Judging Patent Cases

by Dennis Crouch

For the chart below, I tabulated about 7,000 individual votes from the Federal Circuit Judges in patent cases decided 2014-2021. For each judge, I show the percentage of individual decisions that sided with the patent challenger; or patent owner in each case. The results here are not highly surprising.  Judges Moore, Newman, O’Malley, and Stoll are all more likely to favor the patentee while Judges Lourie, Hughes, Dyk, Prost, and Reyna are relatively more likely to side with the patent challenger. Note that the differences between the judges are not huge–at most <10% difference.

This is a very rough-cut of the data. It does not include all decisions, and I used some automated techniques to identify the winner.  This process was greatly aided by Jason Rantanen’s Federal Circuit Document Dataset that he makes publicly available.

The AIA substantially altered the role of the Federal Circuit and post-AIA patent challengers are much more likely to win appeals. That is why the data here is limited to 2014+.

 

58 thoughts on “Judging Patent Cases

  1. 15

    So much for securing to inventors the exclusive right to their discoveries. This is why inventors are boycotting the USPTO.

  2. 14

    The data seems to reflect a meaningful difference in tendencies among the judges. It doesn’t explain the difference, but over a large number of cases, the difference appears to be plain. And it appears correlated to people’s general views on the anti- or pro-patent bias of these judges. Two things come to mind:

    1. It would be great to see some benchmark for the statistical significance of variations in this metric.

    2. It would be great to see some explanation for the overall high level of anti-patentee decisions. 63-24-13 seems huge, and troubling, to me, unless there is some selection bias, as suggested above (e.g., mostly people with strong appeals tend to appeal).

    1. 14.1

      The more I think on it, the more I am leaning toward the selection bias hypothesis. We know that the majority of appeals result in the district court/PTAB being affirmed (1). This makes sense, because the district court/PTAB have all of the same info as the CAFC, and the district court/PTAB are staffed with competent legal minds that are trying to get it right. It should, therefore, come as no surprise that the district court/PTAB gets it right most of the time, and are—therefore—usually affirmed on appeal.

      We also know that the PTO makes up the largest single source of appeals to the CAFC (2). However, the overwhelming majority of PTAB decisions that are available for appeal to the CAFC are cases in which the PTAB rules for the challenger. That is to say, if the applicant wins on an ex parte appeal, then there is no one with statutory standing to appeal to the CAFC. Meanwhile, most of the time that the PTAB holds for the patentee in an IPR/PGR, the PTAB does so at the institution stage, and institution decisions are excluded from the CAFC’s subject matter jurisdiction. Therefore, the overwhelming majority of PTAB decisions that can go to the CAFC on appeal are cases in which the challenger prevailed.

      If the largest single source of CAFC appeals is the PTAB, and the overwhelming majority of PTAB decisions available for appeal are cases in which the challenger won, then if follows that the largest single group of appeals that the CAFC decides are going to be cases in which the challenger prevailed below. As noted above, it is only to be expected that the CAFC is going to affirm more often than it reverses, so it follows that there is going to be a selection bias in favor of most CAFC decisions favoring the challenger.

      (1) link to patentlyo.com
      (2) link to patentlyo.com

        1. 14.1.1.1

          “Disposed of”? Do you have the same success rate when someone other than you evaluates the strength of your argument?

          1. 14.1.1.1.1

            Absolutely.

            Especially seeing as I am willing to engage, as opposed to most naysayers who would rather comment and run away.

  3. 13

    As is typical, Judge Dyk was on the side of invalidating patents. His motto is “I’ve never met a patent I liked.”

    1. 13.1

      What then do you attribute that 18% listed in the graph portend to?

      Sure, that number indicates a less than 1 in 5 on matters that both sides feel strong enough to fight for (which for a Rational Actor viewpoint would hew to a starting point of 50%).

      1. 13.1.1

        What then do you attribute that 18% listed in the graph portend to?

        As always, the devil is in the details. For example, a precedential case involving a 101 decision has a much different impact than a non-precedential case involving a motion to transfer. However, the dataset used to come up with these numbers seemingly treats them identically.

        1. 13.1.1.1

          I entirely agree – context very much matters, and a “here’s 7,000 cases” approach obfuscates more than illuminates.

          Of course, the usual suspect sees this as merely ‘attempted spin’ – but who needs attorneys to understand the law anyway, eh?

  4. 12

    The most striking take-away from these data is that all of the CAFC judges hold for the challenger much more often than for the patentee. On the other hand, Prof. Rantanen’s data show that the CAFC affirms much more often than they reverse. I can see two ways to reconcile Prof. Crouch’s numbers above with Prof. Rantanen’s:

    (1) District Courts and the PTAB rule for challengers more often than for patentees.

    (2) A losing patentee is more likely to appeal than is a losing challenger.

    Obviously, these two are not mutually exclusive. Do you have any sense, Prof. C., how much of the reality of CAFC judgments is a function of #1 above and how much a function of #2?

