$tahirl_
[01]
LabBlogPuzzleMCP↗AboutSupport🇹🇷 TR
tahirl
LabBlogAboutSupportMCP ↗

© 2026 tahirl. All rights reserved

Back to Blog

TurkGambit: A Lichess Bot That Plays the Turkish Gambit

Published on April 1, 2026•4 min read
botTurklichesschess

Related Section

TurkGambit

TurkGambit: A Lichess Bot That Plays the Turkish Gambit

I've played chess since I was a kid. Never competitively, no coach, no rating grind. But I always enjoyed it and I never stopped.

At some point I came across the Turkish Gambit through a video by FM Selim Gürcan, who introduced the opening to the Turkish chess community and gave it its name. In older literature it appears as the Elephant Gambit, but nobody ever really claimed it or gave it serious attention. That changed. It's now the Turkish Gambit, with its own variations like the Crazy Turkish Gambit and the White Turkish Gambit.

The moves: 1.e4 e5 2.Nf3 d5. Black offers a pawn on move two. It looks like a blunder but there's logic behind it: the side that accepts gets their development disrupted, black gains tempo, and the position becomes sharp and hard to calculate. If you know the theory and your opponent doesn't, the rating gap closes fast.

I tried it, my rating went up, and I started promoting it to anyone who'd listen. After wins I'd message opponents:

Did you like that opening?

That's the Turkish Gambit. Look it up.

Some tried it, some didn't care. I kept going because it works.

Why I Built It

One day I was watching FM Yasin Emrah Yağız's livestream and he said something along the lines of

Why don't we have our own chess platform, can't someone build one?

I couldn't build a platform on my own, but I could build a Lichess bot. And my favorite opening is the Turkish Gambit. So I did.

How It Plays

Most Lichess bots run Stockfish and play the strongest move. TurkGambit doesn't. I designed it around a different idea, inspired by Mikhail Tal:

You don't have to prove a sacrifice is correct. You just have to prove it's interesting.

When Stockfish suggests the best move, TurkGambit looks for a sacrifice with enough compensation and goes for that instead. Pawns, exchanges, queens if the position allows it. Sometimes the sacrifice is brilliant, sometimes it's questionable, and sometimes it genuinely drops a piece. It's still learning.

You're not playing against a cold engine. You're playing against a character that takes risks. Sometimes you can't tell if it blundered or sacrificed on purpose.

Not Just the Turkish Gambit

The name is TurkGambit, but its repertoire goes beyond that. It's been trained on over 20 sacrifice-based openings: King's Gambit, Evans Gambit, Danish Gambit, Morra Gambit, Budapest Gambit, Benko Gambit, Halloween Gambit, and more. It picks one based on the opponent's opening.

But when the position allows for the Turkish Gambit, it always chooses that first.

Built for Humans

TurkGambit plays unrated games only. If it played rated it would settle into a bracket and only face similar-level opponents. I didn't want that. 800 or 2000, anyone can challenge it.

It adjusts its behavior based on the opponent's rating. Against lower-rated players it plays more loosely, makes deliberate mistakes, goes for more sacrifices. Against stronger players it becomes more selective. The idea is to make every game enjoyable, not to crush anyone. It only plays against humans.

Learning

TurkGambit learns from every game. It records positions, tracks which sacrifices worked, and uses that data the next time it encounters a similar position.

It's early days though. Expect some bad sacrifices. It knows the fundamentals but its judgment only matures with more games. As of April 2026 it has accumulated 14,545 position patterns, 2,188 of which are specific to Turkish Gambit positions. Almost all from training data. It still has a lot of growing to do.

Technical Setup

The bot runs on my remote VPS with limited resources:

CPUAMD EPYC, 1.5 cores
RAM2 GB
Stockfish Threads2
Hash Table256 MB
StorageRedis
Uptime24/7

For comparison, serious Lichess bots run on 8, 16, or 32 cores. More threads mean deeper calculation. With only 2 threads, sacrifice evaluation is harder since understanding whether a sacrifice pays off can require looking 15 to 20 moves ahead.

In theory it can handle up to 9 simultaneous games, though that hasn't been fully tested yet.

Challenge It

The project is new and any feedback helps. If the bot freezes, behaves oddly, or does something unexpected,

let me know
EmailTelegramXGitHub
.

The bot is active on Lichess as Turk_Gambit, accepting challenges 24/7. Be ready for sacrifices.

Comments

No comments yet. Be the first to comment!

Back to Blog