Stats and stuff

[Dave Sutton's barnet]

Over on the Fulham-Bolton thread, Boris wrote:

According to WhoScored, Pratley's tackling average is 2 won per game, which is 38th out of 78 in the division (sliced by: central midfielder or defensive midfielder, 4+ games). His average attempted tackles per game of 3.3 is 32nd in the same sample set.

As a comparison to his team-mates, per game, Karl Henry wins 1.4 of his attempted 2.8 tackles per game; Karacan 2.1 of his 2.9; Vela 2 of his 2.8; Cullen 1.2 of his 2.4.

Out of (wider) interest, our most prodigious successful tackler is Mark “he can’t defend, you know” Little, with 3.8 successful tackles per game out of an attempted 4.3; Beevers is next on 2.2 of 2.9, then Wheater 2.2 of 2.8, Robinson 2 of 4.3; Ameobi, interestingly, averages 1.4 tackles won of 1.6 attempts per game, so he very rarely gets beaten in the tackle.

More team stats here, should anyone care to delve.

It's (sort of) good that our centre backs don't tackle much. It's worrying that our midfielders don't tackle much.

Nobody tackles as much as I expected. It’s now more than half a decade since I had to start studying football stats as part of my job, and the biggest surprise was how few tackles there were (or were enumerated). Perhaps it’s an English mindset - we win the ball back by muscling opponents off it – but tackling is only a small subset of how possession is turned over. Interceptions are almost as common, but the most frequent way of losing the ball is the clearance. Although it may be different if Opta started tracking Sunday League, no matter which competition I’ve analysed (World Cups, Euros, Champions League, Europa League, la Liga, Serie A, Bundesliga, Ligue Un, MLS, Prem and Champo), tackle numbers are surprisingly low.

For instance, WhoScored has the Championship’s top tackle-winner per game (4+ appearances) as Barnsley left-back Matty Pearson with 4.2; Preston left-back Joshua Earl is second (3.9), Mark Little is third (3.8 ) and QPR’s Massimo Luongo is the top central midfielder with 3.7; Ben Pearson and Calum Woods then make it three Prestonians in the top six, which probably says a lot about their early-season success. But in all the division there’s only 19 regular players who average 3+ tackles made per match.

If you widen it out to tackles *attempted*, the left-backs switch position at the top of the table, with PNE’s Earl (5.5) leapfrogging Baaaarnsley’s Pearson (5.4), while in third is Forest’s Liam Bridcutt – although with only 2.7 wins out of 5.3 tackles, he’s losing almost as many as he’s winning. You do get more centre-mids high up the “attempted tackles” table, but they’re not always very successful: QPR’s Luongo wins 3.7 of 5.2, Prestons’ Ben Pearson 3.6 of 5.1, but Leeds’ Kalvin Phillips only wins 3 of 4.8 and Barnsley’s Joe Williams only 2.8 of 4.7. Our highest-placed players here are Robinson and Little, averaging 4.3 per match, but Little’s 3.8 wins dwarfs Antonee’s 2. With 3.3 attempts per match Darren is our busiest central midfielder and 33rd of 78 in the division; Karacan is 47th with 2.1 from 2.9, Cullen 49th with 1.5 from 2.9, Henry 53rd with 1.4 from 2.8.

As a team, in terms of average attempted tackles per game we’re seventh in the league with 27.4 (Millwall and Hull are top with 28.8, QPR and Birmingham bottom with 21.4). For tackles won we’re 10th with 17.4 (top again are Millwall 19.7 and Hull 19.3, Birmingham bottom with 14).

And finally, as I researched it, you might as well have this:

Opponents – tackles won / attempted (%won) – Pratley’s figures
Fulham – 23/40 (58%) - DP 3/7 (only Little attempted more)
QPR - 18/29 (62%) – DP 3/3
Sheff Wed - 11/14 (79%) – DP 0/1
Villa - 22/27 (81%) – DP 5/5 (nobody made more successful tackles)
Bristol C - 27/34 (79%) - DP 4/4 (only Little made more successful tackles)

[boltonboris]

I get that most play is broken up by interceptions, but FFS, we're bottom of the league with 2 defensive midfielders. You really would want 1 of them to be pressing into tackles

I also get that shapes are more fluid now and rotational so there are always more passing options for teams who get pressed, but with a team as limited as ours, with apparent aggression, I'd expect more

[Dave Sutton's barnet]

All it took was a phrase from BWFCi - to wit, “Games at this level aren't generally full of chances” – and I found myself digging in a number-mine…

As with the first post on this thread, the following research uses expected goals (xG) – a method that not everybody likes, but arguably a more useful statistic than shots (on target) because it weighs each chance as to how strongly a player should have scored from a given opportunity.

