Yeah, it’s not just sample vs variance, there’s also the mathematical law of big numbers
The higher the numbers getting divided, the lower the result is
Like, if you go 10-5, that’s 10/15 = 66,66%
If you go 20-15, that’s 20/35 = 57,14%
If you go 30-25, that’s 30/55 = 54,54%
Note that, even though the win-loss is constant (5 wins more than losses in all cases), the winrate percentage keeps falling
That + sample/variance laws are to blame for such winrates with low samples
53% in 250 games is meaningful. It’s a tier 1 deck
53% in 25 games is a disaster.