Almost a month ago, we discussed how to apply some basic clustering techniques on top of Euroleague Data. However, as I stated multiple times, building AI models just with shotchart data is not enough for the vast majority of purposes. For this reason, this article has two main goals: (a) presenting a new lineup combination dataset, available online for free, and (b) examine different historical lineup patterns in the last 12 seasons of Euroleague.
Once again, thanks to Nacho Gámez (@ngamezj in Twitter), I was available to deal with large amounts of raw Euroleague play-by-play data from seasons 2007–2008 to 2018–2019. In particular, 1379919 events were registered in the given .csv files, which contained the type of action, the team+player performing it and other stuff such as the scoreboard or the game clock at that specific moment. On the one hand, the corresponding lineups for both teams in every single event were assembled by checking the substitution-timeline on court. More concretely, in 12 years of competition, 42090 unique 5-players lineups were obtained. On the other hand, (simple and advanced) statistics were computed for every single combination, not only in offense but also in defense.
However, the main bias of the “plus-minus” stats prevails: roughly, numbers might not express accurately the contribution of a single mediocre player surrounded by superstars (or viceversa). For instance, if your lineup is composed of Milos Teodosic, Nando de Colo, Nikita Kurbanov, Kyle Hines and Adrià Arbués (~1.80m, decent range, but skinny af), this last random guy might be in the Top league rankings: “wow, when Arbués is on court, his team has a NET rating of +20, although he has a 1% FG percentage”. For this reason, not only the team performance of the lineup will be examined as a whole, but we are also gonna break it down into pieces and permutations. As a matter of fact, we will check all possible player combinations in groups of 4, 3, 2 and 1 player. In the given example, if we divided that lineup into groups of 4, we would obtain: group 1 — Teodosic, De Colo, Kurbanov, Hines (arguably the best one), group 2 — Teodosic, De Colo, Kurbanov, Arbués, group 3 —Teodosic, De Colo, Arbués, Hines, group 4 — Teodosic, Kurbanov, Arbués, Hines, and group 5 — De Colo, Kurbanov, Hines, Arbués. Just for the record, dividing into quartets results in 5 combinations, threesomes and couples in 10, and individuals in 5 too. The total number of possible combinations is nuts:
- Whole lineup (5 players): 42090 unique combinations.
- Quartets (4 players): 88968 unique combinations.
- Trios (3 players): 62716 unique combinations.
- Duos (2 players): 19232 unique combinations.
- Individuals (1 player): 8325 unique instances. In this last experiment, I considered that a unique instance represents the performance of a player in a particular team; if that player has been in the roster of N different teams across the seasons, it means that there are N different instances that relate to that player.
Furthermore, the meaning of the obtained results must be clarified: (eg.) imagine we split into couples, and while checking the performance of Diamantidis-Batiste, we see that they scored 78/200 three point shots; this rarely means that between Diamantidis and Batiste shot 200 times from deep, but instead, that the Panathinaikos team, with both Diamantidis and Batiste on court, shot that specific amount of 3’s. These shots could have been attempted only by Diamantidis or Batiste, sure, but they could have also been attempted by any of the other 3 players that coincide with them on court (Fotsis, Nicholas, Jasikevicius…).
In this “article”, we’ll analyse the best and worst lineup configurations in this last 12 Euroleague editions, but you can also do it for yourself! In this Google Drive shared folder, you can find 5 csv’s, which contain all different types of combinations together with their corresponding stats. Handling the shared data with Python (Pandas) or R is highly suggested, since Excel may not excel with such a large dataset. Anyway… Let’s go!
Before starting, it must be stated that we’ll talk about OER / DER / other metrics; if you want to know more, I encourage you to read “Basketball on Paper” or check Basketball Reference / NBA Stuffer.
Number of Possessions
Computing efficiency metrics consists in the normalisation of scored/allowed points per possession. However, large deviations can pop up if the sample size of possessions is small: for instance, if a particular lineup only coincided on court for one possession and they scored a 3 point shot, their OER would be 300! For this reason, we’ll have to filter out those combinations that don’t reach a minimum number of possessions; this is a really hard task, and of course, it will depend if we check 5-/4-/3-/2-/1-player combinations. I am pretty sure that not even Dean Oliver would know by hand which thresholds should be set in this particular scenario, so the best way to find them out is to rank the lineups in terms of possessions. For every single contribution, and taking only possessions of their own team, we may see that:
Logically, as we decrease the number of players in the given combination, the number of possessions increases. While the lineup of Llull-Rudy-Darden-Mirotic-Darden was on court for 312 possessions, Llull by itself was on court in more than 10k! In the displayed figure, the perseverance of Olympiacos can be seen, with the everlasting Spanoulis-Printezis, and Siena emerges in the quartets part, with McIntyre-Sato-Stonerook-Eze, which were often complemented either by Bootsy Thorton or Rimantas Kaukenas.
