Document ranked without size limit

README.md

@@ -25,8 +25,8 @@ Decoder-only models (50K–250M parameters, custom architectures welcome) play t
 side on CPU. Two ways to play:
 
 - **Match** (friendly): pick any 2+ models and the seed. Nothing is recorded.
-- **Ranked**: press Play and the arena picks up to 4 models **at random** from its pool, all within 20M parameters of
-  each other (models with 100M+ parameters can all play each other). The result updates a public **Elo** leaderboard.
+- **Ranked**: press Play and the arena picks up to 4 models **at random** from its pool, of any size. The result updates
+  a public **Elo** leaderboard.
 
 ## How a model plays
 
@@ -59,9 +59,10 @@ Each protocol has its own leaderboard.
 ## Elo
 
 Only the Ranked tab changes Elo. The arena picks the players at random from the suggested models (models with fewer
-ranked games are more likely to be picked), all within 20M parameters of each other, and uses a random seed (when
-every picked model has 100M+ parameters, the 20M gap doesn't apply: big models can all play each other). Nobody
-chooses who plays ranked, so Elo can't be farmed by pairing a model with weak opponents. The match runs on the server
+ranked games are more likely to be picked), of any size, and uses a random seed. Nobody chooses who plays ranked, so
+Elo can't be farmed by pairing a model with weak opponents. Any model can meet any other, so every rating sits on one
+comparable scale (which is what the Elo vs. parameters chart needs); Elo weighs each win by the opponent's rating, so
+beating a much weaker model earns almost nothing once ratings have settled. The match runs on the server
 in the background: it finishes and counts even if the viewer leaves, and only one ranked match runs at a time
 (pressing Play while one is running lets you watch it).
 
@@ -80,8 +81,8 @@ Every ranked match (seed, model commit SHAs, scores, Elo before/after) is stored
 | `MAX_PARAMS` | `250000000` | Parameter limit |
 | `MIN_PARAMS` | `50000` | Smallest model allowed (smaller ones are rejected and removed from the leaderboard) |
 | `MAX_PIECES` | `500` | Piece cap per game |
-| `MAX_PARAM_GAP` | `20000000` | Largest size difference between models picked for a ranked match |
-| `LARGE_FROM` | `100000000` | Models this size or bigger can play ranked against any other model this size (no gap) |
+| `MAX_PARAM_GAP` | `0` | Optional largest size difference between models picked for a ranked match (0 = no limit) |
+| `LARGE_FROM` | `100000000` | Only with a gap: models this size or bigger can all play each other |
 | `ALLOW_REMOTE_CODE` | `1` | Allow models with custom code (`trust_remote_code`) |
 | `TORCH_THREADS` | `2` | CPU threads for inference |