littleply / games & analysis

Tournament field notes / November 10, 2026

Six wins. A difficult field. Useful context.

LittlePly at the 2nd ChessBots Mega Team Battle: rated standard chess, 5+0, playing for Neural Sparks.

The final record

28th of 81 players, with 12 arena points. The official game export records 6 wins, 0 draws and 17 losses. One loss was a no-start, not a played defeat: 22 games contained moves, producing six wins and 16 losses on the board. All six wins ended by resignation; all 16 played losses ended in checkmate.

The two-hour arena ran on November 10, 2026, from 17:30 to 19:30 UTC (9:30–11:30 a.m. Pacific). Arena points are distinct from game score or a rating estimate.

The six wins

All six wins were against human accounts, across five distinct opponents. Two were against the same GM account, DJ_Haubi. Those are interesting games to replay, not evidence that LittlePly has grandmaster strength.

LittlePly started with five minutes in each of these games. In four of the six games, the opponent used berserk, which halves their starting clock to 2:30 in this 5+0 arena. Both GM wins had that clock advantage. No increment applied. Opponent ratings are Lichess ratings at each game, not model strength measurements.

What this tells us, and what it does not

The result captures the deployed CM-14 agent with its existing CPU harness and deterministic assistance. The new shared-GPU search and neural pondering work was separate. Do not attribute these tournament results to that new harness or to the model alone.

The wins illustrate why an imperfect agent can still be an interesting opponent: there are real games to inspect, alongside a substantial number of losses. They do not establish an equal-clock GM result, a general win rate against humans, an Elo estimate, or a causal benefit from imitation learning. Pairings, clock choices and opponents matter.

Play the deployed agent on Lichess · Learn how to interpret evaluations

Sources and reproducibility

Final standings and all 23 game records were retrieved from the official Lichess API after the event. The no-start remains in the official total rather than being silently dropped. No new engine analysis or claims about the best move are attached to this recap.