In the following figures (except some particular cases), the following thresholds will be applied:
- Whole Lineup (5 players): minimum of 150 possessions.
- Quartets (4 players): minimum of 300 possessions.
- Trios (3 players): minimum of 600 possessions.
- Duos (2 players): minimum of 1000 possessions.
- Individuals (1 player): minimum of 2000 possessions.
Offensive Efficiency (OER)
Let’s start where the fun begins: scoring! In this subsection we’ll examine the best/worst teams in terms of offensive efficiency (points scored/100 possessions). Without further ado…
Top Lineups: impressive or not, the top lineup outperforms the 2nd one with +22 points every 100 possessions. That CSKA machine with Teodosic and Vorontsevich in their prime was too much for their opponents. Baskonia was one of the first teams to use a shooting 4, such as Teletovic; the Bosnian guy, together with a top-scorer like Rakocevic and the Prigioni-Splitter pick-and-roll dynamic duo was a huge success. The two finalists of the 2018 Euroleague edition are also there whilst sharing some patterns: offensive point-guards (Sloukas, Wanamaker, Llull, Carroll), stretch-4’s (Melli and Randolph) and excellent rollers + rim protectors (Vesely and Tavares). This pattern seems to be excelling in the last lustrum. Finally, part of the Efes squad that reached 2019’s Final Four is also there: it is quite remarkable that they managed to be in this top-5 without the Larkin-scoring madness of 2020 (data from the 2019–2020 season wasn’t included).
Bottom Lineups: Euroleague coaches are notable team orchestrators, and their hard effort is directed towards building flawless teams. For this reason, lineups that coincided a large number of possessions on court usually performed at least decent in both the offensive and defensive side; thus the offensive efficiency of the bottom-ranked lineups is not dramatic. In fact, those OER numbers could be notable if they still managed to have a low DER. For instance, the “worst” team is composed of 5 amazing Siena players that any GM would hire in 2010; that team, although not being particularly great in the offensive end (McIntyre was the only top-shooter from deep), still managed to have a NET Rating of +6. Similar cases happen with Pana and Red Star teams, which reached (at least) the Euroleague Playoff’s. The last couple of teams that appear in this ranking make a little bit more of sense: both Aris and Cedevita had great players (Massey, Wright, Bilan), but they were bottom-league teams that didn’t have a slight chance of making it to the Final-4. Moreover, we are talking about low-budget teams, which resulted in short rosters, thus forcing them to coincide in many possessions through the season despite not being optimal for their team.
But hey, was that CSKA top lineup compensated? Were there other couples that produced higher offensive efficiency for their teams? Not really…
I swear that the gathered dataset was not CSKA-biased or whatsoever, but damn, they monopolise the OER categories in every single way. Moreover, it is pretty fascinating that players from different seasons emerge (Clyburn in the individuals, for instance). Another fascinating thing is to see that the best offensive efficient couple is Kurbanov-Hines; no one would say that they are top-scorers, but they are the perfect complement for the team engine in order to produce points as hell. Finally, just Simon-Moerman emerge in this CSKA hegemony.
Defensive Efficiency (DER)
In order to contextualise offensive efficiency data, the number of allowed points every 100 possessions (DER) has to be also detailed. As I stated before, coaches may not care building lineups that are not extremely efficient in offense if they clearly outperform their rivals in defense…
Top Lineups: The same Red Star team that showed up in the worse offensive setups leads the defensive ranking! Quincy Miller defined that lineup in Twitter as “some clamps right there”, and I couldn’t agree more; maybe none of their guards-wings (Jovic, Kinsey and Dangubic) could shot decently from the arc, but folks, that Kombak Arena was fire: not even 75 points allowed / 100 possessions. In fact, the home-court conditions of the top defensive teams is vital: crowded stadiums acting as a 6th man on court proved to be crucial in terms of pressuring the opponents. Another example could be Olympiacos: I bet that no one would like to play a road decisive game in the “Pace and Friendship” Stadium with guys like Mantzaris or Papanikolau playing their tough defense during all 40 minutes (if you are lucky and there’s no overtime…). The same pattern of non-shooting guards is repeated in Maccabi squads, specially in the one with Perkins and Pargo in the back-court (Kane and O’Bryant were slightly better despite not being snippers). The Siena vintage team is also included in this top 5 (swapping Kaukenas for Thorton), proving that elite teams change their strategy in this last decade by both scoring/allowing more points per possession with the 3-point madness.
