Previously we’ve looked at which teams tend to do well in the MLS Superdraft so I thought it would be fun to look at the relation between managers and the universities that are developing the players. I’m using percentage minutes played as a proxy for the quality of the draft pick. I’ve restricted the data to draft picks from the first two rounds from the last 5 years. There isn’t enough data yet for the 2011 season (rookies tend to see more playing time later in the season) so I have excluded that from the set. Also, to facilitate displaying the graph, I’ve restricted the set down to the current set of managers, not all of whom have draft picks from the first two rounds prior to 2011. I used to previous linear regression to set a baseline for whether or not a draft pick is under-performing based on their selection number. The color and size of the nodes indicates on average if draft picks for that entity under or over perform (red=under, white=average, blue=over perform). The color and size of the edge represents a single draft pick (for example, the thick line between Hans Backe and St. Louis University is Tim Ream).
Archive for MLS SuperDraft
MLS Draft Social Graph
On the MLS Draft — Top Ten Picks
Here’s part 1 and part 2 of our series on analyzing what a team can expect given a certain pick in the draft. So far we’ve looked at picks 11-20 and 21+ so now it’s time to take a look at the cream of the crop.
Here’s what we found:
- Only 10% of players never started a game in MLS
- 16% of the class are no longer in the league. Of that, half moved on to play in Europe.
- 42% of picks were starters in their first season, with a further 22% growing into starters.
- Expected starts are 12.09 per season
So how does the Top Ten compare to the rest of the draft picks? Here’s a look at how our three segments compare to each other:
Not surprising, the top ten picks do extremely well in MLS. Up next is a look at the tradeoffs of combining picks to move up in the draft. Stay tuned…
On the MLS Draft — Picks 11-20
Previously we looked at what the Sounders could do with their 2nd round picks by examining picks greater than 21 from the last 5 years. You can find the results here. That gives us a good idea of what the Sounders could get if they keep all 3 picks, but not what they could get if they trade for higher spots in the draft. Today we’ll be looking at spots 11-20 in the draft. The Sounders currently have the 11th pick.
For picks 11-20 in the last 5 years:
- 32% have not started a game in MLS. However, unlike spots 21+, some of these guys drafted in 2009 and 2010 are still with the league and could see time this season
- 48% are no longer with the league, however, that figure includes Jozy Altidore who was sold to Villarreal and Dominic Cervi who opted for Europe instead of playing in MLS
- 26% see an increase in games started and “grow” into starters, 42% are starters for at least a season
- 2009 was the best performing class, in stark contrast to their performance in spots 21+
- All classes except 2006 saw growth year over year. 2006 was flat due to most players not surviving in the league for 5 years
- Expected starts is 6.80 games per season
If the Sounders were to trade all 3 of their second round picks for a pick in the 11-20 range, they would have a slightly better chance of getting a starter, but they also increase the probability that their pick will never play in MLS. The shotgun blast approach to the draft is akin to diversifying an investment portfolio and mitigating risk. However, there are other concerns such as roster size and salary cap that make diversification a less appealing strategy. In a later post we’ll look at different draft strategies to see which is the optimal approach the Sounders should take with their picks.
On the MLS Draft — Late Draft Picks
In a previous post, I examined the overall trends of the last 5 years of the MLS Superdraft and created a linear regression to estimate the expected number of minutes played a given draft selection would contribute. This worked reasonably well, but it didn’t tell you much about player growth. There is a myth that even though the later picks don’t contribute right away, a lot of them can grow into starters. Given that the Seattle Sounders currently hold the 21st, 27th and 29th picks of the 2011 MLS Super Draft, I decided to take a deeper look at 2nd round picks from spot 21 and lower to see how likely it is that the Sounders can find someone who can grow into a starter.
In order to normalize across the different drafts, I decided to look at the data based on number of years in the league, where t=1 is their rookie season, t=2 is their second season and so forth. What I found was:
- The expected number of games started for a player is 5.63 per season
- 52.5% of players never start a game in MLS and are out of the league within 3 years
- Of all the seasons played by all the players, only 19% of them had 15 or more games started (majority of games given MLS’s 30 game season)
- Players either contribute in their first season, or don’t contribute at all. Only 10% of the picks increased their number of games started over the 5 years. Those players are Corey Ashe, Marc Burch, Andrew Jacobson and Peter Lowry.
- On average, year over year growth is flat or negative across the draftees, with the exception of 2008
- Much like the 2011 draft class, the 2009 draft class was regarded as being loaded with talent, yet that talent doesn’t seem to be evenly distributed. The late 2nd round picks from 2009 performed the worst out of the last 5 years and only two players started games (5 and 3 games). Those same two players are also the only ones who are still under contract with the league.
