Shadow Fiend

Shadow Fiend has the lowest Divine win rate of any hero in this dataset, at 40.7%. It is also played roughly three hundred thousand times a patch. Those two facts sitting together are the most interesting thing on this page.

The curve never looks healthy. It peaks at 46.0% in Guardian — one of the lowest peaks here, though not the lowest — and falls from there to 40.7%, a drop of 4.9 points on top of a baseline that was already five points underwater. There is no bracket in this sample where picking Shadow Fiend is anywhere near an even proposition, and the gap widens the further up you go.

The matchup table does not rescue it. Six favourable results against eight unfavourable, and the favourable ones are narrow in character: Medusa at 64.0%, Undying at 58.0%, Void Spirit at 57.1%, Razor at 56.7%, Underlord at 55.4%, Disruptor at 54.8%. Against the eight that beat it — Mirana at 38.8%, Monkey King at 39.8%, Phoenix at 40.6%, Batrider, Troll Warlord, Earth Spirit, Sniper and Spirit Breaker — the losses are heavier and better evidenced. Phoenix rests on 293 games and Sniper on 353, where most of the wins rest on 114 to 249.

So why is it picked so often? The numbers cannot answer that, and it would be dishonest to pretend otherwise. What they can do is rule out one explanation. A hero that is merely difficult should improve as the players get better, the way Enigma and Shadow Demon do. Shadow Fiend gets worse. Whatever keeps it in three hundred thousand games a patch, it is not a skill ceiling that eventually pays out.

The most useful thing this page offers is probably the matchup list rather than the verdict. If the enemy draft contains Medusa or Undying, the numbers are genuinely good. If it contains Phoenix or Sniper, they are not, and those are the two results here with enough games behind them to lean on.