Bottom Lineups: even if you have the 2-times Best Euroleague Defender in your roster (Dunston), your team defense might be weak if the other pieces don’t fit in. These pieces could simply be players (such as Heurtel or Derrick Brown, not excellent defenders), but also team chemistry: Efes has struggled trying to build a successful team by acquiring a large number of topnotch players, but they had to keep trying for years in order to establish a solid base of players that can share the ball on offense and focus on defense. I wouldn’t say that Jayson Granger, Cedi Osman or Tyler Honneycut (RIP) are bad EL defenders; in fact, all three have strong legs and (except Granger) large wingspan, but their lack of presence in offense (you needed 4 balls in order to feed that team) was turned into lack of commitment in the other end. The same happens with Maccabi, with Pierre Jackson being a top-scorer but mediocre leader in terms of ball sharing; it is curious enough to see how DeAndre Kane is both in the top and bottom lineups. Once again, we can see a CSKA lineup… Surprising? Not at all. That lineup is the one with an OER higher than 150, which suggested their coach to let them chill in defense, knowing that barely any team could outscore them. Finally, there’s another low-budget team, Chorale Roanne, which had the Alphonso Ford scoring champion (Marc Salyers) but didn’t manage to stop their opponents.
Let’s break it down into pieces now…
Once again, Red Star leads this ranking, but hey… Is that the same team? Not at all! In fact, it is a complete opposite team from the one we have seen before, where the Balkan character of Kalinic, Mitrovic or Marjanovic was strongly present. Moreover, other remarkable things can be spotted: Barcelona did not appear in any 5-player lineup, but their defensive performance with Pascual was memorable. The team that won the Euroleague in 2010 pops un in the quartet, duos and individuals! Athletic guards like Sada or Rubio complemented the natural offensive talent of Navaro and combined perfectly with tough players such as Pete Mickeal. Similarly, the combination of Diamantidis with undersized power 5’s (Lasme or Batiste) proved to be successful protecting the rim. Last but not least, it is pretty awesome to see how the presence of a particular player may change the team attitude: I am talking about Vitaly Fridzon! The duo of Teodosic and Vorontsevich was extremely prolific on offense, but defensively, they were quite sloppy (bottom-2 5-player lineup); instead, with Fridzon, they build the most efficient threesome defense.
Overall Efficiency (NET)
Net Rating is the best way to contextualise offense and defense in one same metric: by computing the subtraction of OER-DER, the outscoring difference per 100 possessions is obtained. In this way, rankings show those lineups that have a better balance of both notable offense and defense.
Top Lineups: as it was previously stated, the offensive CSKA team that had +150 OER also allowed a lot of points per possession; for this reason, that lineup is ranked “only” in the 5th spot. Instead, another vintage version of CSKA proved to be more efficient: watching Holden-Langdon-Siskauskas-Andersen-Goree was an absolute beauty in both ends. In the other three Top-5 NET Rating teams, we can distinguish two cases: (a) Fenerbahce, who had an excellent offense and a notable defense, and (b) Olympiacos and Maccabi, who had a notable offense and excellent defense. Both styles proved to work fine while being different basketball versions.
Bottom Lineups: in this scenario, there are 3 extreme cases: (1) teams without skilled-players and low-budget that ended in the last standings EL spots (Aris, Partizan and Alba), (2) teams with skilled players but with a clear lack of chemistry (Efes), and (3) Real Madrid! Madrid reached the Euroleague Final with Chacho-Rudy-Suarez-Mirotic-Begic, but jesuschrist… That lineup performed poorly. I bet they did some research after losing the final against Olympiacos, resulting in Suarez and Begic leaving los blancos.
