Quote from the paper cited in the post describing this algorithm: > These algorithms do not perform well with adversarial streams, but we have mathematically a

Quote from the paper cited in the post describing this algorithm: > These algorithms do not perform well with adversarial streams, but we have mathematically analyzed the 1 unit memory algorithm and shown fast approach and stability properties for stochastic streams Your criticism is really pretty strident for saying something the authors probably completely understand and agree with.