    1. 12.1

      The most striking take-away from these data is that all of the CAFC judges hold for the challenger much more often than for the patentee.
      The standard for appellate review is substantial evidence for findings of fact. This is about as deferential as it gets. Considering the PTAB is known as a “killing grounds” for patents and the District Courts aren’t too far behind and many issues revolve around factual issues, then the affirmance rates are not surprising in the least.

      (1) District Courts and the PTAB rule for challengers more often than for patentees.
      This is true. However, if the PTAB and District Courts were decidedly pro-patent owners (and their decisions/rulings reflected that), the PTAB and District Courts would still be affirmed at a high rate. Again, it is the standard of review that is important.

      (2) A losing patentee is more likely to appeal than is a losing challenger.
      Both are very willing to appeal. The stakes (i.e., the cost/benefit analysis) usually demand that an appeal be filed. If you go to the Board or go to the District Court, you are almost always going to go to the Federal Circuit — assuming you have the money to do so.

      1. 12.2.1

        Naturally, I will be interested to read your findings. It occurs to me that most of the times that the PTAB holds for the patentee (i.e., when the PTAB refuses to institute), this is not appealable to the CAFC. I wonder if that fact is “skewing” the numbers, so to speak. In other words, if one assumes that the the CAFC is ordinarily going to affirm, but the only sorts of PTAB decisions that are appealable are those from the pile that is enriched in rulings against the patentee—while the pile of PTAB decisions that is enriched in rulings for the patentee is statutorily excluded from appeal—then it follows that the CAFC is necessarily going to end up holding for the challenger more often than not, simply as a function of the going-to-affirm assumption that is baked into the set-up.

        1. 12.2.1.1

          Thanks for reminding us of the many institution-denied IPR petitions that are de facto wins for the patent owners but cannot be appealed and thus do not show up in these statistics.

          1. 12.2.1.1.1

            You do not understand the point that you are trying to make Paul.

            Your tone and phrasing indicates an anti-patent holder bias.

            Painting “as a win” something that is bogus on its face and should not be entertained should NOT count in a CAFC judge by judge evaluation of those judges anti or pro patent proclivities.

            Not only do you seem to want to make it so, you seem to be implying that such SHOULD skew the CAFC to be more anti-patent.

            This is nothing but a logical fallacy.

    1. 11.1

      I recall such a decision a few months ago. The exact name escapes me.
      However, I (strongly suspect) that since that the other two judges went the way of patent owner, he just signed onto the opinion to avoid having to write a dissent. Had the other two judges split, I have little doubt that Dyk would have ruled for the patent challenger.

        1. 11.1.1.1

          Care to venture a guess how many times that happens? BTW, I don’t know the answer so if you don’t know, I don’t know either.

  5. 10

    Note that the differences between the judges are not huge–at most <10% difference.
    I think this dataset understates the real difference.

    I presume the dataset includes all cases. However, not all cases are the same. There are a great many cases that are disposed under Rule 36. While I’m sure a good many have merit, I also suspect that a great many are on appeals that should have never been filed. Moreover, I suspect that out of those cases in which a Rule 36 ruling was made, that the vast, vast majority were not in favor of the applicant/patentee/patent owner.

    Also each case has a different level of importance. A pro-patentee non-precedential ruling on a fact pattern that may impact <1% of all cases is far different than an anti-patentee precedential ruling on an issue that impacts 50% of all cases. Granted, there is no easy way to code the dataset for this. Regardless, the dataset does not take these differences between individual cases into account.

    Let's look at the differences between Neman and Dyk as an example. Dyk rules for patent owner in just 18% of decisions. Newman rules for patent owner in 31%. When one omits loser appeals that should have never been filed, my guess is that Newman is well over twice as likely to rule for patent owner than patent challenger. That is not a minor difference.

    1. 10.1

      “Granted, there is no easy way to code the dataset for this.”

      Is there a reason why limiting the cases to 2-1 decisions and then performing the same analysis wouldn’t produce a reasonable approximation?

      I’m not sure how meaningful that metric would be, but if the differences remained small it seems like we could conclude that the practical differences truly are small.

      1. 10.1.1

        Is there a reason why limiting the cases to 2-1 decisions and then performing the same analysis wouldn’t produce a reasonable approximation?
        The sample sizes start dropping which lessens the chance of getting statistically-relevant result.

        if the differences remained small it seems like we could conclude that the practical differences truly are small
        When you read the decisions over the years, it becomes abundantly clear who likes to fudge the facts/law to get to a certain result. I pointed out Dyk and Prost elsewhere as examples. I would be interested to know what results that particular duo produces.

    2. 10.2

      That’s right. I believe about 1/3 are R.36 judgments without opinion. The vast majority of those situations where the patentee lost below and the judgment affirmed on appeal.

  6. 9

    Is there a way to parse the data in the same manner for each 101, 112, 102, and 103?

  7. 8

    Re: “Note that the differences between the judges are not huge–at most <10% difference."
    If the sampled numbers include the many IPR appeals from the PTAB [subject to APA review standards], and ex parte application appeals from the PTAB, would not these more unanimous and more numerous decisions significantly dilute the percentage differences between judges in the much smaller number of decisions on appeals from District Court patent suits?

