When Losing a Defender Hurts Your Attack: What Will Happen to Hearts Without Callum Paterson?

It would be fair to say that noted laptop enthusiast Ian Cathro has had a rocky start to his managing career. Cathro has seen losses to Rangers, Dundee, and Aberdeen, as well as a draw with Partick Thistle and being taken to extra time in a Scottish Cup tie with patersonstruggling Championship Club Raith Rovers in his short tenure as Hearts manager. Even his surely enjoyable 4-0 win over laptop doubter Kris Boyd and Kilmarnock saw him lose Callum Paterson for the rest of the season to injury and perhaps beyond that.

Paterson had been having a great season before that, among the leaders for Hearts in goals and had the highest expected goals at the club, which for a defender is some feat. Paterson’s contract is up at Hearts at the end of the season. His injury could mean he will stick around at Tynecastle rather than head to pastures new, but regardless Cathro now faces the pressure of bad results piling up without one of his leading scorers to this point. Is the young manager doomed without Paterson or can others pick up the attacking slack in Callum Paterson’s absence?

xg_g-8

Paterson’s goals and expected goals were pretty close, meaning he was not overachieving in an unsustainable fashion. Before his injury, Hearts had the best expected goals for numbers in the league for a club not named Celtic, with an xG total of 41.47 and 1.97 xG per game in the league (for comparison, the league average xG per game is 1.58). While your small sample siren should be going off in your head, in the 3 games (in the league versus Aberdeen and the two ties in the Scottish Cup versus Raith Rovers) Hearts had an expected goals of 0.68, 0.95, and 1.82 (this was in 90 minutes in the 2nd Raith fixture). Before the injury, Hearts had averaged 13.65 shots per game. In these three matches they had 7,8, and 15 shots. You shouldn’t press the panic button yet if you’re a Hearts supporter due to the aforementioned sample size, but you could say their attack has slowed down in those three matches after looking at those numbers. But is it all doom and gloom for Hearts supporters?

xg-p-90_g-p-90-4

Before we go on, can we first have a moment of silence for Tony Watt. Scoring 1 goal while averaging 0.51 xG per game is really some feat in being unlucky. Like a shining star, Tony wattburned quickly, hot and bright before burning out.

Anyways, when we look at Expected Goals per 90 minutes over Goals per 90 minutes for Hearts, Callum Paterson is among the best at Hearts for both. But look, some hope! Bjorn Johnsen split time while previously mentioned burning star Tony Watt and Conor Sammon struggled finding their asses with a map. However, Johnsen is now settled in Hearts’ first XI and has knocked in 6 SPFL goals with an xG total of 6.11. Johnsen’s xG per 90 minutes and Goals per 90 minutes both better than Paterson’s, logical given that he’s an actual striker. Hearts fans can be hopeful that Johnsen can help carry the goal scoring load lost with Paterson (Oh, hi US Soccer, lets go ahead and give him a US cap the next possible chance you have).

Shots p 90.png

I have referred to Callum Paterson as a defender a few times so far, but the defending part of his job might be the weak part of his game. It might be fair to compare him to some attacking midfielders Hearts have instead. Jamie Walker has scored 9 goals, though 4 of those are from the penalty spot, and has a rather respectable xG total of 4.66 and a xG per 90 of 0.27. You can’t see it, but I am nodding my head at these numbers. Furthermore, Walker had already averaged more shots per game than Paterson. There may be even more opportunity for Walker to be the focal point of Hearts’ attack with Paterson out that could see the creative attacking midfielder get more goals, shot quality, and shots improve.

If Jerry Seinfeld has taught me anything, it is that things usually even out. Paterson may be gone due to injury for the season, but Hearts have just got back Sam Nicholson from a long injury layoff. Before getting hurt, Nicholson had a very impressive goal haul at 4 goals in about 600 minutes from an xG total of 1.6 and an xG per 90 of 0.23. I have a word of warning though. Nicholson averaged 1.88 shots per game in his limited time thus far this season. It isn’t feasible to think Nicholson will be able to continue a goal pace of 0.58 goals per 90 minutes with those xG and shot numbers. However, like Walker, Nicholson will get more opportunities in the Hearts attack with Paterson out.

Johnsen, Walker, and Nicholson all missed time due to injury or not being picked. Now that each seem to have a spot in the Jambo’s first XI, there are some numbLee Wallace of Rangers & Bjorn Johnsen of Heart of Midlothian during the SPFL Ladbrokes Premiership match between Heart of Midlothian & Rangers at Tynecastle on 30th Novemberers to suggest together they can replicate the missing goals that Callum Paterson’s absence brings. However, pressure is starting to mount on Ian Cathro to get some results. Things will not get easier for the young manager, as his first match in the league after the winter break is at Celtic Park against a Celtic squad running at full tilt trying to break records that the Lisbon Lions set. They also competing for second place with Aberdeen and Rangers squads that are starting to find their form and have spotted each a 6 and 8 point head starts respectively for the race for second. Cathro will need these three performing at maximum capacity if they have any hope in reaching that second spot.

 

Mailbag: Set Pieces, Transfers, xG, and Who in the SPFL Would Make the Best Bachelor?

Welcome to the first in a possible reoccurring series of posts where I take queScreen Shot 2017-01-13 at 8.17.26 PM.pngstions from you, the reader interested in some combination of Scottish football and stats in football. I was unsure if I would actually get any questions when I asked for them on twitter, so to get too many to questions to feature in one post was quite the surprise and very humbling. People disparage both Scottish football and analytics in football, so to see so many people interested in either/or (OR POSSIBLY BOTH?! You people get the gold star) is great. Now, onto your questions!

 

Screen Shot 2017-01-13 at 8.33.44 PM.png

Daisy here is not the first Jags supporter to ask about how Thistle do at set pieces. I am not sure what has caused Partick Thistle fans to be so concerned about set pieces, but questions about how they do might be one of the things I get asked most. Set pieces are what many in the football analytics community seem to believe is the biggest opportunity to gain an advantage over opponents.

Set Piece Chart.png

To see how Thistle compare to the rest of the league, I counted every set piece each club has had so far in the attacking half of the field. Now, I have discussed the limited public data available for Scottish football, so I used the same BBC live tracker data I use for my expected goals. Since that is what I am limited to, I decided to count any goals resulting around a set piece. It may not have gone in directly from the either the free kick or corner kick, but the resulting play led to a goal, so there is a bit of subjectivity to it but it should give us an idea how the Jags compare to the rest of the league.

