The reason why a pure ELO would be problematic for Splinterlands is because a game of Splinterlands is not pure skill. Card collections (the pay to win component) is important and there is also a heavy dose of randomness.

If Splinterlands was a game where everyone had access to the same cards. And then once both players submitted a team, the computer did one million simulations to decide who which team would win on average. In that scenario, then I agree a pure ELO system would make sense.

But there are lots of cases where the winner of a game is 50/50. And situations where one player chose the much stronger team but due to bad luck (like lots of misses), they end of losing. Chess doesn't have to deal with such scenarios.