It could be made more functional by avoiding the mutation of acc. The effects are contained within judgeCircle so it's not a big deal, but it feels like if you're going to mutate the accumulator, you might as well just use an imperative loop.

I also preferred to be explicit about the final check. I find the intent of !h && !v isn't quite as clear as h == 0 && v == 0;.

I ended up with:

const judgeCircle3 = moves => {
  const [hori, vert] = [...moves].reduce(([h, v], move) => {
    switch(move) {
      case "U":
        return [h, v + 1];

      case "D":
        return [h, v - 1];

      case "R":
        return [h + 1, v];

      case "L":
        return [h - 1, v];
    }
    
  }, [0, 0]);

  return hori == 0 && vert == 0;
};

I also got rid of the default since it didn't seem to be doing anything. You could have done error handling in there (and should in most cases), but if it's a challenge with predefined input, that's probably not necessary unless it's part of the challenge.

It could be made more functional by avoiding the mutation of acc. The effects are contained within judgeCircle so it's not a big deal, but it feels like if you're going to mutate the accumulator, you might as well just use an imperative loop.

I also preferred to be explicit about the final check. I find the intent of !h && !v isn't quite as clear as h == 0 && v == 0;.

I ended up with:

const judgeCircle3 = moves => {
  const [hori, vert] = [...moves].reduce(([h, v], move) => {
    switch(move) {
      case "U":
        return [h, v + 1];

      case "D":
        return [h, v - 1];

      case "R":
        return [h + 1, v];

      case "L":
        return [h - 1, v];
    }
    
  }, [0, 0]);

  return hori == 0 && vert == 0;
};

The need for the switch here is unfortunate, but the only other alternative I could think of was some mess where a you'd do a lookup on a map which returned a function that returned a "altered" accumulator state.

I also got rid of the default since it didn't seem to be doing anything. You could have done error handling in there (and should in most cases), but if it's a challenge with predefined input, that's probably not necessary unless it's part of the challenge.