OMJ: Oh My Jonas

There are great performances and then there is this amazing string of Jonas Valanciunas’ last 5 games:

By comparison the highest Player Efficiency Rating (PER) in the NBA is currently 32.5 (LeBron James). Obviously a five game sample size against lesser quality players is was it is. But its certainly a good sign – and he’s been very good all season:

This (largely offensive focused data) is encouraging especially when several experts compare him to defensive anchor Tyson Chandler.

Ninja Defense

I supported the trade for James Johnson for Miami’s late 1st rounder and had a couple good debates with NBeh (a Wages of Wins network site) here (at the time of the trade) and here surrounding the merits of the deal.

A smart Raptors fan emailed me a few times on the weekend asking why the Wins Produced database has James Johnson as a PF only and what impact that may have on his WP48. He also wondered exactly what’s hurting his WP48. I defer to the Wins Produced experts for the exact impact of slotting him in as a SF vs PF (BTW, this is a small pet peeve of mine as players can be pegged in a position, but perform slightly different roles that the “average” 3 or 4 on both offense and defense, for example). We do know the WP position adjustment can be material and certainly pegging him as a full time Power Forward hurts. We also know that Johnson also plays the “3” more (including the top 6 line ups in terms of MP) more often than the “4”.


So being pegged as a “4” hurts his WP48 score. But what else is hurting him?


Certainly his true shooting percentage is the main problem. We have also “cheated” as he probably should be a “3.4” and compared to both SF and PF. Regardless, his eFG% of 44.0% is below the league average for SF of 48.2% (min 15 mpg) and PF of 48.4%. His FT% hurts him even more: 57.5% versus the SF average of 76.5% is unacceptable. On the plus side, he rebounds well – more in line with “full time” PFs and leads all SFs in blocks. And while he gets “rewarded” for blocks and steals under the WP formula, I believe the “team adjustment” hurts him as his individual defensive contributions are superior to many in the league. This is partly derived from the “eye test”, but we also looked other metrics to help support our point. As an aside, we added Matt Barnes, who happened to screen just above Johnson. It may suggest we dodged a $9-million-over-two-years bullet as the Raptors effectively have a similar contributor in Johnson.

While Johnson shows very well under Adjusted +/-, many in the stats community are moving more to Advanced Statistical Plus/Minus (you simply cannot have enough +/- metrics!). Daniel’s site has a great tool plotting Advanced Statistical Plus/Minus (ASPM) and Value over Replacement Player (VORP). Oh, and its super easy to understand. 🙂

Note Daniel also lists James Johnson as a “4”.


We appreciate Daniel’s work as it parses offensive and defensive contributions. We also point to this work by the “Sport Skeptic” which examined the how various models (e.g. “new” and “old” WP, APM, ASPM, RAPM etc) performed: “…ASPM appears to do a really good job overall. It describes what happened well, it predicts the next season well, and it contributes a good amount to both of the best blends.” (See the link for a full description of his methodology)

Questions? Arsenalist introduced a forum thread dedicated to “Statophile Q&A”. If you prefer to send questions privately, you’re welcome to email me at tomliston [at] gmail [dot] com or find me on Twitter (@Liston).

Share this:
  • Puffer

    Love these articles. Informative, intelligent and based on real world results.

  • James

    Who is the competition that Jonas is putting these numbers against ? Is this the Lithuanian league ? The competition makes a big difference on what PER really means and implies.

    If these numbers are from his team competing in Euro League, then what was Paul or Mark Gasol’s number ? What was Tiagu’s numbers ?

    • Nicolas Denis

      The author states how the given PER numbers are not to be directly compared with the same NBA stats.  It still tells us he’s dominating his competition which is a good sign.

      I would, however, love to have the same stat calculated for the players you named.  Anybody?

    • Yes its the Lithuanian league. I bolded “lesser opponents” for that very reason.
      Do you mean Tiago Splitter? If so:

      • James

        Thanks Tom, yes , I did mean Tiago’s numbers. I don’t think numbers from lit league mean anything !!!! Tiago’s number are from euro league . Is lit league even as good as D league ??? People have to be very careful not get too excited about Jonas . For example, what was the numbers of other Lithuanian players on lit legue before joking nba ? Lina’s or their pg or …. There were few big men from lit which played few season in nba and could not make it at the end

        • Tim

          That is true. If I am not mistaking, Raptor’s had a Lithuanian big man who played for their national team as well. What was his numbers when he was playing in Lithuanian league ?

  • cjmcdnld


  • Nicolas Denis

    If the defensive and offensive VORP are equivalent in value, why would you scale them differently in the graph?  Makes for a skewed analysis…
    I also assume from the graph that a negative defensive VORP is a good thing.  Is that right?
    Good work!

    • I didn’t scale them.  The “presentation” code did – not sure what Daniel used.
      You’re right, it does skew the graph a tad – but you can at least see JJ’s rank on defense vs others.

  • 2damkule


  • Statement


    Surely my reading comprehension skills suck, so please endulge me.

    I understand from reading the Sport’s Skeptic’s work that in order to explan what happened, the old WP and ASPM work well.

    In terms of the prediction side of things, the article states that a weighted combination of 4 variables is useful for prediction.  Which variables is the author referring to.

    Also, when predicting stuff, if you do an in-sample forecast, you come up with a forecast error.  My question is, what is the dependent and “real data” values that the author is using to get his prediction error.  I don’t really get it.