In terms of break-down performances:
Similar patterns can be recognised in this ranking when comparing it with the DER one. First, it is proved that although Vitaly Fridzon averaged 6.8 points per game, he was one of the CSKA’s success in this last decade, being the perfect role-player to share court with Teodosic or De Colo. Second, Rudy-Mirotic, who where included in one of the bottom-5 lineups, step up in the threesome classification, by creating a much better fit with Bourousis than the one they had with Begic. Finally, the lowkey presence of Panathinaikos and Barça emerges in particular situations: the most remarkable one is the presence of MVPete, who leads the individual ranking of individuals with a +22.1 NET Rating (outperforming Teodosic by +4); although Barça did not appear in any kind of lineup combination, the presence of Mickeal was always positive for their team overall’s performance.
Offensive Rebounding (ORB%)
Apart from efficiency at both ends, the other Dean Oliver’s factors are also analysed (well, not really, eFG was strongly related to OER so I left it out at the moment!). First, we’ll see which teams had a stronger/weaker presence in terms of offensive rebounding.
Top lineups: although it has been proved that it is more likely to capture an offensive rebound after a missed 3 than a missed 2, the ORB% lineup leader is the vintage 2009–2010 Zalgiris that didn’t shoot that much from deep. The best feature of that lineup was the strength of their power-forwards: although neither Salenga, nor Klimavicius, nor Watson reached the 2.05 meters high, they were fearless in terms of rebounding, and their athletic conditions made them a tough top-16 team. The situation of Partizan is quite similar: apart from having Pekovic in their squad, the wingspan of Tepic or Velickovic when crashing the glass was really prolific. Olympiacos appears in this top-5 once again, but with new members: in this lineup, the dynamic kamikaze duo of Spanoulis and Law was the main focus of many defensive sets; while the small back-court drove to the basket trying to score tough layups, the energy of the front-court (with a young-fresh Antic) was there to capture 45% of the missed shots. Finally, Madrid and Fenerbahce also made it to this list and personally, their presence is admirable: while all the previous teams had borderline shooters and powerful frontcourts, Madrid’ and Fenerbahce’s lineups consist of accurate shooters with decent shooting-selection. If you know that you are gonna miss 60–65% of your shots, it is worth it to crash the rebound, but if you miss only 40–45%, is crashing worth it? Of course, it depends on the specs of your players, but I assume you can take the risk if you have hand-magnetic players like Reyes and Ayón or skilled long guys such as Melli or Vesely.
Bottom lineups: watch out! In my honest opinion, this particular ranking should be only analysed experimentally, since crashing the offensive rebound is often a coach’s choice. This means that, although we can still rank lineups according to their “poor” performances, coaches might prefer to retreat on defense instead of crashing the board. As a matter of fact, two different lineups of Brose Baskets can be found in this ranking with 10 different players and only one coach; furthermore, Melli, who was previously included in the top ORB% team, is now the power-forward of the weakest one… Suspicious! Lokomotiv Kuban, a team that reached the Final 4 of 2016, is another clear example: no one would say that Claver, Randolph or Broekhoff are included in a team that is not able to capture offensive rebounds. You better ask Georgios Bartzokas.
When checking the decomposed break-down, we can see three common factors all-around: Othello Hunter, Gustavo Ayón and Felipe Reyes, being both of them different types of interior players of this last Euroleague decade. As it is clearly displayed, Madrid monopolises this statistic while being a team that has a notable 3-point shooting selection and optimised spacing. Once again, no one would consider the presence of short guys like Llull, Chacho or Carroll as a key feature to capture rebounds, but getting those boards is not only a matter of skill, but also relocating the defense and take profit of the resulting miss-matches once a shot is attempted.
Assist Percentage (AST%)
Every single coach wants their players to take the best possible available shot, which is normally an open attempt from a comfortable position. These shots are generally assisted (on the contrary, we have the midrange pull-up 2, but I won’t go into that…); let’s see who made it to the list!