Given the information above, if the Sounders were to keep their picks there is
- 14.5% chance that none of their 3 picks ever start a game in MLS
- 38.8% chance that at least one of their picks becomes a starter
- 46.7% chance that they end up with several role players
38.8% of drafting a starter might sound like decent odds, but the question is, can the Sounders improve those odds by trading up? Stay tuned as we take a look at the draft odds for spots 1-10 and 11-20. In the meantime, take a look at our previous draft posts.
.
On the MLS Draft — Which Position To Pick?
The MLS Superdraft is one of four ways in which teams can reinforce their squad. They can also sign players from their academis, sign players from outside the league that they scout themselves or be allocated players from the US National Team pool. Given that there several options on how to get new players, it’s important to think about what type of player you want to get through each method. For the MLS Superdraft, I wanted to see if any patterns appeared that would indicate drafting a certain type of player would be more successful than others.
I started by looking at the breakdown of what types of players were being drafted and compared that to what the breakdown would be for a 4-4-2 formation. In the MLS Superdraft, forwards and midfielders are over represented while defenders and goalkeepers are under represented. For this analysis, I am excluding goalkeepers because the sample size (n=6) is too low to get meaninfgul results. What this means is that teams are biased towards so-called “flair players”. Are they making the right decisions?
Keeping with using minutes played as the basis for our success metric, I looked at the distribution of players over/under performing based on the expected number of minutes they should be playing based on a linear regression (more information on that can be found here).
I noticed two things looking at the distribution. One, the number of midfielders who greatly underperform was quite high, and two, there seem to be a lot of defenders who overperform by quite a bit (look at how the data tails off to the right compared to sparse data for M/F). Given that defenders seem to be performing better than their flair counter parts, I wanted to see if there was a drafting strategy that would emerge, something like “Pick flair players early and save your later picks for defensive players”. I decided to look next at how players perform based on both their position AND their selection number.
Looking at defenders first, it seems like across the board they are a good pick. No matter where in the draft you are, it looks like there quality defenders available. Moving on to midfielders, it looks like only in the first few draft spots will you consistently get someone good who can start. Late in the second round it looks like there are some guys available who can sub, but in general, midfielders tend to underperform (which begs the question, if there are so many midfield flops, why do teams keep drafting them?). Lastly, looking at forwards, they really seem to be a mixed bag. In the first few spots, they perform as expected, but after that it’s hit or miss as to whether or not they will see playing time. In fact, looking at the significance of the linear regression by position, the regression is not significant for forwards (p=.12). If you look at the top point scorers (points = 2*goals+assists) from the draft for the last 5 years, the top 10 scorers are:
- Robbie Findley
- Sacha Kjlestan
- Kei Kamara
- Dane Richards
- Yura Movsisyan
- Dominic Oduro
- Mehdi Ballouchy
- Steve Zakuani
- Jasan Garey
- Jozy Altidore
Not exactly a Who’s Who of goalscorers in MLS. What can we take away from this?
- Starting midfielders can be found in the first 5 spots. After that the odds are against you.
- Defenders are undervalued in the draft. They perform well, but teams tend to ignore them.
- Forwards are a gamble. High risk, moderate reward.
Thinking about why midfielders and forwards perform so poorly in the draft, I think part of it has to do with the talent drain that occurs before the draft. Talented flair players like Charlie Davies, Alejandro Bedoya and Marcus Tracy opt to skip the draft and go straight to Europe. The best midfielders and fowards aren’t participating in the draft. It’s very rare for a defender to make the jump straight away. Also, in looking at the migration into the league, most of the players being imported are flair players. Guys in the draft are competing with designated players and other new comers like Alvaro Saborio and Fredy Montero. They’re finding it a lot harder to get playing time and teams are finding really good options on the open market.
On the MLS Draft: Linear Regression
First I wanted to look at how the percentage of minutes played changes based on the selection number of the player. Aggregating the data from the first two rounds from 2006 to 2010, a nice linear pattern emerges. We can use a linear regression to estimate what a player’s expected percentage minutes player should be and whether they are under or over performing.
Another thing I noticed was how bad teams were at predicting talent. Approximately 25% of the first two round draft picks never make an appearance in MLS. Given that the draft is one of the main sources of acquiring talent (although this is changing), I found this number appalling. I wanted to see if any teams were good at drafting players or if it was a random toss of the dice. To estimate draft success I looked at how each team’s picks compared to how the linear regression estimated they should do.
