Thanks for joining me on the latest part of my football data-science journey! I hope you found my previous two blogs useful, where we investigated Goals and xGoals (Expected Goals) over 15-minute time segments of a Premier League match. This time around, I’m going to look at corners over the course of the 2016/17 season so far, with a view to helping make your corner trading this weekend more profitable.
My hope is that you’ll be able to apply what we’ve found through our rigorous StrataData analyses to your own trading, even if you’re only entering the corners market for the first time. With this in mind I’m going to be looking both at the number of corners teams have gained and conceded so far this season, alongside each team’s expected corner (xCorner) metrics. This will further help us understand how differently a team is performing versus expectation. In addition to this, we’ll also be able to find trends over the team’s last 10 games (either at home or away) and once again I’m going to pick three games that will be broadcasted live this weekend to look at in more detail.
As mentioned above, I’ll be using xCorners to identify when teams are over-performing or under-performing versus expectation. You may have seen us use expected goals in our previous blogs, and in the same light xCorners simply indicates the number of corners a team is expected to have based on the number of key entries they make.
If you need a refresher on what key entries are and how we collect them at Stratagem, I’d strongly encourage you to read one of my colleague Rich’s earlier blogs. However, to give a brief explanation: a key entry is an instance where the attacking team achieves possession of the ball in the last 18 yards of the pitch where they are attacking. This is further classified by the location on the pitch where the team achieved possession (either “right”, “left” or “box”) and also the method of the key entry (either “pass”, “run”, “turnover” or “shot”).
We have calculated that the average team in our database wins a corner once in about every 5 key entries (0.22 corners per key entry). We then used this number to multiply against the amount of key entries a team makes in a match. So the more key entries a team has, the greater their number of xCorners would be, whereas the fewer key entries a team has, the lower their xCorners would be. For instance, in Tottenham’s last home match against Chelsea, they had 19 key entries. To calculate their xCorners, we can multiply 19 by 0.22 to give us an xCorner value of 4.18. In actuality Tottenham only had one corner in that match, so greatly underperformed against expectation!
So, let’s jump straight into the first fixture and put this simple way of calculating xCorners into practice:
Tottenham Hotspur vs. West Bromwich Albion (Saturday 12.30pm on Sky Sports 1)
Staying with Tottenham, they will surely be looking forward to hosting West Brom this coming Saturday considering their good form of late. Indeed, investigating Tottenham’s corner data unearths some rather interesting patterns.
Looking across Tottenham’s fixtures this season (top-left graphic), the number of corners they have won seems to fluctuate greatly, with an average of 6.7 and standard deviation (SD) of ±3.95. Although they are winning more than the league average of 5.49 corners, Tottenham’s fluctuation is highlighted when compared to the league’s SD of ±0.57. Diving a bit deeper into the data, we can see that against the somewhat “smaller” teams when they will be heavy favourites in the 1X2 market, they tend to rack up the corners. The two peaks in the top-left graphic show 14 corners against Sunderland (a 1-0 home win in fixture number 3) and 11 corners against Burnley (a 2-1 home win in fixture number 9), whereas against Manchester City they only had 4 (a 2-0 home win in fixture number 4) and against Chelsea they only had 1 (a 2-0 home win in fixture number 10).
This trend is further highlighted when we look at their xCorners, as against Sunderland when they racked up those 14 corners they should have actually only had ~9 corners. These came from their 43 key entries during that game, whereas against Chelsea (fixture number 10), they only managed 19 key entries, giving them 4 xCorners (although still 3 more than what they actually won!) So whilst they still won these fixtures, it maybe sensible to look at their opposition before deciding whether to place a trade over or under on the 2-way market!
West Brom on the other hand have a smaller SD of ±2.53, conceding an average of 5.2 corners so far away from home this season. Whilst this is better than what they are expected to be conceding (6.18 ±1.71), their fixture-by-fixture pattern appears more steady than Tottenham’s, with only one stand out ‘blip’ against Arsenal in fixture 9 (top right graphic) where they allowed Arsenal to rack up 45 key entries generating ~10 xCorners compared to the 6 Arsenal had.
Having investigated both teams’ corner patterns this season, a two-way corner line of 11.5 initially seems to be fitting for this fixture. However, given Tottenham’s propensity to over-perform at home when they are the favourites, and West Brom’s relatively consistent behaviour, a line of 12.5 is more reasonable. A sensible trade on this fixture’s corner market would be to back under if the market line is greater than 1 above this projection (under 14), or over if the line is greater than 1 below this projection (over 11).