Set Piece Graph.png

I have got some good news for the Firhill faithful. Your club is above average in the league at set pieces! The league average is around 3.86% conversion rate on corners and free kicks combined. Partick Thistle come in at a whopping 5.53% and they compare favorably to the top 4 clubs in the Premiership. So rest easy Jags supporters, your club is doing ok when it comes to set pieces.

“Are there any players Celtic should sign from the rest of the SPFL?”- Julio Lèon via email

Are there players good enough to play for Celtic on other clubs in the SPFL and be successful? Yes. Are these players better than what they have? I’m not so sure.

spfl-premiership-xg-leaders-8

Who are these players that I think would be successful at Celtic? Well, have I ever mentioned this striker at Ross County? Liam’s his name…I can’t quite recall his last name. Boyle? Boyd? Bourne? Oh, Liam Boyce, that’s right. Boyce has 11 goals from 6.35 xG this season, outperforming his xG. Could he see a regression? Possibly, but he also outperformed his expected goals last year. Typically, great players will outperform their xG.  Looking at his shooting stats, Boyce averages 1.94 shots per 90 minutes this season. Compared to Leigh Griffiths, who is averaging around double the number of shots per 90 minutes. However, Boyce averages an expected goal of 0.19 per shot, which is higher than Griffiths’ 0.13 xG per shot. Given that you would expect a striker playing for Celtic to naturally have more chances on average than a striker for Ross County, as well as his superior xG totals per shot, one would expect Boyce to not only replicate Griffiths’ success with Celtic, but perhaps exceed it as well. Is Boyce better than Moussa Dembele? No, but I do not expect Dembele to be at Celtic past this summer.

I would also point Julio to Callum Patterson, once healthy, as “good enough for Celtic, but not better than what they have”. There is no denying Patterson’s attacking prowess would fit in with Celtic’s high octane attack. Patterson’s 8 goals before he was injured is about what we would expect from his xG of 7.23. For a fullback, Patterson’s weakness is definitely the defensive side of his responsibilities. However, much like one would expect Liam Boyce to get more opportunities lining up for Celtic, you could also hypothesize that Patterson would have less opportunities to be exposed defensively playing for Celtic.

Screen Shot 2017-01-13 at 8.34.08 PM.png

So when I previously mentioned the lack of public data for Scotland? Yeah, something like this is sadly not available. Though as interest in stats in Scottish football has grown, so have those interested in producing content on it. Here relatively new guy on the block “The SPFL Radar” twitter account has created a stats radar for McKay.

McKay Radar.jpg

I would also really quickly like to discuss the “community” surrounding Scottish football stats. One could accuse some of those involved in EPL stats twitter to be over saturated and unwilling to help those just learning about the basics of analytics in football. I have found the absolute opposite of those interested in stats in Scottish football. The most well known accounts are always willing to lend a hand and share info for those looking for it. I would suggest tapping these resources if you have any interest in Scottish football or stats (OR BOTH!)

Screen Shot 2017-01-13 at 8.36.34 PM.png

xg_g-6

I asked Michael what he wanted to use to compare Garner to the rest of the league with and requested expected goals. Garner currently is 12th in the SPFL Premiership in xG total at 5.58, 0.41 xG per 90, and 0.16 xG per shot. With GarnerOuchie.jpgonly 3 goals this season in league play, Garner is underperforming his expected goals. This would suggest he has either been unlucky, not great at finishing, or some combination of both. One would expect his goal total to improve, you know, once he comes back from…

We’ll close out our first mailbag with two questions from emailer Michael M., concluding with perhaps the most important question facing Scottish football today.

How do the SPFL teams rank in terms of expected goals per shot?

screen-shot-2017-01-13-at-8-20-22-pm

Motherwell’s prolific duo of strikers have had another great season for the Steelmen so far. Louis Moult and Scott McDonald have contributed 15 of Motherwell’s league goals this season and are the biggest reason why they have the highest xG per shot in the league at 0.16 xG per shot.

Which SPFL player would make the best Bachelor?

I was unsure if the readers of this blog would know what “The Bachelor” is. After consulting with my “sources”, I was told that The American edition of “The Bachelor was on Channel 5. That is the channel of shit programs.” If you are unfamiliar with it, one male dates 25 females, eliminating a few of them weekly, until he finds his “winner” who he proposes to. The whole idea is, of course, ludicrous.

However, sometimes in a relationship you have to compromise. My spouse puts up with my sports viewing and “sports spreadsheets” (her term for the excel sheet where I house my stats), so I can watch 2 hours of a reality dating show.

Now, most know I support Celtic but I would like to believe I am impartial when it comes to discussing stats in the SPFL. However, in choosing the first star of “The SPFL Premiership Bachelor”, I could only think of one midfielder from Celtic.

 

screen-shot-2017-01-13-at-7-53-38-pm

Now, while I have stats for every SPFL Premiership player, I must admit my knowledge of the marital status of each player and if they would do well on a silly reality dating show is lacking. Do you have a better suggestion for the first star of “The SPFL Bachelor”?

B.U.R.L.E.Y. Tries to Sort Out a Close Relegation Fight in the SPFL Premiership

By now, many of you have realized B.U.R.L.E.Y. hates your team. Celtic fans taking thistleexception to B.U.R.L.E.Y. thinking the Bhoys won’t finish the year undefeated, Rangers fans mad that B.U.R.L.E.Y. is of the opinion they will finish far behind their Glaswegian rivals. Aberdeen fans perplexed by a possible 4th to 5th finish predicted by B.U.R.L.E.Y. The list goes on.

After you get past the top 5 in the SPFL Premiership, things have been incredibly tight. Five points separate last place from sixth place as of writing this. Heck, Ross County who, by both stats and the naked eye, have struggled this season currently sit sixth and only 4 points behind fifth place St. Johnstone, albeit having played one more match than the Saints. Things are equally tight in the fight for a top six place and fighting to avoid relegation. The always tricky promotion/relegation playoff awaits the club that finishes in eleventh and is another thing teams will be trying to avoid.

