Not really. It's true that few big winners have used Lisp, but that would only disprove the thesis if there were also companies who tried using Lisp and failed. AFAICT, no one is even doing the experiment.
I actually know of two notable counterexamples: Barefoot Networks has an internal design tool written in Common Lisp that is a significant source of competitive advantage for them. Also Orbitz.
Keep in mind the at that Beating the Averages is from 2001 (revised in 2003). This means that it predates Java 1.4 and obviously C++11 (let alone C++14, 17, etc)
Most languages have the features of CL that made it so useful. Even Java has first class functions, lambdas, partial application, async IO, etc. Java even has a repl now. The only things left afaict are macros (non-hygenic in CLtL2) and code-as-data/eval (a security hole).
Aside from pulling from functional languages, Java also learned Python's 'with' using try-with-resources. Meanwhile the tooling of Java went from strength to strength and it's a serious blub factor for people who haven't used Java.
Beating the Averages was fairly spot on at the time it was written, but since then it's lost it's power as features of blub languages was merged into existing languages.
If not commercial Lisp, what you think about using Racket for both the IDE and so they can pre-scale by giving HtDP to good developers who want to learn Scheme?
Not really. It's true that few big winners have used Lisp, but that would only disprove the thesis if there were also companies who tried using Lisp and failed. AFAICT, no one is even doing the experiment.
I actually know of two notable counterexamples: Barefoot Networks has an internal design tool written in Common Lisp that is a significant source of competitive advantage for them. Also Orbitz.