BLACKJACKRANKED

Basically Elo for blackjack.

The game itself: 6-deck shoe with a real cut card (75-83% penetration). Persistent bankroll across sessions. An actual discard tray you can track. Dealer stands on soft 17, 3:2 blackjack, late surrender, DAS allowed. As close to sitting at a real table as I could make it without the cocktail waitress.

The rating system is called VAR (Variance-Adjusted Rating). This is where most of those 3 months went. I didn't want to just slap a number on a blackjack game, so I went deep into the research. Here's how it actually works:

Every decision you make gets compared against the composition-dependent optimal play for the exact remaining shoe state. Not basic strategy charts. Not index numbers derived from a counting system. The actual mathematically optimal move given precisely which cards are left at that moment.

The formula is built on Peter Griffin's research into what actually drives a counter's edge. Griffin proved that in 6-deck S17, bet sizing contributes roughly 1.0-1.5% of total edge while play deviations beyond basic strategy contribute about 0.10-0.15%. That's a 5-6x difference. So VAR weights them accordingly using a weighted geometric mean:

Play quality (~13% of your rating weight): every hit, stand, double, split, surrender, and insurance decision scored against composition-dependent optimal play.

Bet sizing quality (~87% of your rating weight): this evaluates three things:

Positive-edge pressure. When the shoe favors you, did you bet bigger? It doesn't demand exact Kelly-to-the-dollar precision. There's a tolerance zone around a practical bet center based on your bankroll and table limits. Human-realistic sizing within that zone gets full credit.

Negative-edge discipline. When the shoe is against you, did you stay near the minimum? Overbetting weak shoes is one of the most expensive mistakes a counter can make and the formula penalizes it accordingly.

Bet ramp quality. Did your bet sequence across the shoe make sense as a disciplined progression? Did stronger opportunities receive meaningfully more pressure than weaker ones? Did you preserve headroom to press when the shoe got really good instead of firing too big too early?

The geometric mean is what makes the whole thing work. Play and bet multiply, they don't add. Perfect play with zero bet sizing skill produces a low rating, not a 50% rating. Perfect bet sizing with random play decisions also scores poorly. You need both. This matches reality because a counter's actual edge is the product of playing correctly AND betting correctly.

Win with sloppy play? Rating drops. Lose with perfect decisions? Rating goes up. The leaderboard ranks purely by decision quality. You can't just run hot and claim you're good.

This is why it's system agnostic. It doesn't care if you use Hi-Lo, KO, Omega II, Wong Halves, or no named system at all. It measures whether your actual decisions align with the true mathematical edge.