Philly’s poor performance can be attributed to their strategy of drafting young players and only having one season for them to develop. A few more seasons are needed to determine whether or not their players are panning out. I was surprised to see the Seattle Sounders performing below expected because Steve Zakuani has been a wonderful pick. However, looking at their other picks, David Estrada and Evan Brown have not performed up to expectations and Brown has been released from the team. I was really impressed with the LA Galaxy’s record. 3/4 of their backline this season came from the draft, including newly capped Omar Gonzalez. In fact, half of their picks were starters in the conference semifinals.
I also wanted to look at which universities produce the most successful players. I was curious to see if some universities were talent pipelines to MLS.
The darker colors indicate the number of players drafted. The data was filtered down to only universities that had 2 or more players drafted. Wake Forest and Notre Dame tend to have a lot of players drafted, but they aren’t very successful. I was a little shocked that year after year teams picked from these universities. University of Maryland, however, seems to consistently produce talent in large numbers. Also of note is that players who didn’t attend college tended to under perform. Definitely there have been some players that were drafted and didn’t pan out, but others like Brek Shea, Fuad Ibrahim and Jack McInerney still show promise and might take longer before they become everyday starters.
On the MLS Draft Part Two
I decided to take a look at how last year’s draft class was doing. Specifically, I was wondering if I would see the same patterns in year two or if last year’s class was behaving differently. More or less the same patterns appear for the universities producing the draftees. Overall, the same exponential decay seems to happen with playing minutes as the selection number increases. However, we don’t see the same pattern of defenders taken late in the draft out performing their expected minutes. Instead we see defenders outperforming for middle picks instead of just in the tail. For the class of 2009, the tail is almost completely filled with players who were cut from their team before ever making their MLS debut. In fact, 13 out of the 30 players taken in 2009 had little to no minutes in their first 2 seasons. The fact that they are no longer with their MLS clubs rules out the theory that they will provide future value to the club. Basically, MLS teams are really bad at figuring out who is good enough to play in the league. It will be interesting to see how these numbers change with the reintroduction of the reserve league and larger rosters. Teams might be more patient with developing young players and have the roster space to do so.
I’m going to continue to track how draftees perform over time with the end goal being a model to estimate the value of a draft pick given its position. For example, Seattle traded Stephen King to DC United for their 2nd round pick, 19th overall. Based on my preliminary findings, it looks like this player should expect to play about 15% of the time. You can also look at the variance of the playing time to estimate the probability you’ll find someone who could be a consistant starter vs. someone who will be complete rubbish. With this sort of information, it becomes easier to make data driven decisions. Did Seattle make the right call trading King for a draft pick? Seeing as he was down on the depth chart and getting 0 playing time, the Sounders were able to give themselves a shot at getting someone who could crack the lineup.
On the MLS Draft Part One
I’ve been playing around with Tableau Public a little more and I have to say I’m impressed. I decided to revisit some work I had done on the MLS draft. Mid-season I decided to look at the correlation between a player’s selection spot and the amount of impact they are having with their new team. Currently there is no good metric to estimate impact, so I used minutes played. Yes, it is a very imperfect metric, but it does provide an easy way to compare players of any position. The logic behind using minutes played is that it shows a baseline ability that player X is good enough to make it onto the playing field. If they perform well, they will be selected again, if not, then they won’t see much playing time. It tells us nothing about potential or future performance nor about the quality of those minutes played. Certainly there are unique circumstances in each team that could affect a players minutes, but as a stake in the ground to get started I think it’s a decent metric.
When I first looked at the data, I noticed that minutes played seemed to decay exponentially as the selection number increased, with a handful of outliers. The drop off was a little surprising. It shows that there are only a handful of players in the draft class that are able to come in and make an impact straight away. Looking at the data again, but this time with the ability to filter by position, I noticed that defenders taken later in the draft outperformed their expected minutes. Something to keep in mind if you’re looking for cover in the back and need someone to step up immediately.
I wanted to play around with Tableau’s mapping features so I decided to plot the draftee’s university or club and see what patterns appeared. There’s a heavy East Coast bias with Southern California getting good representation as well. Given that a lot of the traditional college soccer powerhouses are located in the Mid-Atlantic region and Southern California, this wasn’t too surprising. However, it doesn’t represent the current ranking of college teams. Midwestern schools like Tulsa and Drake who were in the top 10 of the NSCAA rankings at the end of 2009 didn’t have a single player selected while schools outside the top 10 had 17 players selected (out of 32). Most surprising was Notre Dame, with 3 players selected. Notre Dame wasn’t in the top 25 at the end of last season. The players selected have yet to make an appearance in MLS.
This is just the tip of the iceberg. I can’t wait to look at the previous years’ drafts as well as how players progress over the years. Take a look at the viz below and let me know if you notice anything I missed.
