This is exactly the kind of behavior that you'd see with reward hacking in reinforcement learning.

I think humans are pretty good at finding the optimum strategy. It's only really recently that bots have overtaken us in a lot of highly environmentally complex games.

they should have made a rule that a counter attack point wouldn't count if the attack was initiated within the first say 3-5 seconds after the signal.

> counter attack point

The rule was primarily aimed at epee though, where there is no such thing as "counter attack point", since there's no right-of-way rules.

Unless I'm gravely misunderstanding your idea.

Just paint 2 lines midway to the center - whoever crosses first gets the attacker designation and the 5 second window during which his opponent's point if such scored would be good only to break the attack and wouldn't be counted. Add to that that the fencers in many situations wouldn't be completely sure who crossed first.

So, the very first attack gets better protection on their attack than a foil fencer? And simultaneous attacks just look like sabre?

And what about subsequent attacks after that? They're just normal where the threat of the counter attack drags us back to fencers not doing anything for 2 minutes?

Or what if play is halted and restarted somewhere other than the centre en-garde lines?

Ok yea but then you've just turned epee into foil

Which, as a foil fencer in a prior life, I am totally on board with. But I don't see it happening lol.