Of the 40 league games we’ve played, we’ve only had a greater xG than our opponents – ie we should have won - eight times. We’ve actually won nine, but the other side of that is that we’ve deserved to lose 32 of them, and we’ve only lost 19.

We had greater xG in two of our first three games - home to Leeds and at Birmingham – but then we didn’t again for nearly four months, even when results picked up.

Let’s split the season into three by using two big dividing lines: the autumn international break when we regrouped around Karl Henry and Sammy Ameobi, and the transfer deadline day when we sold Madine and had to reshape again.

Screen Shot 2018-04-05 at 13.49.21.png (89.16 KiB) Viewed 4172 times

Over the first 11 games of the season, ie before the autumn international break, our average xG was 0.84 and our opponents’ was 1.44.

From that break until the January transfer deadline, our average xG jumped up to 1.25 - ie we were creating enough chances to be about 50% more likely to score. Our opponents’ xG also jumped up to 1.75, but overall, it worked.

Since Madine, our xG has dropped back to 0.88 - indeed, only twice (against Villa and at Reading) have we created 1.0+ worth of chances. Our opponents’ xG has slipped slightly to 1.56 – intriguingly, marginally higher than it was in The Bad Old Days of autumn when everybody beat us. I guess they didn’t have to work hard, but we were remarkably efficient at losing. (Props perhaps to Alnwick: our record with Howard this season is W0 D2 L5, although to be fair he hasn’t played a league game since the autumn break.)

Speaking of being remarkably efficient at losing, it’s worth noting that of the eight games we’ve ‘won’ on XG, we’ve only actually won twice: home to Cardiff and Villa. We drew at Birmingham and Reading and home to Ipswich, losing at home to Leeds and Burton and at Forest, where we somehow managed to concede three times from a Forest xG of 0.9.

For the record, these numbers were mined from Experimental361, who diligently creates pleasing graphics from each game's accumulating xG scores, with a vertical jump for each chance (and a dot for a goal). For instance, here's our home game with the Owls, easily our highest xG of the season but also the sixth-highest we've allowed our opponents:

2017-10-14-bolton-sheff-wed.png (10.92 KiB) Viewed 4172 times

https://experimental361.com/category/vi ... timelines/

[Prufrock]

Genuinely think xG is meaningless to the extent it is utterly worthless. How on earth do they expect to put a figure on whether someone "should" score. The variables are beyond our capabilities.

[Dave Sutton's barnet]

Genuinely think xG is meaningless to the extent it is utterly worthless. How on earth do they expect to put a figure on whether someone "should" score. The variables are beyond our capabilities.

That's your right. I understand and I share some of your concern, principally over the qualitative assessment of quantitative statistics. (A colleague wrote a reasonable piece about it here.) But as they're done scientifically and carefully by an unbiased third party, I think there's some use in them; certainly you could argue that a 45-yard pea-roller which gradually bobbles into a goalkeeper's arms "on target" is much less of a "chance" than, say, Buckley heading wide (thus off-target) from two yards. It's all shoulda-woulda-coulda and we could argue forever on the individual assessments but as with other stat-sets, if it's assessed the same way over a decent length of time and a decent number of games it gives a good idea of trends.

And frankly the trend is that we're creating a lot fewer chances than before January 31st, almost back to how bad it was pre-Henry. I don't like that result, and I'll be perfectly honest and say I was hoping to discover we were creating just as many chances as with Madine thankyouverymuch. But having done the research, I would be a churl to withhold it.

[BWFC_Insane]

Genuinely think xG is meaningless to the extent it is utterly worthless. How on earth do they expect to put a figure on whether someone "should" score. The variables are beyond our capabilities.

That's your right. I understand and I share some of your concern, principally over the qualitative assessment of quantitative statistics. (A colleague wrote a reasonable piece about it here.) But as they're done scientifically and carefully by an unbiased third party, I think there's some use in them; certainly you could argue that a 45-yard pea-roller which gradually bobbles into a goalkeeper's arms "on target" is much less of a "chance" than, say, Buckley heading wide (thus off-target) from two yards. It's all shoulda-woulda-coulda and we could argue forever on the individual assessments but as with other stat-sets, if it's assessed the same way over a decent length of time and a decent number of games it gives a good idea of trends.

