fancied' home teams.

That would leave you with Pompey, Colchester and Luton as the genuinely underperforming home teams last weekend.

Not really, the percentages I put in my post are the ones that e361 are giving; he already takes account of home advantage according to his explanation (though clearly not a specific team's home form) and since I'm critiquing how well his predictions have performed this weekend I won't take account of home form again. I'm simply saying whether giving these teams likelihoods of over 40% to win was realistic.

 

[shoddycollins]

Wherever e-ratings are proving a poor indicator of a team's performance then the usual thing is that the team either has problems putting away the chances they create or has problems saving the chances their opponents create. According to his explanation this is expected and he considered trying to include each teams effectiveness at either end, but prefers instead to have a table based on chance creation as this is more informative when compared with what actually happened. He says finishing and shot-stopping abilities are difficult to quantify, I guess what he means is that even if you did quantify it, you'd be measuring the ability of maybe only one or two players who take the majority of the shots, and that could prove too erratic, especially as different players tend to take shots from different areas of the pitch.

 

[BeesKnees]

Some good new stuff from Ben on experimental 361
Analysis of each teams attacking threat. No surprise that Grimsby and ourselves are one man teams.

https://experimental361.com/2016/11/24/visualising-each-clubs-attacking-threat/

 

[BeesKnees]

Also a position probability after tomorrows games

 

[Luke Imp]

Bit of a bump on this for those interested in 361:

Ours is a bit misleading because we had a very, very settled first XI last season. It was only really two positions that changed.

 

[shoddycollins]

Bit of a bump on this for those interested in 361:

Ours is a bit misleading because we had a very, very settled first XI last season. It was only really two positions that changed.

I don't think it's misleading really, it tells you exactly what it's supposed to tell you. Lincoln's percentage is high, because of both a settled first eleven making up most of their minutes and the retention of those players, with the players who have left being largely further along the 'clock' where the players who didn't play much are. Other teams are lower down because either a few key players have left or lots of squad players have left. Notts County and Accrington are the other two teams to have kept all the key players that accounted for over half their season but have released more bit-part players.

Three of the relegated teams are in the bottom four with almost everyone who featured in League One for them departing, Coventry seemingly more confident that their players can do a job in League Two.

Wycombe, Grimsby, Carlisle and Morecambe are the only teams who's top appearance-makers have departed (probably goalkeepers) though Wycombe are still pretty high up, mainly because they haven't released squad players. The other three have large chunks missing from the right side of the graph indicating many key players departing.

Cheltenham would have been near the top of this last summer as they went into their return to League football with pretty much the same squad that romped to the National League title the previous season, and therefore this is the summer that most of those players have departed and they look to build a League-standard squad. Something for Lincoln fans to bear in mind. It begs the question since Lincoln are 5th favourites and Cheltenham weren't particularly fancied last year (apart from one Cheltenham fan) why Lincoln are so different?

Have the bookies seen something in them that suggests they'll fare better with their non-league squad, or are they just over-reacting to the Cowley brothers' media-friendliness, Lincoln being a bigger club than Cheltenham, and last season's cup run?

Both won the National league pretty comfortably, but Lincoln did it in a more eye-catching manner.

 

[Luke Imp]

Probably a combination of all 3.

 

[Son of Cod]

Wycombe, Grimsby, Carlisle and Morecambe are the only teams who's top appearance-makers have departed (probably goalkeepers)

We've lost the likes of Disley, Pearson, Andrew and Bogle in Jan. Still have McKeown in goal though.

 

[PaulHaddock]

Andrew is easily the big chunk of that.

 

[Luke Imp]

Early graph.

Sums us up - not being clinical as the opposition are being.

 

[Trapdoor]

Bit of an odd looking graph considering Swindon are top of the table....?

Edit: after thinking about it all this graph shows is that there is not enough data to conduct any meaningful analysis. Having looked at his scatter plots for deriving expected goals per match since 2012 you can observe that the data set in question is VERY noisy. There is only just about enough signal to observe a trend when using all games since 2012.... Maybe better to look at this again in December before making conclusions about how your team is performing.

 

[Madejski]

Even in December it is useless. This website and this statistic rated us as the 23rd best team in the Championship last season.. we finished 3rd.

Like any stats, use them but don't get hung up on them. They only prove so much.


Sent from my iPhone using Tapatalk

 

[BeesKnees]

Forest Green really not looking good.
What is going on with Newport.

