AI Xiangqi handicap challenge | Chinese Chess

Play Red against AI that gives up pieces or opening moves. Choose a handicap level and practise converting your advantage for Chinese chess (xiangqi).

In handicap chess the engine gives up pieces or opening moves. Play Red, choose a level, and develop your ability to convert an advantage. This page is built for Chinese chess (xiangqi).

Author: Sachess Development Team · Updated: 2026-11-18 · 0 Sources

Highlights

  • AI thinks for a fixed 2 seconds per move, overriding global depth/time preferences. Device performance can still affect playing strength.
  • A campaign win unlocks the next level; a draw does not. Free play is practice and never counts toward the ranking. Level N has a base of N × 1000 points: a draw earns base + 200, a win earns base + 1000. Your best result is retained, not added to previous results.

Steps

01

Start at a level you can challenge, review mistakes after each game, and move on when you win consistently. Handicap games complement normal games and endgame practice.

Details

Rules and scoring

A campaign win unlocks the next level; a draw does not. Free play is practice and never counts toward the ranking. Level N has a base of N × 1000 points: a draw earns base + 200, a win earns base + 1000. Your best result is retained, not added to previous results.

Asian rules: after three complete cycles, a user must vary a perpetual check or prohibited chase. Other threefold repetitions and 60 moves without a capture are drawn; alternating check and chase is allowed.

How to improve

Start at a level you can challenge, review mistakes after each game, and move on when you win consistently. Handicap games complement normal games and endgame practice.

Why practise handicap Xiangqi against AI?

Against a stronger engine, an even game can become defensive before you have time to practise a plan. A material or opening-move advantage gives you room to work on calculation, coordination and converting a lead. The aim is to understand how you keep an advantage and where you give it away, rather than merely collecting wins.

What regular practice can develop

Work on spotting threats, choosing exchanges, coordinating rooks, horses and cannons, and keeping your general safe. A fixed two-second search lets you compare repeated attempts on the same device, although hardware still affects strength. Challenge points are not an Elo rating or a promise of improvement; use reviewed games and normal play to judge progress.

Learn from the AI in the middlegame

Before replying, ask what the AI threatens: an active file, a rook-and-cannon attack, a horse manoeuvre or a forcing check? Check its checks, captures and direct threats, then choose your own plan. After the game, find the first point where you lost the initiative and compare your idea with the actual counterplay. Treat engine moves as material to investigate, not unquestionable model answers.

Convert your lead in the endgame

Avoid checking without a plan. Secure your general, reduce counterplay and decide which exchanges help you. Keep enough coordinated material to win instead of entering an ending you cannot convert. Watch pawn advances, rook activity and stalemate threats. A draw is useful evidence: review whether tactics, exchanges or an unclear winning plan cost the advantage.

Use handicap training carefully

Choose a level where you can compete seriously and sometimes win. Different material combinations teach different skills; the 15-level order and points are not a calibrated strength scale. Handicap openings differ from normal openings, so do not copy their plans blindly. Respect counterplay and repetition rules, and alternate this practice with even games and independent endgame studies.

A practical routine for each game

Pick one goal before starting, such as checking captures or making favourable exchanges. Look at the opponent’s threats and compare two or three candidate moves before playing. Free practice allows takebacks; scored campaigns do not. After the game, click move pairs or use the replay slider, identify one important mistake and download the record. Repeat the level and check whether you avoid that mistake.

What this page provides

Play Red without signing in, with 15 material or opening-move handicaps, a sequential campaign and free practice. Games and progress save automatically in this browser, unfinished games can be resumed, and a PGN download includes the starting position for your own backup. Replay completed games and optionally inspect the current AI search depth, Red-perspective evaluation and principal variation. Online campaign scores enter rankings only after server verification. Local progress does not sync across devices; download records before clearing browser data. A PGN is a game backup, not a campaign-progress backup. Online verification checks legal play and scoring, not that every Black move came from the AI. The leaderboard is for recreational practice and does not certify competitive play.

Read the handicap and campaign guide

FAQ

Can I play without signing in?

Yes. This version saves progress in this browser. Clearing browser data removes that local progress.

How long does the AI think?

The browser engine uses a fixed 2 seconds per move. Device performance can still affect playing strength.

Does a draw unlock the next level?

No. Only a campaign win unlocks the next level. A draw can receive the configured local score.

Sources

Related Pages