Plenty of people want to play nardi against a bot, to learn the rules and the moves without making their mistakes in front of a live opponent. Let us be straight about how it works on tamashebi.cc: tables in the nardi room are opened by real players and joined by real players, and there is no button yet to pick a bot as your opponent in nardi. The bot does exist in the game engine, at three levels of difficulty, and it plays on the site at particular moments. Below we explain when it steps in, how the levels differ, why it cannot cheat, and how to set up practice that gives you the same benefit without one.
Why practising against a bot is worth it
Practising against a bot has three advantages: nobody is waiting for you, nobody minds your mistakes, and you see the same situations many times over. In nardi you can get all of that against people too, if you set the game up right: a free table with no coins, a one-point match, and a friend you invite to a password table. In return a live player teaches you what a bot cannot: human mistakes, psychology, and the moves a simple formula would never consider.
Three levels: what changes
The nardi bot on tamashebi.cc plays at exactly three levels. They differ in how much the bot thinks before it chooses a move:
- Easy plays any legal move, chosen at random. It plays both dice whenever that is possible, but it has no strategy.
- Normal scores every position a turn can end in: the pips left, how many blots are exposed and how close they are to danger, how many points are closed (home points especially), how long the prime is, and who is on the bar. It picks the move with the best score.
- Hard does the same scoring, then looks one roll ahead for its few best candidates: for each of the opponent's 21 possible rolls it works out the opponent's best reply, and it chooses the move that comes out best on average.
Where does it play on the site? Most often when your time for a turn runs out. If the time you chose (20, 30, 45 or 60 seconds) expires, the server finishes the rest of the turn for you using the Normal bot's logic. The seat stays yours and you play your next turn yourself. Watching such a move is useful, by the way: you see what the formula would have picked in the same position.
Does the bot cheat?
No, and this is guaranteed in the code. The server rolls the dice with a cryptographic random number source, the same source for everyone, bot or human; the client, meaning your browser, has no influence on the roll at all. The bot does not know what comes next: even the Hard level weighs all 21 possible rolls by their probability rather than one "real" one. And finally, the bot's moves come from the same search the server uses to check every move, with the rule that you must use as many dice as possible, the larger-die rule, and the Georgian rule about hitting in your own home. It simply cannot make a move outside the rules.
Practice drills
Below are three drills that make your game grow fastest. Each works against a live opponent, especially in a free one-point match where the result means nothing and you can experiment.
The prime drill
Goal: build three or four closed points in a row across your yard and home, starting with the 5, 4 and 7 points. For the whole game, ask yourself one question: "After this move, does a new point get closed, or do my chances of closing one next turn go up?" For example, 3-1 on the first move: 8 to 5 and 6 to 5. The 5 point is closed, and together with the 6 you already have a two-point block. At the end of the game, count how many times you held a block of three points or more.
The back checker drill
Your two back checkers start on point 24 and have the longest way to go. The drill: in every game, either bring both back checkers out past your opponent's block, or keep them together in their home as a closed point. If one is left alone and gets hit, count how many turns it took to enter it again. You quickly learn that running a back checker from 24 to 13 on a 6-5 is often the best play, and that a checker stuck in your opponent's home is also a koks risk.
The bearing-off drill
Many games are lost at the end of the race to a wasted move inside home or a needless blot. The drill: once contact is over, bear off as many checkers as you can with every die; while an opposing checker is still on the bar or in your home, safety first, speed second. Two things on the site help: a checker that can be borne off goes to the tray with a single tap, and the undo button takes back your last step before you confirm the turn. The rules of bearing off are set out precisely in a separate article.
Keep a practice log
Practice works when you measure it. After every game, take one minute to note three things: how many times you were hit and on which point you left the blot; what both pip counts were when contact ended; and how the game finished, with oini (a single win, 1 point), mars (2) or koks (3). After ten or fifteen games a pattern shows: some players keep leaving the same point exposed, others are ahead in the race and still lose because they waste time bearing off. That pattern tells you which drill to work on next.
When should I move on to live opponents?
In nardi the question is a little different: on tamashebi.cc your very first game will be against a live opponent. So start in a way that makes mistakes cost nothing:
- Create a free table: set the stake to "Free", so no Domino Coin is involved.
- Choose 1 point: one game, a quick finish, then a new one.
- Choose 45 or 60 seconds per move, so you have time to count.
- To practise with a friend, set visibility to "Password" and send them the link.
- Once the logic of the moves becomes habit, move up to 3 or 5 points and 20–30 seconds: managing the score then becomes part of the training.
If you specifically want to play a bot, tamashebi.cc offers that mode in domino. In nardi, the best "trainer" today is a live opponent and your own observation.
Frequently asked questions
- Can I choose a bot as my opponent in nardi?
- Not yet. Nardi tables are played between live players. The bot plays your turn only when your time runs out, and the seat stays yours.
- Does the bot know which dice will come?
- No. The server rolls the dice with cryptographic randomness. Even the Hard level averages over every possible roll and does not know the next number.
- Which level plays for me if my time runs out?
- Normal: it scores every position the turn can end in and picks the best. The move follows the rules, and your next turn is yours again.
- Which drill should I start with?
- The prime drill. It teaches fastest which points are valuable and why a blot is dangerous. Then move on to the back checkers, and finally to bearing off, which finishes the games you have already won in the race.
- Play nardi online
- Nardi strategy: tips
- Bearing off in nardi
- Playing nardi on a phone
- Multi-point nardi matches
Nardi on tamashebi.cc is free to play. Some tables use Domino Coin, a virtual in-game coin: it is never money, cannot be bought or cashed out, and means nothing outside the site. The game is for fun and for skill.
The games on Tamashebi.cc are played in a Georgian-language interface. They are free and use a virtual coin, never real money.