Since I brought B.U.R.L.E.Y. out to forecast who was going to finish where in the battle for the Premiership title and the European spots, I thought I would do the same thing to get a grip who is going to finish where at the bottom end of the SPFL Premiership. I’ll add the methodology of how I came up with B.U.R.L.E.Y. for anyone who missed the first post (Hi Partick Thistle fans who retweeted B.U.R.L.E.Y.’s prediction for the Hearts match but then also liked me stat-splaining Azeez’s miss against Celtic!) Again, I warn you if the nitty gritty of how I am calculating you will bore you to tears, I would skip ahead of this part. I will give you the results after it.

To come up with B.U.R.L.E.Y., I borrowed Mark Taylor, from the Power of Goals Blog, simulations method with a few tweaks. Using the expected goal data from the SPFL I have gathered, I took the xG average for the league, the average xG both home and away for mcdonaldevery club, and the xG for and against for every club to come up with a calculation for xG for every match up between every team in the SPFL Premiership.

With these expected goal figures, I use Poisson distribution to come up with the probability of every scoreline for every match up in the league. For example, there is a 0.000000000046% chance that Hamilton will beat Partick Thistle 10-7! If we sum the winning score lines for each team in each of these match-ups, we can determine the probabilities of who B.U.R.L.E.Y. thinks will win each match.

Once we have the win probabilities for every match up, we can then run simulations for every game each club has remaining in Excel, as Mark details how to in the link above. We
run each club’s remaining season (up until the league splits into a top half and bottom half, since we won’t know who each club will face in their last 5 matches, because LOLSPFL) 1000 times (…and BOY are their legs tired). We take the point total B.U.R.L.E.Y. suggests will happen most frequently and boom, that’s how many points B.U.R.L.E.Y. thinks your team will end up with.

(End of methodology section if you wanted to skip that part!)

sheet-1-49

You will notice I updated B.U.R.L.E.Y.’s thoughts on the top half of the league. He still sees Celtic as run away favorites for the title, but now thinks Rangers will pull away from the pack and be 7 points ahead for 2nd at the time the table splits. Hearts will be behind Rangers in third, but will be in that position comfortably B.U.R.L.E.Y. foresees. Aberdeen and St. Johnstone will go into the split equal on points, which would likely be a disappointment for the Dons.

Now let us look at the bottom half of the table. B.U.R.L.E.Y. thinks that the fight against relegation will be far from over as we head into May. Motherwell is projected to finish 6th, largely on the attacking prowess of Louis Moult and Scott McDonald. Perhaps a bit of a surprise is B.U.R.L.E.Y. predicting Ross County will be seventh at this junction as well, but B.U.R.L.E.Y. is predicting that the Staggies will be only 6 points off the bottom when 5 matches remain. No word what B.U.R.L.E.Y. thinks will happen to Liam Boyce’s back…you know…SINCE HE’S CARRYING THAT ENTIRE TEAM ON HIS BACK!

Dashboard 1-5.png

As far as the rest of the relegation peloton, B.U.R.L.E.Y suspects there will be all to play for after the split. He predicts only 3 points will separate Partick Thistle, Hamilton, Dundee, Inverness CT, and Kilmarnock. That is 3 points between 7th and top. Partick Thistle and Dundee have showed some improved form, so one would think they would be the favorites from that group to stay up. However, if B.U.R.L.E.Y. is correct in how razor thin the margins will be between safety and relegation, it might not be smart to call any of these clubs safe.

 

Introducing “B.U.R.L.E.Y.”, a Prediction Model for the SPFL and Handicapping the Race for 2nd

We are a few days removed from Hearts decisive 2-0 victory over Rangers. The win put Hearts even with Rangers on points and ahead on goal differential to earn 2nd in the table. Aberdeen had a disappointing League Cup final against Celtic last Sunday, but find muirheadgoalthemselves very much in the fight for 2nd in the SPFL, being 2 points behind Hearts and Rangers while playing 2 games less than both of those clubs. Celtic seems to be coasting to another league title, but all three of these clubs have a good shot to finish 2nd. With a week of hyperbole surrounding all three of these clubs, it seemed like the perfect opportunity to introduce the prediction model I have been working on and use it guess where these clubs fighting for 2nd will be towards the end of the season.

My prediction model has been, most definitely for petty reasons, named B.U.R.L.E.Y. This stands for “footBall Using Reliable anaLytics, Even You!” Was this a giant stretch of an acronym just to make sure it fit Craig Burley’s name? Maybe. But if we can’t be petty about Craig Burley, who can we be petty about?

(If you don’t want to read/don’t care about how I came up with these figures, I’d skip this next part!)

To come up with B.U.R.L.E.Y., I borrowed Mark Taylor, from the Power of Goals Blog, simulations method with a few tweaks. Using the expected goal data from the SPFL I have gathered, I took the xG average for the league, the average xG both home and away for every club, and the xG for and against for every club to come up with a calculation for xG for every match up between every team in the SPFL Premiership.

With these expected goal figures, I use Poisson distribution to come up with the probability of every scoreline for every match up in the league. For example, there is a 0.000000000046% chance that Hamilton will beat Partick Thistle 10-7! If we sum the winning score lines for each team in each of these match-ups, we can determine the probabilities of who B.U.R.L.E.Y. thinks will win each match.

Once we have the win probabilities for every match up, we can then run simulations for every game each club has remaining in Excel, as Mark details how to in the link above. We swansonrun each club’s remaining season (up until the league splits into a top half and bottom half, since we won’t know who each club will face in their last 5 matches, because LOLSPFL) 1000 times (…and BOY are their legs tired). We take the point total B.U.R.L.E.Y. suggests will happen most frequently and boom, that’s how many points B.U.R.L.E.Y. thinks your team will end up with.

(End of methodology section if you wanted to skip that part!)