And frankly the trend is that we're creating a lot fewer chances than before January 31st, almost back to how bad it was pre-Henry. I don't like that result, and I'll be perfectly honest and say I was hoping to discover we were creating just as many chances as with Madine thankyouverymuch. But having done the research, I would be a churl to withhold it.

I mean that is obvious to anyone at the games. The only people who swear blind that selling Madine had no effect are those who never liked him.

Whether he was a good, bad or whatever player he made our team better. Objectively better. And as much as ALF has worked hard, it simply isn't a replacement.

[Prufrock]

Genuinely think xG is meaningless to the extent it is utterly worthless. How on earth do they expect to put a figure on whether someone "should" score. The variables are beyond our capabilities.

That's your right. I understand and I share some of your concern, principally over the qualitative assessment of quantitative statistics. (A colleague wrote a reasonable piece about it here.) But as they're done scientifically and carefully by an unbiased third party, I think there's some use in them; certainly you could argue that a 45-yard pea-roller which gradually bobbles into a goalkeeper's arms "on target" is much less of a "chance" than, say, Buckley heading wide (thus off-target) from two yards. It's all shoulda-woulda-coulda and we could argue forever on the individual assessments but as with other stat-sets, if it's assessed the same way over a decent length of time and a decent number of games it gives a good idea of trends.

And frankly the trend is that we're creating a lot fewer chances than before January 31st, almost back to how bad it was pre-Henry. I don't like that result, and I'll be perfectly honest and say I was hoping to discover we were creating just as many chances as with Madine thankyouverymuch. But having done the research, I would be a churl to withhold it.

Sure, and cheers for the link. I do still think they've properly jumped the shark though.

I'm broadly pro-stats, they have their limitations and their fetishization is annoying but I think the nerds are usually more on the money than the philistines.

Some thoughts:

1) I don't agree that any stat set assessed the same way over a decent length of time and a decent number of games gives a good idea of trends. You could track the eye colour of goalscorers in that way and it wouldn't indicate a "trend".

2) more importantly, xG, however many times it is described as one, is not a stat. How many shots a team has, while reductive and subject to some quibbles about how you define a shot is a stat. It's an objective recording of a thing that happened. xG is a conclusion, an interpretation.

3) Alan Shearer, bless him, is spot on. Any system that can't distinguish between the same chance falling to Harry Kane and falling to Shane Long is not fit for purpose. Never mind Harry Kane on his left foot v right, or when on a hat trick, or carrying a knock.

4) it's badly named. What it might be getting towards is "quality of chances" in relation to a mythical "average" player. Now broadly with caveats for accuracy that might be useful info. Has a tactical change resulted in creating more or fewer chances. But it doesn't help beyond that. The average player doesn't exist.

5) re: us. Not accounting for who chances fall to is a big reason to be wary of xG. Sure, ALF offers less all round than Nadine, and so you would expect us to create fewer and less good quality chances, but you would also expect (that word) him to take more of the chances we do create. So xG doesn't help predicting how many goals you'd expect us to score based on those chances. Which is a flaw.

[Dave Sutton's barnet]

All fair enough. I still disagree though.

[Dave Sutton's barnet]

Noticed this on our top-performing players this season (source: Squawka)

Goal Scorer: Gary Madine (10 Goals)
Top Assists: Gary Madine (5 Assists)
Highest Pass Accuracy: William Buckley (78% Pass Accuracy with 195 Comp. Passes)
Highest Shot Accuracy: Zach Clough (83% Shot Accuracy with a Total 6 Shots)
Most Completed Passes: Karl Henry (589 Completed Passes with a 77% Pass Accuracy)
Most Chances Created: Filipe Morais (43 Chances Created: 2 Assists and 41 Key Passes)
Most Cards: Karl Henry (12 Yellow Cards and 0 Red Cards)

Also, per @BWFCStats:

At an average of 10.66, @moraisfilipe20 provides more crosses than any other player in this season's @SkyBetChamp.

Is the Postman injured?

[Burnden Paddock]

Noticed this on our top-performing players this season (source: Squawka)

Goal Scorer: Gary Madine (10 Goals)
Top Assists: Gary Madine (5 Assists)
Highest Pass Accuracy: William Buckley (78% Pass Accuracy with 195 Comp. Passes)
Highest Shot Accuracy: Zach Clough (83% Shot Accuracy with a Total 6 Shots)
Most Completed Passes: Karl Henry (589 Completed Passes with a 77% Pass Accuracy)
Most Chances Created: Filipe Morais (43 Chances Created: 2 Assists and 41 Key Passes)
Most Cards: Karl Henry (12 Yellow Cards and 0 Red Cards)

Also, per @BWFCStats:

At an average of 10.66, @moraisfilipe20 provides more crosses than any other player in this season's @SkyBetChamp.