Top lineups: Jasikevicius hasn’t been coaching for a long time, but their teams really know where to place the ball in their offense. That lowkey team that reached the final-4 in 2018 had 2 skilled point guards (Pangos and Micic) plus a talented combo (Westermann), and Saras took profit of it; with a clear “extra-pass” concept and grounding many plays on the Pangos-Jankunas pick and pop, Zalgiris assisted almost 1 of every 3 baskets when these guys played together. Madrid and Fener are there as always, with the additions of role-players such as Taylor, Maciulis or Kalinic, and center with quick passing hands, like Ayon or Udoh. Finally, someone has to build a Calathes’ statue in Athens: if you take a closer look of the 2nd ranked assisting team, you’ll find a snipper (Feldeine), a role-defensive players (Pavlovic) and 2 rollers (Gist and Radulijca)… That’s all Calathes needs! The greek point-guard is the only one in that lineup excelling in ball-handling (I wouldn’t rely on Feldeine and Pavlovic playing a PnR sequence if I need 2 points to win), and they assisted almost 1 of every 4 baskets. Completely Calathes’ fault.
Bottom lineups: oh wow! Did you recall that Baskonia was a top-2 team in OER? It was definitely not when they had Mickeal and McDonald on their team… Although their lineup had a lot of natural scoring talent, only Planinic could share the rock; by playing a lot of iso, not even 10% of scored shots were assisted. We find once again the Law-Spanoulis team here… Can someone guess why? It’s pretty simple: if your team captures a lot of offensive rebounds, many of those will be tip-in’s, putbacks, or just a KyleHines-like-shot were he muscles up and scores when surrounded by 4 opponents. These shots have no reward in terms of assists, so generally, lineups with high ORB% also have a low assist rate. Finally, we find our old friends from Siena, Partizan and Aris again: those teams struggled to death to score, and were placed in the bottom-5 rank of OER; visualising these data, it is crystal clear that ball movement was one of the main causes of their lack of offensive flow.
Apart from the common CSKA and Madrid hegemony, the top Zalgiris lineup appears twice when splitting lineups: it is frankly amazing to see how Pangos’ outstanding season was partially grounded on their confidence when popping with Jankunas or finding Milaknis wide open (he never had this setup in Barça, for instance). As it has been mentioned in the ORG% scenario, basketball fans would expect only guards in this graphic; instead, Ayón appears in quartets, duos (with Reyes!!!!) and individuals. Despite we can all agree that Ayón is a center with notable passing skills, one would expect that Diamantidis, Chacho, Calathes or Jovic had a greater influence… Nope guys! Assisting baskets is also about clearing space and sharing the rock from the high post, and there’s few guys like Ayón that can let you develop this game with such an ease. Finally, it is also remarkable to see how new CSKA players pop up: Sonny Weems and Sasha Kaun took profit, in their own ways, of Teodosic’s creativity.
Ball Control (TOV%)
We reached the end… The last factor: ball control! In order to measure how teams took risks, the percentage of turnovers (TOV%) is checked for every single lineup.
Top lineups: What do the first and second lineups have in common? You might got it… Xavi Pascual! The coach of Barça and Panathinaikos is known for being strictly methodic and having a huge playbook for every specific game: in this graphic we see that his top-2 teams did not lose more that 7.5% of their possessions…. Impressive! The modern CSKA version with Higgins and Clyburn joins the club of usual suspects, composed of Real Madrid terrific lineup and Udoh’s Fenerbahce season.
Bottom lineups: friends, physical power and a fast-break-based basketball approach comes at a cost of risk. As we have seen before, that Zalgiris team of Salenga-Watson was the best one when crashing the offensive board, but did they know how to play organised basketball, or were they only focused on fighting under the boards? Moreover, every time you get the ball, you are also getting a new opportunity of losing it… The same happens with that crazy team of Partizan that reached the Final Four with McCalebb’s explosion: that team was going up and down 40 minutes, I honestly don’t know how Slavko Vranes managed to run the floor. Finally, we can see the first appearance of a Ljubljana’s lineup, with some classical European veterans such as Markota, Ilievski or Saso Osbolt; it wasn’t their best season, by far.
Well folks, do you need more evidence that Laso’s Madrid has been the main alpha-team when it comes to the 4 Dean Oliver’s factors? They also rule the separate turnover ratio combinations, with some key players like Llull or Carroll: the first one would rather attempt a high-accuracy impossible shot before losing the ball, while the second can find perfect equilibrium in the air at any given instance. Furthermore, Ayón also excelled in this game aspect… It is quite obvious that basketball is changing and Tavares is more dominant than Ayón for Madrid’s current version, but the Mexican (in his prime) has been performing excellent in a large number of facets without being in the spotlight of Llull or Chacho. The only outliers in this last chart are Lazic-Bjelica, role players of Red Star that based their game on tough defense and easy offensive concepts.