When it comes to prediction models, compared to others B.U.R.L.E.Y. is simplistic. I haven’t nailed down how to account for things such as players joining clubs in January or, more pertinent at this moment, managerial changes. B.U.R.L.E.Y. will be blissfully ignorant if Pep Guardiola says to himself, “Forget Manchester, I want to see Edinburgh Castle!” and becomes the next Hearts manager. However, since B.U.R.L.E.Y. is using the expected goal data I am collecting throughout the season, I am hoping that B.U.R.L.E.Y. will continue to get more accurate as the xG sample grows larger. In maybe the first time anyone has ever put these letters in the same order, B.U.R.L.E.Y. will get smarter. But enough of that, you are here to see where B.U.R.L.E.Y. says your club will finish.

sheet-1-47

Unsurprisingly, B.U.R.L.E.Y. is predicting Celtic to have clinched the SPFL Premiership title by the time the split comes. B.U.R.L.E.Y. thinks that the race for 2nd will be much tighter though. Going into the split, B.U.R.L.E.Y. is predicting today that Rangers will be 2 points ahead of Hearts and 4 points ahead of Aberdeen going into the last 5 matches. Those last 5 matches will decide who else will be joining Celtic in Europe next season most likely. St. Johnstone won’t feature in the fight for 2nd according to B.U.R.L.E.Y., but will comfortably make the top 6 in 5th. I will continue to update these projections as the season continues.

Since we can use B.U.R.L.E.Y. to predict each club’s position in the table, we can also use him to predict to outcome of individual matches. While I am no professional gambler (archaic, outdated laws here make that illegal), I have been using B.U.R.L.E.Y. for the very legal BBC SPFL prediction game the last two weeks, going 8-3! Here is what B.U.R.L.E.Y. thinks will happen this weekend.

screen-shot-2016-12-01-at-8-47-21-pm

Celtic are unsurprisingly heavy favorites in their match away to Motherwell, with B.U.R.L.E.Y. giving their chance to win at 56.82% (Draw 25.09%, Motherwell 16.29%)

screen-shot-2016-12-01-at-8-47-58-pm

B.U.R.L.E.Y. gives an edge in the fight for 2nd to Rangers this weekend. Unsurprisingly, teams typically fare better at home than on the road, so when two relatively equal teams meet, B.U.R.L.E.Y. usually thinks that the home team will prevail, giving Rangers a 53.73% chance to win (Draw 28.95%, Aberdeen 17.32%).

screen-shot-2016-12-01-at-8-51-11-pm

As stated earlier, B.U.R.L.E.Y. currently doesn’t take into effect any coaching changes, so despite Robbie Nielson moving to Milton Keynes, B.U.R.L.E.Y. still expects Hearts to win against Ross County. He gives the Jambos a 43.76% chance to win (draw 37.47%, Ross County 18.77%).

screen-shot-2016-12-01-at-9-00-53-pm

Tommy Wright has built a solid foundation in Perth at St. Johnstone, where he has given the Saints an underdog fighting mentality (as well as seemingly able to find good keepers out of Canada). On Twitter, a Saints supporter told me the club was on their way to becoming the “establishment diddy team”. Here B.U.R.L.E.Y. thinks the “establishment diddy team” will be able to put aside Inverness Caley Thistle, giving the Saints a 49.24% chance of winning (Draw 29.56%, Inverness Caley Thistle 21.20%). Even B.U.R.L.E.Y. knows that Richie Foran has the best beard in the league though.

 

 

 

 

Conversion Rates in the SPFL Championship: The Finishing Fairy Can’t Find Dunfermline on a Map

Gathering stats for the SPFL Championship for the first time this year has been very interesting. Sometimes the stats do not make sense, sometimes it is a hassle to make sure cummingsshinnethe data is correct, and sometimes the data is not even there. Yet, it has been interesting following the league, especially since some Scottish media outlets have reduced the coverage of the league.

Seeing as I compared the conversion rates for the Premiership in Scotland, I thought it would be interesting as well to compare them for the Championship as well. Here we can see some teams who has seemed to make lady luck angry, while others seem to be reaping the benefits of an extended visit from the finishing fairy.

championship-conversion-rate-table

As a reminder, Mark from Every Team Needs a Ron says when discussing conversion rates, “As a rough guide, anything within +4% to -4% conversion difference seems like noise that is hard to discern meaning from.” We see first place Hibernian right below that +4% conversion rate difference. With a positive expected goal difference and a relatively sustainable conversion rate difference, Hibs can be expected to continue to compete for the title.

cship-conversion-rate-graph

In recent weeks, Dundee United has emerged in the table as Hibs main competitor for the top spot in the Championship. However, looking at both their conversion rate and expected goal differential, we see them with numbers that might not be able to sustain their undefeated form in the last five matches. United have a negative expected goal andreudifferential and a very unsustainable +9% difference in conversion rate.

In those last five matches, the Tangerines have had a lower xG than their opponent in 3 of those matches. Their overall conversion rate is at 20%, a number they usually would not be able to continue. We would expect Dundee United to regress throughout the season, but they certainly could be in line for a promotion play-off spot.

Now to the other end of the table. Things are looking hopeless for St. Mirren right now. The Scottish League Cup winners of 2013 seem destined for relegation to League 1. The Buddies have an xG difference of over -5 and a conversion rate difference of over -10%! Wow! St. Mirren can expect that to improve a bit, but dreadful xG for and against numbers means you shouldn’t expect much of a change of fortune for the team from Paisley.

Above St. Mirren in the table is Dunfermline. Ayr, Dumbarton, and Dunfermline were most’s picks for relegation from the Championship this year and those 3 clubs are definitely in danger of that. However, the Pars are the only one with the distinction of having a positive expected goals difference and a negative conversion rate difference. DUNFERMLINE V ST MIRREN
LADBROKES CHAMPIONSHIP
EAST END PARK
 DUNFERMLINEDunfermline has been able to create some chances, as they have the 3rd highest xG total in the Championship and have scored 15, the same as Queen of the South.

However, they have also conceded the 4th highest xG against total and the 2nd most goals in the league. The finishing for the Pars hasn’t been relegation level, but the defense and goalkeeping certainly has.

If Hibs can continue these numbers going forward, they should win the league comfortably. However, from the promotion playoff spots to  relegation seem to be up for grabs for numerous teams. If the first three months of the season are any indication, it should be a wild ride.

Conversion Rates in the SPFL Premiership

The football analytics community most recent civil war with itself has been on the moussadembele-cropped_1mhegto7ovgp01i1t0ul0cjkhhsubject on finishing skill. Whether or whether not a player can improve, do something, be better at finishing than fellow pros has been a hot topic. The idea of expected goals heavily relies on there not being a discernible difference in finishing ability amongst teams in a league, since all chances are judged the same in a model no matter who takes it.

