Analytics in Football - Official Thread

[Cloudy O'Ranieri]

FBRef are changing data providers, from Statsbomb to Opta.

On the whole it looks like bad news. Broader coverage of leagues traded off for inferior xG data and loss of some useful data points below.

FBref: Leagues! 🇵🇹 Leagues! 🇧🇷 Leagues! 🇲🇽 Expanded Women’s and Men’s Data! New Data Partner! | Sports-Reference.com

FBref: Leagues! 🇵🇹 Leagues! 🇧🇷 Leagues! 🇲🇽 Expanded Women’s and Men’s Data! New Data Partner!

www.sports-reference.com

Data Removed

passes_pressure, passes_by_foot, passes_by_body_part, passes_by_height, passes_by_specific_result, pass_targets, nutmegs, players_dribbled_past
carries, carries_completed, carry_distance, carry_progressive_distance, progressive_carries, carries_into_penalty_area, carries_into_final_third,
blocked_shots_saves, blocked_carries
pressures, pressures_def_3rd, pressures_mid_3rd, pressures_att_3rd, pressure_regains

This one hurts in particular, as I was tweaking a fun Python visualisation around the top nutmeggers in the top 5 leagues

Ah balls. The inferior xG data will be a bit of a pain but hopefully still good enough in most instances for betting purposes.

 

[Cloudy O'Ranieri]

Having that data for some non top 5 leagues is of some interest though.

 

[Philby]

Ah balls. The inferior xG data will be a bit of a pain but hopefully still good enough in most instances for betting purposes.

Yep. I haven't seen any direct comparisons, but the consensus seemed to be that the Statsbomb xG numbers were the industry standard.

They really went deep on their latest methodology update -> https://statsbomb.com/articles/soccer/upgrading-expected-goals/

Having that data for some non top 5 leagues is of some interest though.

From a betting perspective definitely, it's a chance to find some angles in leagues where the lines haven't already been hammered into shape by guys with stats PHDs

Data Removed

passes_pressure, passes_by_foot, passes_by_body_part, passes_by_height, passes_by_specific_result, pass_targets, nutmegs, players_dribbled_past
carries, carries_completed, carry_distance, carry_progressive_distance, progressive_carries, carries_into_penalty_area, carries_into_final_third,
blocked_shots_saves, blocked_carries
pressures, pressures_def_3rd, pressures_mid_3rd, pressures_att_3rd, pressure_regains

This one hurts in particular, as I was tweaking a fun Python visualisation around the top nutmeggers in the top 5 leagues

This is the nutmeg related visualisation that I ran a few weeks ago.

Was looking forward to re-running this a few times later in the season to see who jumps out :-(

 

[Cloudy O'Ranieri]

From a betting perspective definitely, it's a chance to find some angles in leagues where the lines haven't already been hammered into shape by guys with stats PHDs

Hmm...I think you'd probably be unpleasantly surprised there. I noticed Goias (Brazil) were regularly priced highly relative to their league position during the summer. On checking the new fbRef xG figures yesterday, they're fairly shite alright, and haven't much in terms of valuable players either so the prices were probably bang on and they were just riding some + variance.

I'd say there are very few leagues of any sort of significance that aren't being hammered into shape, tbh. Big syndicates can afford to pay statsbomb whatever they want to hose up any low hanging fruit.

Love the viz by the way.

 

[Donald Trump]

Hmm...I think you'd probably be unpleasantly surprised there. I noticed Goias (Brazil) were regularly priced highly relative to their league position during the summer. On checking the new fbRef xG figures yesterday, they're fairly shite alright, and haven't much in terms of valuable players either so the prices were probably bang on and they were just riding some + variance.

I'd say there are very few leagues of any sort of significance that aren't being hammered into shape, tbh. Big syndicates can afford to pay statsbomb whatever they want to hose up any low hanging fruit.

Love the viz by the way.

Are you saying xG isn't much use anymore?

 

[Donald Trump]

Btw, anyone else get the link to the world cup webinar on Nov 15th. I've accepted, but unsure if I can make it yet.

A recording will be available also.

 

[Cloudy O'Ranieri]

Are you saying xG isn't much use anymore?

No, just that there's no easy money to get from it. In fairness, there probably never was for non Bloom or Benham types.

 

[Philby]

Btw, anyone else get the link to the world cup webinar on Nov 15th. I've accepted, but unsure if I can make it yet.

A recording will be available also.

Didn't know about it. Cheers for the heads up.

I noticed that they've got a placeholder section for the 2022 WC and data already in place for previous WCs.

 

[Cloudy O'Ranieri]

Just looking at the data for last season's English Championship and one thing stands out: so many teams underperformed on the attacking end versus xG, some teams to a comical degree.

I expected an upturn in goalscoring so far in this season's version, but the opposite is true.

I wonder if that means we shouldn't make the same assumptions about lower leagues due to the drop-off in talent. Either that or opta's data is shit...or it could be that variance stays wilder for longer down there.

It's certainly making me reluctant to discover data-led betting opportunities from the Championship anyway.

 

[Donald Trump]

Just looking at the data for last season's English Championship and one thing stands out: so many teams underperformed on the attacking end versus xG, some teams to a comical degree.

I expected an upturn in goalscoring so far in this season's version, but the opposite is true.

I wonder if that means we shouldn't make the same assumptions about lower leagues due to the drop-off in talent. Either that or opta's data is shit.

It's certainly making me reluctant to discover data-led betting opportunities from the Championship anyway.

I've heard Mark O Haire make that very point about the data on the betfair podcast. I actually meant to ask you about it, but forgot.