Tournaments, \(\bar\omega \)-criticality, and directed paths
The formal statements of the digraph-theory library, for mathematicians

3.2 \(k = 3\): infinitely many, and Question 5.9 fails

Theorem conjecture_5_10_at_k3 (N : nat) :
  exists T : tournament, kcritical 3 T /\ (N < #|T|)%N.

Source: theories/applications/unified.v:33

Decoded. First statement: for every \(N\) there is a 3-critical tournament on more than \(N\) vertices (the witness is \(\mathrm{AC}_n\) itself, \(n = 2\max (2,N){+}3\)). Second: moreover the witness can be chosen so that its only subtournament with \(\bar\omega \ge 3\) is the whole tournament — formally, any vertex subset \(S\) whose induced subtournament has \(\bar\omega \ge 3\) must be all of \(V\) ([set: T] is the full set). Hence no function \(\ell (3)\) as in Question 5.9 can exist: witnesses have unbounded size but no proper witnessing subtournament at all.

Theorem question_5_9_fails_at_k3 (L : nat) :
  exists T : tournament,
    [/\ kcritical 3 T, (L < #|T|)%N
      & forall S : {set T}, (3 <= ω̄(sub_tournament S))%N -> S = [set: T]].

Source: theories/applications/unified.v:47