One of the things to like about expected goals is that it does not treat every chance the same, such as a a stat like conversion rates (goals/shots taken) does. Teams may have a higher and lower than average conversion rate at points of the season, but usually teams conversion rates regress to the mean over the season.

Of course, Scotland is a bit different than most leagues. Friend of the blog Seth Dobson compiled 10 years worth of shots in the SPFL and found that Celtic consistently had a higher conversion rate than the league average. However, even Celtic has a mean that it regresses to. Since their 5-1 Glasgow Derby victory, Celtic have not lost and only drawn once in the league. Yet despite their impressive form, Celtic had a conversion rate of 25% before the derby and now have a conversion rate of below 19%. Regression happens to everyone.

conversion-rate-table

After being inspired (well, taking his idea and applying it to the SPFL) by the EPL stats blog “Every Team Needs a Ron” (as in Harry Potter), I decided to compare each club’s conversion rate and the conversion rate of their opponent. In his blog, Mark from ETNAR says “As a rough guide, anything within +4% to -4% conversion difference seems like noise that is hard to discern meaning from.” Keeping this in mind, we see most of the league is within that range, with two outliers of Celtic and Ross County.

xG Differential_Conversion Rate Differential.png

I thought it would be interesting to compare these conversion rate differences to each club’s expected goal difference, as well as their goal differential. A club’s spot on the graph is based on their expected goals and conversion differences, as well as the size of their circle is based on their goal difference.

Celtic, Aberdeen and Hearts all have positive conversion rates and expected goal differences. It is no surprise that these clubs are in first, second, and fourth in the league.

Dundee and Ross County look marooned on the relegation islands with their negative conversion rate and expected goal difference. County particularly look in trouble with their conversion rate falling outside of that + or – 4% conversion rate difference that we could chalk up to noise. Blog favorite Liam Boyce has missed some time with a knee injury, so his return will be a welcome site for County, but Boyce will need some help to pull County out of a relegation fight.

The only team that has the dubious distinction of a positive expected goal differential and a negative conversion rate is Rangers. They currently sit in 5th place in the league, well miller.jpgbelow the standards their fans demand. Much of the blame for the Bear’s place in the table has fallen on the shoulders of their defense. The back line has been a middle of the league defense, with the 4th best expected goals conceded total and the 6th best conversion rate conceded.

While the defense has been ok but not spectacular, I might suggest much of Rangers issues have stemmed from their attack. With the fourth best expected goals for in the league, Rangers should have a conversion rate a lot better than just above 8%. Have Rangers been unlucky or are their attackers not good enough? I would say a little of both.

Harry Forrester has the highest xG total at the club, and is 8th in the entire league, but only has 4 shots on target all year. Rangers leading scorer is 36 year old Kenny Miller. You would expect some more bounces to go Rangers way as the season goes on, but even if their finishing moves up towards the mean, it likely still will not be enough to be a title contender. It could see Rangers challenge the likes of Aberdeen and Hearts for 2nd though.

Finally, looking at the top left quadrant of the graph we see Kilmarnock and St. Johnstone in what I like to call the “lucky quadrant”. St. Johnstone are on the edge of the “lucky quadrant”, with only a -0.53 xG difference and conversion rate difference of 3.14. Much of this is thanks to the wonderful play by their young keeper Zander Clark who is the owner of a 88% save rate. Tommy Wright has managed to buck the stats for his entire tenure with the Saints and this year is no different, though some of the Perth clubs stats are still good this year.

Meanwhile Kilmarnock were many’s pick for relegation this year and their xG total is of a_91368317_coulebaly_cel_sns club fighting relegation. Yet, Killie currently sit in 7th place in the table, seven points safe. Killie’s place in the table has much to do with their conversion rate of over 16%, the 2nd highest in the league behind Celtic. Souleymane Coulibaly is good, but he’s not that good. Expect Kilmarnock’s finishing to regress and likely will coincide fall down the table, but with a 7 point cushion, Killie could prove the pundits wrong again and avoid relegation.

 

xG Leaders in the SPFL: What is it Going to Take to Get Liam Boyce and Jason Cummings a High Profile Move?

Hi Hibs and Ross County fans. Let us get this out of the way right now. I hope Jason liam-boyceCummings and Liam Boyce stay in Scotland. I really enjoy watching them play, think they are great players, and think it is great they are at both of your respective clubs.

Both are leading the SPFL Premiership and Championship in goals and expected goals. When looking at both players’ impressive stats thus far in the season, I started thinking about how long they would be in Edinburgh and Dingwall. More over, I was a bit perplexed they were still at these clubs at largely successful campaigns last year. This made me recall an interaction I had on twitter with a Birmingham City fan.

761205017709862912

Rightly or wrongly, those down south think there is a risk in purchasing talent from the SPFL. While there are plenty of examples of players who have succeeded in England, people (and likely clubs) will think of Andrew Shinnie (who has been good at Hibs this year), David Goodwillie, Aidan McGeady, and others who did not fare as well down south as they did in Scotland.

Expected Goals per 90 minutes over Goals per 90 minutes in the SPFL Premiership

xg-p-90_g-p-90

Strikers like Liam Boyce and Jason Cummings are clearly among the best strikers in their respective divisions. Along with their hot goal scoring form to start the season from both,
both have the best-expected goal totals in the SPFL Premiership and Championship respectively. As we have discussed, expected goals is a better indicator of future success than goals are. Both have taken the highest number of shots and shots on target in their division. Both seem to have stats to suggest they could continue their form at a higher level.

Expected Goals per Shot over Goals per Shot in the SPFL Premiership

xg-p-shot_g-2

Now, this is not to say that there haven’t been bids for either Boyce or Cummings. Hibs have reportedly turned down a seven-figure bid for Cummings. Clearly any bid for either Jason Cummings.jpghas not met either Hibs or Ross County’s valuation of their star striker. Both have recently signed contract extensions, neither club needs to sell anytime soon (and one would hope County wouldn’t continue the recent trend of giving players other clubs want for free).

Hibernian and Ross County have lofty goals this season. Hibs is looking (and honestly needs) to win promotion back to the SPFL Premiership. Ross County is looking to venture into uncharted territory with back to back top 6 SPFL Premiership finishes. Both are the holders of the Scottish Cup and League Cup respectively. Much of their success this season will come down to the performance and health of Cummings and Boyce.