Is the Postman injured?

The commentator on Tower wasn't certain, but seemed to think so today. I'm not sure how many games Morais has played in his best position this season. The poor bloke gets shifted from position to position when he does play!

[Dave Sutton's barnet]

Is the Postman injured?

The commentator on Tower wasn't certain, but seemed to think so today. I'm not sure how many games Morais has played in his best position this season. The poor bloke gets shifted from position to position when he does play!

Ta. Aye, Parky has played him on the left and also as right wingback and (for short spells when losing) right-back; that versatility normally helps him stay in the matchday squad more than if he were merely a one-footed crossing machine.

I actually set out to find out how many goals we scored from set-pieces last season, because if one can be so simplistic as to reduce a season to a statistic, I fancy that's how we got promoted – and I fancy it might be how we avoid getting relegated.

[Prufrock]

Yep. He's a one trick pony, but it's a good trick. We're not an imaginative team, most of our attaching threat comes from banging in crosses.

Clearly Sammy prefers the right, but given how much he drifts anyway, surely worth a change.

[Dave Sutton's barnet]

He's a one trick pony, but it's a good trick.

Ha, that's exactly the conversation I had at Brentford. Bloke in front of me was semi-benignly grumbling that Morais was "a one-trick pony". At precisely that moment, the Postman delivered a cross right on to the bonce of a player who, obviously, ballsed it up. I leant forward and said "pretty useful trick though, isn't it?"

Frankly I don't care if we survive by getting three more 1-0 wins from set-pieces, Morais crossing for Wheavers to nod in. It can't last forever but at the moment it might be our best bet. Eight outfielders have played more minutes than Morais; Ameobi has played nearly double (2459 minutes to Fil's 1530). Sammy brings different things and I'm not for a moment saying drop him, but when I saw Fil was still leading chance-maker I was genuinely surprised and then genuinely frightened.

[nicholaldo]

Iles tweeted an interesting set of figures this morning comparing our results after scoring first with our results after conceding first:

2018/2019

We score first: W4 D2 L0 (PPG: 2.66)
We concede first: W0 D2 L8 (PPG: 0.2)

2017/2018

We score first: W10 D5 L1 (PPG: 2.18)
We concede first: W0 D5 L22 (PPG: 0.18)

It tells us nothing that we don't already know, obviously, but it's incredible to see just how stark the contrast is when it's written down. In short, we better hope that we score first and then hold on. If they score first, we might as well all go to the pub.

Earlier this week, he also tweeted the number of passes we've made per match this season with the result alongside it and looks like this:

West Bromwich Albion – 192 (W)
Bristol City – 224 (D)
Reading – 274 (W)
Birmingham City – 357 (W)
Sheffield United – 377 (L)
Preston North End – 324 (D)
QPR – 475 (L)
Middlesbrough – 330 (L)
Ipswich Town – 307 (D)
Derby County – 248 (W)
Stoke City – 379 (L)
Blackburn Rovers – 483 (L)
Rotherham United – 243 (D)
Nottingham Forest – 297 (L)
Hull City – 461 (L)
Aston Villa – 330 (L)

This isn't enough on its own to prove that we're better off playing a more direct style because the figures above will be heavily affected by who scored first, whether the opposition sat back and let us play or controlled possession, how strong they were compared to us etc.

That said, we certainly appear to be less successful the more passes we attempt. We played fewer passes in the early stage of the season, resulting in us picking up two wins and a draw before our average increased sharply at home to Birmingham City. Although we won that match, we laboured towards victory by a single goal and only really threatened them when Ameobi was brought on after an hour. If we combine our last two home matches, we attempted to make almost a thousand passes, which resulted in few shots on target, no goals and no points. Granted, in both of those matches we went a goal down fairly early, as well as against QPR, and so we spent significant periods in possession trying to get back into it. However, trying to pass our way through doesn't look to be yielding any kind of return (though we might be better at it, I suppose, when we throw Ameobi into the mix?).