 

[shoddycollins]

Morecambe occupying their usual spot I see, they still won't go down. This early in the season though sporadic incidents like red cards and the fluctuation in performances you expect week to week can cause a team to move around a lot, also there isn't enough data yet to adjust it based on quality of opponents faced and so that can mean clubs a long way from where they'll find themselves over the season... just like the league table really

 

[PaulHaddock]

The fact Cheltenham and Exeter are next to each other despite being at opposite ends of the table shows this graph is pointless as this stage.

 

[BeesKnees]

The fact Cheltenham and Exeter are next to each other despite being at opposite ends of the table shows this graph is pointless as this stage.

I disagree.
The graph factors in how many and how good the chances are. It does not factor in how good a team is at putting away the chances they create, nor does it factor in how good the opposition is at converting chances. If you have a striker who never misses or a goalkeeper who fumbles easy saves then the stats won't change, but the goals scored / conceded will.

You need the other half of the story

https://experimental361.files.wordpress.com/2017/09/2017-09-02-l2-att.png

Exeter are far more likely to convert a chance than Cheltenham, suggesting they have more clinical strikers

https://experimental361.files.wordpress.com/2017/09/2017-09-02-l2-def.png

The defensive side of the coin shows Exeter and Cheltenham conced a similar amount of chances but you are twice as likely to score against Cheltenham.

 

[shoddycollins]

I disagree.
The graph factors in how many and how good the chances are. It does not factor in how good a team is at putting away the chances they create, nor does it factor in how good the opposition is at converting chances. If you have a striker who never misses or a goalkeeper who fumbles easy saves then the stats won't change, but the goals scored / conceded will.

You need the other half of the story

https://experimental361.files.wordpress.com/2017/09/2017-09-02-l2-att.png

Exeter are far more likely to convert a chance than Cheltenham, suggesting they have more clinical strikers

https://experimental361.files.wordpress.com/2017/09/2017-09-02-l2-def.png

The defensive side of the coin shows Exeter and Cheltenham conced a similar amount of chances but you are twice as likely to score against Cheltenham.

True but at this stage of the season we can't say for sure how clinical Exeter's strikers truly are or wasteful Cheltenham's as we're basing it on a relatively small number of chances. It could also be that Exeter's strikers have just been luckier than most and they won't continue to score at this rate. The guy who puts these together has often said that based on past experience those differences even out, no striker is so clinical that they keep scoring at a decent rate even if not provided with many chances and so on, so as the season goes on expected goals has become a decent indicator of where a team should be.

 

[Trapdoor]

he should show a regression for all strikers of number of shots on taget vs goals scored.

i bet it's not as clear cut as it sounds.

 

[BeesKnees]

he should show a regression for all strikers of number of shots on taget vs goals scored.

i bet it's not as clear cut as it sounds.

The BBC help a bit
http://m.bbc.co.uk/sport/football/league-two/top-scorers

They also have the stats for each club.
http://m.bbc.co.uk/sport/football/teams/cheltenham-town/top-scorers

Reid currently gets 82% on target and scores around 50% of them.
Eisa is 63% and around 40%

 

[CCFC_Charlie]

Experimental 361 is very interesting stuff, but like all stats must be looked within context and not as a independent marker.

As for the graph posted in the thread, it tells us Cov fans pretty much what we already know. Our defence is class but we can't score for shit. Without two appalling keeping mistakes we'd have only conceded one all season which is encouraging at this stage, but the goalscoring problem really does need to be addressed.

 

[shoddycollins]

he should show a regression for all strikers of number of shots on taget vs goals scored.

i bet it's not as clear cut as it sounds.

He actually has done, though that suffers from the same early season random noise. He did it last season too, (I was the one suggested it to him) it did show the majority of strikers starting each week were in a band. There were outliers who were bagging lots of goals from relative scarcity of opportunities but over the season more and more of these dropped into the average band as things averaged out.

 

[Trapdoor]

get him to post here! we could have a stats discussion thread...

 

[AdamStag]

Forest Green really not looking good.
What is going on with Newport.

So it reckons we're the best?

 

[BeesKnees]

So it reckons we're the best?[emoji38]

Second best, you concede relatively few chances compared to other teams but you are hampered by the fact opponents score a high proportion of the chances they create

 

[AdamStag]

Second best, you concede relatively few chances compared to other teams but you are hampered by the fact opponents score a high proportion of the chances they create

That's because our GK is fat and useless

 

[shoddycollins]

get him to post here! we could have a stats discussion thread...

He is aware of this thread at least

 

[BeesKnees]

Happy days

 

[SkyBluesRose]

Missing Mansfield vs Grimsby

 

[The Jovial Forester]

No idea what that means but if it says we're a bit shit he's not far wrong