Been scanning CS2 round handicap lines systematically for a while now, and there's one mistake that shows up over and over in people's spreadsheets: using blended round-win rate instead of map-specific, side-adjusted round-win rate.

Here's the difference in practice:

Say Team A wins 54% of rounds overall across their last 20 maps. Looks solid. But if you break it down by map, they're winning 61% of rounds on Mirage (a map they've played 9 times) and only 47% on Inferno (3 times). If Inferno is the map in play tonight, using their 54% blended number massively overstates their edge.

Same goes for side. A team can have a strong overall round-win rate that's almost entirely driven by CT-side dominance on a specific map. If they're on the T side to start, that blended number is misleading you before the match even loads.

The fix is simple but tedious: always condition your round-win rate on (a) the specific map being played and (b) the side they're starting on, using their last 6-10 games in that exact context. If you don't have 6 games of sample in that context, flag it as low-confidence rather than trusting the blended average.

This one change alone cleaned up a meaningful chunk of false-positive edges in my own tracking. Curious if others have found similar patterns — what's your minimum sample size before you trust a map-specific split?