If any conclusion can be drawn, I think it's that we're still better off when the opposition come on to us and we hit on the counter as soon as we retain possession by going long to our target man, who's knockdowns we can play off/wide to our wingers who can cross into the box, taking advantage of the extra space created by them having men further forward because we aren't capable of creating that space on our own by passing and moving. Naturally though, if we concede first - especially at home - it's more difficult to do that.

[Dave Sutton's barnet]

Meanwhile in barn-door news:

[Dave Sutton's barnet]

Barn-door update

Last night on another thread I posted the below...

What kind of return are we heading for this season in terms of goals scored? And is it likely to be record breaking?

In 29 league games we've scored 19 (0.655 gpg). Multiplied by 46 that has us on course for 30.137 goals.

In terms of pure goals-scored:
• We scored 28 goals in 1897/98, but that was from a 30-game season. (Out of interest, we still finished 11th in a 16-team top flight; we went down the following season, 17th out of 18, the first season with automatic relegation rather than the proto-playoff Test Matches.)
• In 1970–71 we scored 35 in 42 as we finished bottom of Division Two after a chaotic season of managerial upheaval. Send for Armfield!
• Fewest goals over a 46-game season was 39, er, last year. (We'd scored fewer, but over fewer games, in 2007–08, 1898–99, 1902–03, 1893–94 and 1979–80)

In terms of goals per game:
• We're on 0.655
• That 1970–71 was the worst we've had (35/42 = 0.833 goals per game)
• Second-worst was last season (39/46 = 0.848)
• Third-worst was 2015–16 (41/46 = 0.891)

In other words, unless we suddenly discover a bag of goals, three of the last four seasons have been three of our worst four seasons *ever* for goalscoring – and it's not entirely Parky's fault...

• We're now on 0.633 goals per game
• 1970–71 (35/42 = 0.833)
• 2017-18 (39/46 = 0.848)
• 2015/16 (41/46 = 0.891)

This isn’t just our bluntest season ever, it is so by quite a distance - not far off 25% less able to score than our previous worst.

[officer_dibble]

Parky can add it to his ever growing list of achievements on his CV

[Dave Sutton's barnet]

Thought I'd have a quick stat round-up compared to the rest of the division. According to WhoScored:

In league games this season we've averaged 9.6 shots per game. The only other club to average below 10 is Ipswich, 9.1, also being relegated.

For shots on target we're also second-bottom (2.7 pg, Ipswich 2.5, Norwich top with 5.5).

We're also second-bottom for possession (43.3%). Below us, notably, Birmingham (hence, perhaps, 3-5-2 working against them).

Obviously we've scored the fewest with 21. Then it's Similarly Hapless Ipswich with 25, then Wigan with 32. In other words, the third-bluntest team has scored nearly 50% more than us.

Goals from set pieces: we've bagged 5. Only Stoke and Wigan (4 each) have fewer. Villa, top, have 20. (Last year we got 11, which was 19th-best.) Rotherham have scored three fewer goals than us from open play, but eight more from set-pieces. Them's the differences.

Goals conceded from set-pieces: mid table with 10. So are Rotherham, and Reading.

Pass success percentage is fourth-worst with 66.4%. One in three of our passes fails. (Below us: Millwall, Birmingham again and Rotherham.)

Aerial battles won, we're third from top! 33.8 duels won per game (Rotherham are top, Birmingham second). So it could, I suppose, be a lot worse if the back lot start losing duels – but looking at our peers in that table, it's basically teams who try to sit back and bonce away oppo attacks. It worked in League One. It generally doesn't now.

Number of times fouled per game is a tight stat set but only Sheffield United (10.2) get crocked - or at least referee-admonished crocked - fewer than us (10.3). Again, I guess it's hard to get fouled when you never have the ball. As for the Blades, maybe refs have realised Wilder's a wanker.

[Lost Leopard Spot]

Could you do an internal round up of stats after 33 games from this, last, and season before? ( I was thinking of where we stand on like shots on target, points, minutes of heart attack, etc... Rather than comparisons with Barcelona and/or Oxford United).
Would be appreciated.
Ta.

[Dave Sutton's barnet]

Could you do an internal round up of stats after 33 games from this, last, and season before? ( I was thinking of where we stand on like shots on target, points, minutes of heart attack, etc... Rather than comparisons with Barcelona and/or Oxford United).
Would be appreciated.
Ta.

Might be difficult to drill down into "partial" stats from midway through previous seasons as opposed to season-end totals. I'll see if I can dig anything out, as it might be interesting to compare this season to each of the previous three (relegation, promotion, survival).