Leicester City vs. Chelsea (Saturday 5.30pm on BT Sport 1)
Champions Leicester City host Antonio Conte’s Chelsea at the King Power Stadium on Saturday, with the visitors looking to maintain their strong lead at the summit of the Premier League. Leicester’s rate of winning corners so far this season has been steady (top-left graphic), averaging 4.5 corners at home with a SD of ±2.12. Their average xCorners of 4.64 (±1.15) highlights this impressive level of consistency so far this season. We can expect the home side to maintain their status quo against Chelsea on Saturday, as Chelsea so far this season are conceding 4.4 (±3.31) corners away from home, with an xCorner average of 4.53 (±1.69).
Chelsea’s ability to win corners away (bottom-left graphic) tells us a similar story, as they are averaging 4.6 (±2.91) corners, and 6.49 xCorners (±1.58) so far this season. However, they may look to take advantage of Leicester’s inability to keep teams out of their 18 yard area, as Leicester are conceding slightly more corners at home on average than would be expected of them (5.9 versus 5.43).
This fixture on the whole looks steadier than West Brom’s visit to White Hart Lane, with both teams boasting a relatively small SD in both corners and xCorners. A two-way corner line of 10 would be fitting, even after considering both teams’ xCorners, so a trade under anything above this wouldn’t be unwise.
Manchester United vs. Liverpool (Sunday 4.00pm on Sky Sports 1)
Sunday brings to us one of football’s biggest rivalries, when Jose Mourinho’s Manchester United side host Jürgen Klopp’s Liverpool. United’s ability to win corners at home (top-left graphic) appears unsteady – caused partly by an enormous peak at fixture 5. They are averaging 8.5 (±5.02) corners and 7.33 (±1.83) xCorners at home this season. However, what strikes me here is the difference in consistency between actual corners and xCorners (the SDs for actual and xCorners is ±1.83 vs. ±5.02).
Whilst United have looked to push on in matches, it’s not always the case that this translates to the number of corners they win. For instance, in fixture number 4 against Stoke City they won 13 corners, but were expected to win only 7 given the 32 key entries they had. Despite this increased rate of corner production, they only managed a 1-1 draw. Similar is their story against Burnley in the very next fixture (number 5), where they amassed a monumental 19 corners, but were only expected to win ~10 corners (9.68) based on the 44 key entries they had. Again they only managed a draw, 0-0 this time, pointing to significant inefficiency in their attacking play. Indeed, when United have won fewer corners than expected, they’ve actually won the game. Against Tottenham Hotspur (fixture number 8), they won only 3 corners but were expected to win 6 (rounding up from 5.72) from their 26 key entries and managed to win 1-0. This is true again in their next fixture against Sunderland (fixture number 9), where they won 5 corners but were expected to win 8 (from their 35 key entries) and won the game 3-1.
United will be looking to break into Liverpool’s 18 yard area further on Sunday, despite Liverpool showing a gritty ability to not concede many corners away from home (usually due to being the favourites). Liverpool average conceding only 3.10 (±1.85) corners so far this season, and are only expected to concede 3.67 (±1.27) corners. Irrespective of which United side turn up, it seems as though Liverpool will put up a mean defence of their 18 yard line, largely through their high pressing game and forcing turnovers higher up the pitch.
Whilst Liverpool maybe good at keeping teams on their favoured side of the line, they also show an ability to win corners at the other end. Liverpool are averaging 6.4 (±3.24) corners so far this season, but perhaps should be doing better, as they are expected to win ~7 (±1.21) corners per game. Given that United are only expected to concede 4 (±1.37) corners per game on average, this part of the pitch will truly play a key role in determining the number of corners in this game. Like in the more traditional markets, the game will likely be decided by the effectiveness of Liverpool’s press and United’s ability to bypass it by playing a more direct style.
The higher number of corners on average so far this season for both teams, given their style of play, suggests a corners line of 11.5 maybe fair, but given how closely matched the teams are in terms of ability and because neither should be able to dominate to their usual degree this line should be set nearer the 10.5 mark. The volatility of a derby-type game increases the deviation of corners, which become hugely dependent on intent and game state, but as a deadball view we would seek to back anything under a line of 12 or greater to begin with.
My hope for this post is that not only will it provide some interesting data driven observations for those of you who enjoy trading corners (or are looking to do so), but also highlight where teams are over-performing and under-performing relative to the number of corners they win/concede.
We can see that if you’re planning on trading over or under in a Tottenham match, it’s most important to look out for whom they are playing against, as this appears to dictate their corner stats more than any other factor. Both Leicester and Chelsea have been relatively consistent with their corners this season, so we can potentially expect both to maintain their status quo in a match totalling less than 10. On the other hand, Manchester United are somewhat less easy to predict, as their performances vary wildly along with their chance conversion metrics, even given their relatively consistent xCorner trend.
Until next time,
Sagar Jilka (@DrSagarJilka)
(P.S. Here’s the table of averages and standard deviations for all the teams discussed above – hopefully it will inform any decisions made this weekend!)