Expected Goals per 90 minutes over Goals per 90 minutes in the SPFL Championship

xg-p-90_g-p-90-cship

Yet, if someone comes in with a bid for either that is above Hibs or County’s valuation, could we fault them for accepting? Hibs, one of the biggest clubs in Scotland, are now in their second season in Scotland’s second division with second division revenues. County, despite a top 6 finish and League Cup win, is one of the smallest clubs in the top flight and has only been in the top flight since 2012. Financially, a £2-4 million bid would be tough for either club to turn down.

Expected Goals per Shot over Goals per Shot SPFL Championship

xg-p-shot_g-p-90-cship-2

I do think that both clubs would be wise to not let past failures from players hold them hostage when or if they sell Liam Boyce and Jason Cummings. These players have shown consistency in numerous metrics over multiple seasons now and have the tools to succeed anywhere they go.

Perhaps, Scottish clubs should be looking to sell elsewhere besides the English Championship and EPL. However, if that’s where Boyce and Cummings end up, can they be the players to change the minds of those in England about the quality of player in Scotland? That is, quite literally, the million pound question.

(Peter Lawwell, if you’re reading this, both would be great additions to Celtic!)

Follow The Backpass Rule on Twitter at @TheBackpassRule

A Quick Delve into Keeper Stats in the SPFL Premiership

craiggordonAs the 2016-2017 SPFL Premiership Campaign is mere days away from getting underway, it seemed like an opportune time to discuss new stuff I am hoping to accomplish this year. This year I am making an attempt to track stats on keepers in the SPFL. Thus far, most of the available data available to play with for the SPFL deals with the attacking side of the ball. Expected Goals, shot ratios, and the like give us insights on the teams and players who have the best opportunities to put the ball in the net, so why not reverse that data and look at those trying to stop the ball going into the net. Stats for keepers have been developed by a few in the “analytics community”, such as American Soccer Analysis  and Deep xG. It has been very interesting to see these talented analysts and writers develop these stats for our friends in between the sticks. I started this blog after not finding any stats for the SPFL that I was looking for, and these keeper stats are along the same lines. I used the data I gathered for last season’s SPFL Premiership campaign and applied it to American Soccer Analysis and Deep xG’s ideas.

First, I use inspiration from American Soccer Analysis, the very well done blog on MLS and analytics. Is it a little depressing that Major League Soccer has more data and stats to play with than the SPFL? Yes, yes it is. But regardless, you should check out the fine work they do there and keep tabs on the various former SPFL players in MLS now. Below are the minutes, Shots on Goal Faced, Saves, Goals Against, Expected Goals Against, Goals Against minus Expected Goals Against, and Goals Against minus Expected Goals Against per 90 minutes for every keeper who appeared in at least 1,000 minutes last season.

2015-2016 SPFL Keeper Stats-3

partick-thistle-mascot-tomas-cerny_3320439.jpgLast season, we see that Celtic’s Craig Gordon had over two less goals conceded than his expected goals against total, putting him at the top of the league. The general consensus (or “eye test” if you will) was that Gordon, Danny Ward, Neil Alexander, and Scott Bain were the best keepers in the league last year, and these numbers back that up, with those four keepers amongst the top five in these metrics. The surprising inclusion amongst the top five is Partick Thistle’s Tomas Cerny. Cerny had some injury issues last year, but despite appearing in about 700 less minutes, faced the same amount shots that Craig Gordon did and made only two less saves. With the Partick Thistle defense much more porous than Celtic, Cerny’s GA-xGA total of 0.98 is very impressive and the second best total in the Premiership last season.

The next stats we are applying to the SPFL keepers are taken and developed by the great blog Deep xG, who does wonderful analytical work on the EPL. These stats include minutes, shots on goals, goals against, saves, expected saves, which uses the xG model to predict how many saves each keeper should make, how many saves they made above expected, and a “keeper rating” which is saves over expected save as a percent. Again all keepers listed appeared in at least 1,000 minutes last year.

2015-2016 SPFL Keeper Stats 2

scottfoxWhen we look at these stats, we again see the likes of Craig Gordon, Danny Ward, Neil Alexander amongst the top keepers in the SPFL, as well as Tomas Cerny again towards the top, further confirming the Partick Thistle’s great season for the Jags last year. All four of these keepers made more saves than they “were expected to make”, with a positive Saves Above Expected total. They also have a Keeper Rating of over 100 (with a keeper rating of 100 meaning a keeper made the same amount of saves they were expected to). Towards the other end of the table, we see Ross County’s Scott Fox and Gary Woods. Last year, this blog was very impressed with Ross County’s season. For a club of their size, Ross County had some impressive attack numbers and their defense conceded the 3rd best expected goal against total in the league, yet their keepers let them down for much of the year. Fox was the Staggies first choice keeper, although he had some injury issues throughout the year, yet was a -9.22 for saves above expected and a keeper rating of 87.58. His deputy when he was out injured was Gary Woods and Woods didn’t fare much better, with a saves above expected of -9.58 and a keeper rating of 76.96. County manager Jim McIntyre saw the need to improve the keeper position this offseason and brought in Aaron McCarey from Wolverhampton. If County are to continue their unprecedented success from last year, they will need improvement from their keepers this year.

I am looking to track these stats throughout the upcoming season and provide weekly updates to how the keepers of the SPFL Premiership fare.

Using Expected Goals and Simulations to Guess How the SPFL’s Top Strikers Will Perform

stone cold lock of the weekAs we move through the summer and closer to the beginning of a new SPFL Premiership season, it is time to look at what we think will happen in the 2016-2017 season. While the SPFL has less data available to play with than our neighbors down south (as I swoon looking at passing data that’s available in the EPL) we can still use what is available to us, including the shot location data that Opta and the BBC provide to come up with expected goal numbers. I have done a lot with expected goals for the 2015-2016 season, but in case you are new here, expected goals (shortened to xG) measures more or less how many goals a team and/or player “should” score based on the type and location of a shot.

