Eyeballs v stats

[bytholwyn]

The issue I mostly have is with people who quote, tweet, repost, whatever stats without having the first inkling of what went into generating them.

The one that springs immediately to mind is the recent one with Tymon, where he apparently created the most chances of any player in the division. This was immediately used by some as a stick to beat Vipotnik with.

Ask these people what actually constitutes a chance, what methodology was used in the collation of these stats or where these chances were and they have no idea. None whatsoever. It's a bright shiny number that they think backs up their argument so they throw it in there. Without context it's a completely meaningless number, and football is chock full of those these days.

I noticed early on this season that Tymon had very good expected assists stats and therefore paid particularly close attention to his efforts from then on. And what my eyeballs revealed were a steady stream of excellent crosses that were begging to be buried, which Vipotnik, Bianchini, Ronald and Eom consistently squandered. So, observations absolutely backed up the stats. This is a perfect example of how stats can provide useful feedback that help gain a deeper understanding of what unfolds.

The fact that Tymon's crossing excellence has not been more widely acknowledged is because when the chances that he's created have been routinely missed, the emphasis, understandably, is on the failure to score rather than the quality of the cross. Which is a limitation of the eyeball test, the natural cognitive bias to put almost exclusive attention on good or bad shooting or defending.

 

[CygnusX1]

Shits on target?

No-one likes a shit off target

 

[Dr. Winston]

Shits on target?

Like Shechters?

 

[J_B]

Assists last season:
Tymon 7
Ronald 7
Peart-Harris 4
Darling 3
Cullen 3
Key 3

 

[bytholwyn]

Assists last season:
Tymon 7
Ronald 7
Peart-Harris 4
Darling 3
Cullen 3
Key 3

Expected assists:

Solomon 10.3, Tymon 9.4, Josh Murphy 9.4, Dan James and Max Bird 8.7.

 

[Swans Jack]

I'll leave others to enjoy this thread. The Grimes one and now this one, talking about stats feels like I have just got a job in the stock exchange. More interested in talking about the summer ahead and who is going or comings

 

[Dillwyn the Dog]

Assists last season:
Tymon 7
Ronald 7
Peart-Harris 4
Darling 3
Cullen 3
Key 3

Grimes?

 

[J_B]

Grimes?

2
1 for us and 1 for Coventry.

 

[bytholwyn]

2
1 for us and 1 for Coventry.

Expected assists 6.63. Another victim of our woeful finishing. Only Luton had a bigger difference between goals scored and expected. We need a clinical finisher apart from Cullen.

 

[J_B]

Expected assists 6.63. Another victim of our woeful finishing. Only Luton had a bigger difference between goals scored and expected. We need a clinical finisher apart from Cullen.

He's ever so unlucky to be at two different clubs whose strikers can't finish his brilliant passes.

 

[bytholwyn]

He's ever so unlucky to be at two different clubs whose strikers can't finish his brilliant passes.

Indeed.

 

[JackSomething]

The thing about eyeballs is that recollections are infamously inaccurate, people will more consistently remember what annoys them and forget what goes their way. It takes more training than people realise to properly scout and analyse just based off of what you see, almost like watching a highlights reel of a player and using that to form the basis of your scouting report.

To be fair, this is true.

I remember being at the Barnsley Capital One Cup game (the year we won it) and Mark Gower coming off the bench for what I think was one of his last appearances with us. I was well aware of the group of Plantswans posters who disliked him and what was likely to follow online later that evening, so I watched him carefully and was sad enough to count his passes etc.

I can't remember the exact stats of course, but it was something like he made 12 passes and 11 reached their target. That evening on PS, one of the usual suspects posted that he was terrible when he came on and gave the ball away regularly. They didn't like it when I posted the stats I had recorded because it didn't match with their preconceptions.

 

[Colin-swansea]

Training Ground Guru | Donough Holohan: My 10 rules for monitoring

Training Ground Guru | This is THE portal for staff working in – or aspiring to work in – professional football.

trainingground.guru

An interesting article by Donough Holohan from a High Performance Consultant. He was previously the Head of Physical Performance for Manchester City, working for the club for almost 13 years.
My 10 simple rules for monitoring:

1. For now, human connection beats data collection – every time.
2. If you’re short on humans, collect something that has a chance to make a difference. But be very selective.
3. Don’t collect data that you can’t action in an appropriate timeframe.
4. Being the scientist means you’re also the one with the responsibility to call out the limitations of the science.
5. Understand all of the costs of your monitoring approach. Nothing in this process is truly free.
6. Time of day matters for all measures.
7. Data gathered through subjective questionnaires will be influenced by the use or expected use of that data. Don’t be naïve.
8. Take time to understand how your manager and coaches view fatigue. Then build monitoring solutions that will help their decision making.
9. Pick up to five key strands to your physical performance plan that you believe will maximise player preparedness. Monitor player adherence to these strands.
10. Be brave, be bold, be original. There’s so much groupthink in this space. Don’t be afraid to challenge the dogma.

 

[Dillwyn the Dog]

2
1 for us and 1 for Coventry.

I’m sure someone was saying he was better than that because of his other stats. Eyeballs win again

 

[J_B]

I’m sure someone was saying he was better than that because of his other stats. Eyeballs win again

Considering the fact that he took most of the dead ball situations for both his clubs, two assists all season is extremely poor.