Counter Intelligence

No. This page aggregates what happened in 90M rated games across 200 positions. It's a practical counter dataset: "What tends to score best for Black here?" not "What is theoretically best?".

Move 3 is an early decision point where many openings branch. Our 200 positions include 145 against 1.e4, 52 against 1.d4, and 3 against flank openings—each with sufficient sample sizes for stable comparisons.

Outcomes are shown from Black's perspective for the response move (because Black is to play 3...). Win = Black wins, Loss = White wins. For example, Caro-Kann Defense: Hillbilly Attack after 3...Bg4 shows 68.4% Black wins.

Positions are keyed by the move prefix (the exact moves leading to the position), not just the opening name. That reduces "same name, different position" mixing.

Prefer higher n and tighter confidence intervals. Our most-played position (French Defense: Knight Variation with 10M games) is far more reliable than positions near the 50-game minimum threshold.

Because the page is position-first. The same named opening can be reached via different move orders (different positions). This page keeps them separate so the results don't get blended incorrectly.

Edge is a conservative advantage estimate for Black: (Black win% p5) - (White win% p95). In our data, Edge ranges from -24.2 (Queen's Pawn Game: Chigorin Variation) to +36.2 (Caro-Kann Defense: Hillbilly Attack), with an average of -1.5 across all 200 positions.

Because low-sample buckets are hidden. Filtering slices the dataset; if a move's sample size is too small in that slice, it's excluded to reduce noise. When filtered, expect fewer rows—but higher trust per row.

Yes. Many openings have different practical results across rating bands. Of our 200 positions, performance varies significantly by skill level—use the Rating filter to see counters that actually score for players at your level.

Yes. Bullet rewards simplicity and traps; rapid/classical rewards precision and endgame conversion. Use the Time filter—some counters gain or lose edge depending on the clock.

Use sample size as a tiebreaker. If Edge is close, prefer the move with larger n (more stable). Our largest sample is French Defense: Knight Variation with 10M games—that's far more reliable than positions near the minimum.

Search the opening name, then play the top 3... response with the highest Edge and strong sample size. For example, Caro-Kann Defense: Hillbilly Attack shows 3...Bg4 with +36.2% Edge from 1.9K games.

This page covers 200 Black move-3 (3...) responses. If you want move-by-move recommendations for both sides, use the main TrueElo analyzer and the per-opening stats pages.

Use this page as a "first punch" tool: pick your preferred 3... responses from our 200 positions. 83 (42%) are true counters where Black is favored; 117 are equalizers for minimizing White's advantage. Then go deeper on per-opening stats pages for follow-ups.

Yes. Rating and time control filters are encoded in the URL query string, so you can copy/paste the URL after setting filters and someone else will see the same slice.

This is computed from 2.1B rated Lichess games (2023-12 to 2025-11). TrueElo aggregates outcomes and confidence bounds so you can compare moves with different sample sizes.

This page was last regenerated January 2026 from the underlying stats dataset. The timestamp at the top shows the last refresh; data is updated every few months.

No. This is a practical scoreboard based on 90M games, not a proof of correctness. Use it to choose moves that score in real games; use theory (or an engine) when you want to understand long-term soundness.

We cover 200 positions (145 against 1.e4, 52 against 1.d4, 3 flank). Try searching broader labels or the move prefix. If it's still missing, there may not be enough games for stable move-3 comparisons (minimum 50 required).