While there has been a lot of debate about the validity and place for analytics in football, the expected goals metric has gained more public acceptance. Much of the reason of the metric’s acceptance is due to the fact that it works! In 2015-2016, the six teams with the best xG difference (expected goals for minus expected goals against) finished in the top 6 of the SPFL Premiership. Furthermore, there has been plenty of research done by the likes of Ben Pugsley, Opta, and 11tegen11  on how expected goals is a much more repeatable metric than other stats like scoring rate, shooting accuracy, shots, and even goals from the previous year! The smart guys who wrote these articles were able to find after looking at data for lots and lots of strikers across Europe that there is nothing to suggest a player who has a high scoring rate, shooting accuracy, high number of shots per game, or scores a lot of goals will be able to repeat those numbers the next year. Basically, they found that goal scoring is often times a random occurrence that many times has a lot to do with luck. However, they did find that a striker who has good expected goal numbers one year is more likely to be able have good expected goal numbers again the next year. We can roughly expect a striker to have similar expected goals numbers as the previous year. And since I have the expected goal totals for every player who played in the SPFL Premiership last season, we can use that data to predict how many goals a player will score. There are of course numerous things that could throw off these predictions, such as injury, the arrival of new managers who don’t fancy a player, the arrival of a new night club that may distract the player, etc, yet I am deciding to ignore all those very real instances that might skew my predictions and declare all of the following information MY 2016-2017 SPFL PREMIERSHIP STONE COLD LOCKS OF THE YEAR!!!!!!(LOUD LOCK NOISE SOUND EFFECT HERE!) 

In order to predict just what the Leigh Griffiths’ and Adam Rooney’s of the world might do this season, I had to run some simulations. I had plans of using the summer to learn the powerful computing, graphing, and data management program “R”, which could handle this function with ease. I was able to get some basics down, but not as far as I wanted to be. Luckily, Mark Taylor from the The Power of Goals blog published instructions on how to use expected goal data to run simulations in order to make our best guesses definitely and authoritatively quantify how many non-penalty goals your favorite SPFL striker will score this year using Excel. The simulations use the xG data to give the percentage chance of each non-penalty goal total a striker could get, with the highest percentage being “the most likely” goal total. So with my trusty expected goals data and the help of Mark and Excel, let’s take a look at what we can expect this season from the 10 strikers who led the SPFL in expected goal total last season.

griffiths-pricelessLeigh Griffiths – Celtic: 28 Non-Penalty Goals, 20.39 xG last season

2016-2017 Non-Penalty Goal Prediction: 21

Sheet 1-31

Last year’s player of the year was able to knock in 28 non-penalty goals last season and had far and away the highest expected goal total in the league. Much of that expected goal success had to do with the sheer quantity of shots Griffiths got due to Celtic’s domination of the ball and the higher number of opportunities they created. While easily the best striker in the league last year, the xG numbers suggest Griffiths might have got a bit fortunate scoring 28 goals and we might expect him to still net a very respectable 21 goals next year.

hemmingsKane Hemmings –  Dundee FC: 19 Non-Penalty Goals, 12.26 xG

2016-2017 Non-Penalty Goal Prediction: 10

Kane Hemmings xG Simulation

While two Dundee strikers made the PFA Player of the Year team last season, in this author’s humble opinion Kane Hemmings was the only deserving Dee on the team of the year. Hemmings was second in the league in both goals and expected goals in the league last season, keeping Dundee in the hunt for a top 6 spot much of the season. However, the simulations don’t think Hemmings xG output will allow him to hit the heights he did last season, putting him in the 10-13 non-penalty goal range.


moultLouis Moult – Motherwell
: 12 Non-Penalty Goals, 11.02 xG

2016-2017 Non-Penalty Goal Prediction: 8

Sheet 1-25

Louis Moult quickly became a blog favorite, though retweeting a blog praising his play last season may have a bit to do with his favored status. That favorite status makes it that much harder to predict that Moult will see his goals likely decrease this year, though it won’t be a drastic decrease, with the simulations predicting a likely goal total between 8-10 non-penalty goals this year. Moult is able to get into decent scoring positions, averaging 0.18 xG per shot, but he did not get the same number of shots as a Leigh Griffiths or Kane Hemmings did last season. Don’t worry Louis, The Backpass Rule thinks you can be the outlier this year and the 0.1% chance that you’ll score 19 goals this year will happen!!!

rooneyAdam Rooney – Aberdeen: 14 Non-Penalty Goals, 10.41 xG

2016-2017 Non-Penalty Goal Prediction: 12

Sheet 1-23

Adam Rooney probably deserved a spot on the Team of the Year last season, scoring 20 goals (though 6 were from the penalty spot) and helping Aberdeen keep pace with Celtic for much of the year before falling short. Rooney could have seen more goals but missed a few weeks due to injury, but when he was healthy Rooney feasted on high probability chances, netting a 0.19 xG per shot he took, the highest amongst the xG Top 10 in the SPFL. His ability to get in high probably shot locations has led the simulations to believe that Rooney can continue his good form from last season, with his most likely output being around 10-13 non-penalty goals.

magennis.jpgJosh Magennis – Kilmarnock: 10 Non-Penalty Goals, 9.94 xG

2016-2017 Non-Penalty Goal Prediction: 10

Sheet 1-29

 

With Killie bringing 11 new players this off-season, I am afraid this prediction might miss due to the whole “manager doesn’t fancy player anymore” reason. But there’s no time for uncertainty, these are STONE COLD LOCK predictions. Joking aside, Magennis seems to be slagged a lot, but he had a very solid season for a very frustrating Kilmarnock club scoring 10 goals. Magennis scored a decent and had good xG numbers for a club that struggled as much as Kilmarnock last year, though his 2.27 shots per 90 minutes, which was 2nd highest in the league, helped contribute to his high xG numbers. The simulations think that if he gets playing time, Magennis will be able to replicate his goal scoring stats from the previous season in the neighborhood of 10 goals.

boyce.jpgLiam Boyce – Ross County: 12 Non-Penalty Goals, 9.21 xG

2016-2017 Non-Penalty Goal Prediction: 10

Sheet 1-24

Ross County made club history last season finishing in the top 6 for the first time in club history, as well as winning the League Cup also for the first time. Much of the club’s unprecedented success was due to Liam Boyce. Boyce had impressive numbers at every level he has played at, and has continued that good form last season with impressive goal totals, xG per shot, xG per game, and total xG numbers last season. Those solid numbers should continue this season for Boyce as he and the Staggies look to build on their unprecedented success from last year.