    1. 8.1

      agreed – another area of context matters.

      Or as marty would snidely put it – “hey, how about cherry-picking there?”

  8. 7

    The Four Horsemen of the Innovation Apocalypse:

    Congress . . . SCOTUS . . . CAFC . . . PTAB

  9. 6

    This chart needs a row at the bottom that shows the percentages for all votes. This can be used as a baseline for comparison with the individual judges.

  10. 5

    Knowing this, who would ever take the chance as as patentee with a 30% odds of winning. How awful. Truly only lawyers making money by bringing cases to the Federal Circuit for patentees. This plus IPR and the fix is in.

    1. 5.1

      Cases are not decided randomly, so you can interpret this data as showing the odds of success.

    2. 5.3

      Knowing this, who would ever take the chance as as patentee with a 30% odds of winning.
      30% odds of winning is usually better than 100% odds of losing. If you lost at the PTAB as a patent owner (the far likeliest scenario) you can either concede (in which case your invalidated claims are dead forever) or you can attempt a ‘Hail Mary’ by appealing to the Federal Circuit.

      Sophisticated clients understand the likelihood (or not) of prevailing. Unsophisticated clients are likely to lose regardless because the case probably isn’t salvageable by the time it gets to the Federal Circuit. Oftentimes, a good case is lost by unsophisticated client well before it reaches the Federal Circuit. Frequently, the application/claims are a mess because the unsophisticated client went with a low-cost patent attorney/agent. Also, because unsophisticated clients rarely have money to pay for good trial counsel, they are stuck with patent law’s version of ambulance chasers who do as little as possible in order to get a settlement from the defendant. They make bad arguments and/or fail to make good arguments. Even if the client “sees the light” and actually hires competent appellate counsel, appellate counsel is stuck trying to make lemonade out of lemons.

        1. 5.3.2.1

          I hope you aren’t in the stock market or a gambler if you’re accustomed playing odds that poor. 2:1 aren’t odds worth taking, especially when the buy in is hundreds of thousands, if not millions of dollars.

          1. 5.3.2.1.2

            You are not accounting for the fact that the issue is money–not winning a particular case. It wouldn’t make since to bet $1 for a chance to win $1 with a 30% change of success, but it would make sense to bet $1 to win $10 at those odds.

            1. 5.3.2.1.2.1

              Makes sense only for the lawyers. They are the only ones that get more than one at bat to beat those odds. For the patentee it’s one and done.

              1. 5.3.2.1.2.1.1

                My point is that the issue is expected payout–not likelihood of success. In any case, patentees make one-off probability assessments all the time, which is how they determine whether to settle or appeal. The actual odds depend on their specific case, not a 30% win rate in the abstract.

    1. 4.1

      LOL – yes marty, I can see that you, with your penchant for wanting to not understand the terrain of patent law — would want to spin it that way.

      But no, the point remains that when the issue at hand is a contested one, there is CLEARLY a statistically significant difference among the CAFC judges.

      But you be you and try to pretend otherwise.

      1. 4.1.1

        Yes, because people pay for appeals all the time without intention to contest the matter. Or do you mean cases that aren’t just R. 36’d into the dustbin of history?

        Not that it matters: with a sample size of 7,000, you can cherry-pick however you wish, but that’s a sizable corpus & it’s saying something real.

        1. 4.1.1.1

          You are quite missing the point there marty that lumping all 7,000 in together is obscuring the picture.

          As I said – you be you.

    2. 4.2

      Scene from Rex Stout’s The Doorbell Rang

      Mrs. Rachel Bruner : [gushing] Incredible! Is there anything you can’t do?

      Nero Wolfe : Yes, madam, there is. I couldn’t put sense into a fool’s brain. I’ve tried.

      1. 4.2.1

        Says the guy who insists on commenting all the while claiming that he cannot see the actual content to which he is commenting…

        That’s “I Use My Real Name” Greg for you…

        ¯\_(ツ)_/¯

  11. 3

    This is interesting. Thanks for assembling these data. I confess that I am surprised that the difference between (e.g.) Newman and Dylan is not wider.

    1. 2.1

      That certainly is the desired narrative.

      But that narrative is an obscured one – for reasons provided.

    2. 2.2

      You cannot measure favoritism based solely on outcome. Because appeals are not all equal, favoritism will show up at the margins. By analogy, a biased referee can change the outcome in close games, but in many instances the outcome will be the same, and the better team will win. Assuming the sample size is large enough, we can infer that Newman leans in favor of patentees, and Lourie leans in favor of patent challengers relative to the other judges.

      1. 2.2.1

        NS II

        Given how much I ride you, let me give you credit here for your perspicacity.

        Kudos.

  12. 1

    1) did you include enough caveats there?

    2) no, this “mere ten percent at most” difference does NOT excuse the lack of listing of judges on CONTESTED decisions.

    In view of 2) and in view of the items for which will generate the most interesting and pointed conversations on this blog, you would need to winnow out those decisions that would hold no controversial views.

    When you properly account for a more meaningful population of cases, I would bet you that the differences in opinions will be FAR greater than ten percent.

Comments are closed.