One Stat To Rule Them All Revisited

His poor shooting over the last 3 games has brought his accuracy way down, but clearly on his day he is capable of great hauls

It's great. Check out Statsbomb and @mixedknuts on Twitter to start.

I haven't looked into it too much but a lot of statisticians and people that play similar Fantasy Football games swear by it.

It basically analyses all shots being taken by a certain player and judges how likely he is to get a shit away that threatens the goal. So the likes of Townsend would probably not rank very well.

It looks at factors like the type of balls being played through for the chance, the type of shot in relation to the goal, the whole works. It takes a lot of effort to reproduce the information as far as I know since it is so detailed, but apparently it is the dog's danglies

In a nutshell it assesses all of the shots being taken by a particular player, the way they are being created and how threatening they are in order to analyse how likely a player is to score.

Most xG sites are not updating properly this year due to issues with opta.

more like I miss Suarez

As with everything, context is paramount

For each table there were a good few players that ranked highly for minutes per chance and had a fantastic shot accuracy.

Namely, Balotelli and Dzeko in 2011/12. But the context is that they were always in rotation, yet delivered 13 and 14 goals respectively.

This method will pick up a lot of players that score well over a season, but will not necessarily take value or team status, etc into consideration.

Take Son as a prime example. Kane is out, Son is playing as CF and he is getting many chances and hitting them accurately. A prime example of a player that we expect to be among the top scorers by this method, right? Well for the present, yes. But what about when Kane returns? This is what the table and the method doesn't take into consideration and relies on the user to look out for.

I just don't think PP90 is a reliable stat. It's simple predicating what IS happening, not what WILL happen. Take Hazard for example. Last year we would have saw Hazard's good PP90 and invested in him, me included. He flopped. Likewise Mahrez or Vardy might not have had a fantastic PP90 and we would have disregarded them.

By that logic, Capoue is one to have for the current season based on the PP90 he was clocking up in his first few games. But does PP90 look at sustainability or positional changes (touches in the box even)?

Capoue went from 11 penalty touches, 6 shots, 4 shots in the box in the first 4 gamesweeks (where he hauled) to 3 penalty touches, 3 shots and 1 shot in the box over the last 3 gameweeks. But he is still averaging 6.0 points per 90 minutes, so should we still invest in him by a PP90 statistical model? No, imo.

Stats such as minutes per chance (in conjunction with a good number of chances and time played) and accuracy from a decent field position is a much better indicator of what is expected of the player as we can track any progression/ regression over a given time frame.

Take Redmond as an example. Fantastic underlying stats but his weak finishing is what stops him from being a "Mahrez" and keeps him as a "Redmond" this year. Not saying he won't haul, but it is dependent on his accuracy and clinical ability improving fast.

Not true at all tbh, no reason you can't do both. The two aren't mutually exclusive.

This statement couldn't be more ignorant & dismissive.

And a lot of people are a poor judge of talent, ability and growth.

PP90 is generally a better indicator than PPG because it gets rid of abnormal modifiers like injuries or randomly being subbed off after a red or getting their own red. Of course for certain players these are genuine concerns (someone like Sturridge is injured consistently enough that he could go off early in any match and his PPG therefore may be more relevant. Someone like Coutinho on the other hand is regularly subbed off early so his PPG is also more important because he never actually plays 90 and that would be a misleading figure.

Yes people frequently use chances created and look at average baps for assist and bonus predictions.

Of course using multiple stats to make a decision/prediction is better.

Yes many people have compared the use of different stats and found PPG on its own to generally be more effective than any other stat on its own because it takes all point sources into account. Where it fails is only in a player who is either highly overperforming (Capoue) or someone in a sustained period of underperforming (assuming they come out of their slump. someone like an Ozil here, although he does have 2 goals).

I personally find points per "insert unit of time" almost completely useless as a future predictor, because if you're looking at it in tandem with a future-facing stat (such as shots in the box or on target, etc.) then you're pretty much basing your decision on the latter but the former is giving the more frightful a warm blanket to wrap themselves in that caresses them and says "well he has scored before so it's not too big a gamble".

Yeah I think assists are harder because you could have a player creating lots of quality chances for a donkey of a player regularly.

Maybe minutes per chance created and big chances created would be a good starting point.

I wonder if you then factored in the conversion rate of the respective strikers, would you get a better metric?

I haven't looked into it but a good topic for down the line definitely.

"just by looking at previous year's PP90"

No. Potentially different manager, different teammates, different fixture schedule, different teams in the league

"and age of player."

cmon man lol..

"I was in the top few hundred players in FPL for the early months last year by taking players at breakout ages with high PP90 on a good sample from the previous season (such as Mahrez whom I had from the start - I dropped some as the year went on as I play in a draft mini-league without open transfers)."

Pure coincidence and last year was an outlier.

"PP90 is less reliable in-season with a smaller sample"

It's always unreliable.

Who's taking captaincy decisions based on just PP90?
Most FPL managers will either take a number of factors into account, or else captain Kun or the Poll Leader regardless.

There are of course exceptions to PP90 or PPG, as there also are to predictions based on underlying stats (such as in the cases of Ramsey and Yaya in 2013/14), but they are far from being random.

PP90 or PPG should however ideally be considered in conjunction with other factors rather than entirely on their own.

Similar to what RL has already said - why would you assume that Sigurdsson would score what his "season average PP90" across "a random set of 7 games". Players don't just hit averages. Over the season, yes you would expect there Sigurdsson to do similarly to previous season, if other factors remain the same. But we know that Swansea aren't scoring many goals at the moment (selling Bony one year, Ayew the next are huge factors in that, plus managerial changes, plus possibly any consequence to Iceland's relative success at the Euros (which apart from anything else meant he wasn't playing in the first few games). Over the season though, yes you would probably expect that his PP90 will revert to something closer to his usual, along with the other underlying stats.

xG based on how many fixtures?

Similarly, most of those players rank very highly on my own table too.

Rashford and Kane are one of a large group of players that are on the fringes of this model.

Benteke is far down on my own table too as his shot accuracy is very poor in comparison to the others, but he has 3 goals nonetheless.

But xG is a very good system and I hope to get familiar with it a lot more

Would be interesting to see how strong a correlation there is between this group and the 10 furthest forward players on heat maps. Just scanning the list, and especially with the inclusion of Antonio (who has practically been playing keeper for the opposition he has been so far forward when I've watched them recently), these seem to be the lords of the six yard box.

This!!!