juanamaJuanama – Heart of Midlothian: 9 Non-Penalty Goals, 7.95 xG

2016-2017 Non-Penalty Goal Prediction: 9

Sheet 1-28

Five Hearts players had at least 5 or more goals last year, with Juanama leading the club with 9 non-penalty goals. Yet, I can’t help but think that if the Edinburgh club had a more prolific striker, they could had a better chance of leaping over Aberdeen for the 2nd spot. Juanama is certainly a decent striker and 9 goals is a respectable total, though he has one of the lower xG per 90 minutes last year. This year, Hearts have brought in Connor Sammon and Robbie Muirhead to add to their striker depth. Perhaps one of the aforementioned new additions can form a partnership with Juanama that sees them both flourish, but the simulation’s best guess is that Juanama largely repeats his performance from last year with 9 goals this year.

doolanKris Doolan – Partick Thistle : 14 Non-Penalty Goals, 7.5 xG

2016-2017 Non-Penalty Goal Prediction: 7

Sheet 1-27

Partick Thitsle had a famous name on their roster last year, with Mathias Pogba, brother of French international Paul, at the club. Yet, it was not the famous name who led the club in scoring but rather longtime Jag Kris Doolan who led the club in goals. Doolan’s 14 goals always seemed to come at crucial times to keep Thistle from getting caught up in the relegation fight as Dundee United continually shot themselves in the foot before going down. However, the xG numbers suggest that the 29 year old Doolan and Thistle were a bit fortunate to see him get that many goals, with the simulations suggesting that if he repeats his xG output from a year ago he will likely net 7 goals.

macleanSteven MacLean – St. Johnstone: 12 Non-Penalty Goals, 7.49 xG

2016-2017 Non-Penalty Goal Prediction: 6

Sheet 1-30

It is looking like Steven MacLean will be in First Class of the good ship Regression next season. St. Johnstone have been a club that have “gone against the stats”, though last year they did have the 4th best xG difference in the SPFL and turned that into a 4th place finish the league. Yet, even without looking at MacLean’s numbers in depth, one could argue that you probably could not expect another 12 goal season from the 33 year old and the xG numbers back this up, guessing MacLean will score 6 goals for the season.

mcdonaldScott McDonald – Motherwell: 10 Non-Penalty Goals, 7.37 xG

2016-2017 Non-Penalty Goal Prediction: 10

Sheet 1-26

It seemed like Scott McDonald was on his way to join Neil Lennon at Hibernian, but McDonald controversially re-signed with Motherwell at the beginning of July. Lennon was a little irritated with this, but Hibs loss is Motherwell’s gain, as the Australian striker was a solid contributor to Motherwell’s top 6 finish last season. Rounding out the top 10 of xG total in the league last season and forming a solid partnership with Louis Moult, the simulations largely expect McDonald to be able to replicate his goal scoring form of 10 goals from last year.

So there you have it, the absolute, definitely going to happen predictions for how some of the best strikers in the SPFL season will go. Definitely no chance of something else happening, either better or worse, for any of them. We probably shouldn’t even bother playing the season since I’ve predicted all of this.

 

 

Celtic Conversion Rate by Location

So, anything interesting happen in Scottish football since I have last posted? A quiet end to the season, right? No? Oh. Anyways, we have reached the summer break for Scottish football. It will be a short break for the clubs heading to Europe, as all will have to run the gauntlet of qualifier matches to make the group stages of both the Champions League and the Europa League. I have spent the off season looking at the shot data I have gathered this past season from every SPFL Premiership match from the BBC (who utilize Opta to provide this data) in hopes of improving the Expected Goal model I use for next season. It is times like these that I am thankful there are only 12 clubs in the top flight of the SPFL, as categorizing every shot all 12 clubs had this past season is long work, yet I am nearly done doing so. In the meantime, since I am a Celtic supporter and they are first on the magical spreadsheet of stats I decided to take a closer look at their shots by location and goals scored from them this season.

With Celtic adding Brendan Rodgers as their new manager, discussion about the tenure of Ronny Deila in that position was soon replaced by looking towards the future for Celtic. However, a joke I made about Deila’s time at Celtic had me thinking about shot locations and conversion rates during the Norwegian’s time in charge.

734046872722132993

Deila’s time at Celtic has been has seen mixed reviews, though I suspect as time marches on, we will see his time remembered more fondly, such as Gordon Strachan’s time at Celtic has been. I was particularly interested in where Deila’s Celtic squad had most of their shots this past season and their conversion rate at those spots. If you aren’t familiar with BBC match reports for the SPFL, the Beeb shares how the shot was taken (header, right footed, etc.) and a general location of the shot (right side of the six yard box, left side of the box, outside the box, etc.).

Celtic Shots 2015-2016

Celtic Shots by Location Foot

Celtic Shot Locations Head

Celtic Shots on Target 2015-2016.jpeg

Above we see both the percentage of goals scored versus the number of shots taken and shots on target in each description the BBC gives us, as well as a chart of total number of shots, shots on target, goals, conversion rate from shots, conversion rate from shots on target, and shots on target percentage. Looking at the numbers, my joke about Deila and his teams relying a lot on long distance shots seems to be true. Celtic often times were either stopped from getting better quality shots or settled for those long distance shots. Last year, Celtic took double the amount of shots “outside the box” as the did from the “center of the box”, yet scored more than double the amount of goals from the “center of the box”. It is no surprise that the highest conversion rates were seen from the six yard box, scoring 46% shots from the left side of the six yard box by foot (though scoring none from that same spot via header), nearly 33% from “very close range” from shots by foot and 67% from headers, and 20% from the right side of the six yard box by foot and 14% by header. Much of Celtic’s success came when Leigh Griffiths was able to get into the box. Celtic would often have trouble when Griffiths was unable to get service in the box and he and his teammates would let the shots from distancefly .

It has been accepted that a team getting shots in the “danger zone” in the box is a key to success in football. Most of Celtic’s success and goals this year went along with this and you would think that Brendan Rodgers will try to increase the number of opportunities in that area while limiting shots outside the box and further. Of course it would be very interesting to see where the previous pass that led to these shot came from, but as they say, beggars can’t be choosers. I had plans this offseason of keeping similar stats that for the Irish League that I did for the SPFL, but I soon realized that not only is there no shot location data like this for the Irish League but they do not even have reliable numbers for shot attempts, so I will count my blessings for what the SPFL have.