New way of playing odds chess on Lichess

[GambitCat]

You may have heard of the Lichess Leela odds chess bots that have been drawing attention of the chess community up to Nakamura himself. I am developing a Lichess bot that is sandboxing ideas to bridge the A/B & MCTS/Puct worlds for the better &, as part of that private project, I have added a chess odds module dedicated to human players. It is built around Leela hybrid search contempt code with dynamically "mutating" settings & the ability to summon A/B helpers when needed, as opposed to the other modules of the bot, where the A/B helpers are always active. FRC castlings are not supported by the odds module, only standard castlings.

If you are familiar with Leela odds (or not) & willing to experiment this offspring, you're welcome. It's not intended to be 24/7 online & can play just one game at a time, so game duration is capped to Rapid. It is accepting only casual 'from position' challenges from humans, 60s+1s min: https://lichess.org/@/LeelaMultiPoss. Challenge & enter a starting fen, time & chosen side. The conditions to trigger the odds module at startup are:

1) No earlier move (no move odds)
2) Minimum of 2 pieces along with the King on the starting row
3) All pawns & Kings are sitting on their standard starting squares
4) The bot must have a material disadvantage

If any of the above conditions is not met, the bot will use standard non-odds engines

Playing chess odds is recognized as beneficial for training. Standard engines have very poor performances when playing with an initial disadvantage such as starting w/o queen, hence the need of specialized code & data to offer an enjoyable experience.

Feedback on your experience against the bot is welcome !