WHAT YOU WILL BE ABLE TO DO
Take three ideas to the table.
- Compare useful draws under explicit assumptions rather than declaring a universal best discard.
- Keep tile efficiency, the scoring route and opponent risk as separate checks.
- Use opponent-specific Riichi safety without treating a probability clue as a guarantee.
A HAND YOU CAN READ
Choose a discard you can explain
An early-hand Riichi shape exercise, not a safety guarantee
Table setup. Your closed hand has just drawn its fourteenth tile. No opponent is threatening, and your narrow goal is to retain a ready hand with the red-dragon yaku. Compare discarding 9 dots, 3 dots or a red dragon. Real table danger can override this exercise.
1. Fourteen tiles: protect the useful blocks
14 tiles · before discard1 · Sequence
Keep a finished group.
2 · Sequence
Another finished group.
3 · Dragon triplet
A real Riichi yaku is already present.
4 · Pair
Preserve the pair.
5 · Two-sided wait
1 or 4 dots can complete the last sequence.
Extra tile
The 9 dots does not connect to 2–3.
You have two sequences, a red-dragon triplet, a pair and 2–3 dots. The extra 9 dots does not join those groups. Among the three choices in this example, discard 9 dots: it leaves the useful shape intact. Discarding 3 dots breaks the last sequence shape; discarding a red dragon breaks both the triplet and the guaranteed yaku.
2. Thirteen tiles after discarding 9 dots
13 tiles · between turns1 · Sequence
Keep a finished group.
2 · Sequence
Another finished group.
3 · Dragon triplet
A real Riichi yaku is already present.
4 · Pair
Preserve the pair.
5 · Two-sided wait
1 or 4 dots can complete the last sequence.
Now either 1 dots or 4 dots completes the final sequence. There are at most four copies of each, so this wait starts from eight potentially available copies before visible tiles are subtracted. “Eight” does not mean eight are definitely in the live wall; some may be in opponents’ hands or the dead wall.
3. Visible tiles change the opportunity
13 tiles · between turns1 · Sequence
Keep a finished group.
2 · Sequence
Another finished group.
3 · Dragon triplet
A real Riichi yaku is already present.
4 · Pair
Preserve the pair.
5 · Two-sided wait
1 or 4 dots can complete the last sequence.
Already visible outside your hand
Count each physical tile once, including a discard that was claimed into a set.
Suppose three 1 dots and two 4 dots are already in exposed sets or discard piles. Subtract those five known copies from the maximum eight: only one 1 dots and two 4 dots remain unseen, three possible copies in total. Count each physical tile once; a called discard now in an exposed set is not a second copy. This is availability information, not proof a discard is safe.
Remember: Use three checks: what useful draws remain, what makes the hand eligible, and whom the discard might help.
The blocks show how to read the hand, not its order in the game rack. Every example here has no kong; a declared kong adds one physical tile to the usual count.
Read groups before judging individual tiles
Start with completed sets, then your likely pair, then incomplete groups. This helps you see what each candidate discard would break. An isolated 8 dots is a different decision from the same tile in 6–7–8 dots or beside another 8 dots.
Do not force every tile into your first interpretation. For example, 2–3–3–4 bamboo can be viewed as a 2–3–4 sequence plus another 3, but that extra 3 may also help a future pair or overlapping group. Flexible shapes often deserve time before you commit to a call or discard.
Compare the tiles that would help next
A 4–5 dots group can become a sequence with either 3 dots or 6 dots. A 1–2 dots group needs 3 dots. With no other information, the first group has more completion types. That is one reason middle sequences are often easier to develop than edge sequences.
Now bring in the visible tiles. There are four copies of each core tile type. If several 3 dots are already in discards or exposures, that route has fewer remaining copies. Count only information you can see: your own tiles, public discards and exposed sets. You cannot know exactly what is in another player’s closed hand or the remaining wall.
Useful draws are broader than immediate winning tiles. A draw can improve an incomplete group, create a pair or connect two loose tiles. Avoid labeling a hand “bad” only because it cannot finish on the very next draw.
A worked count: more tile types can mean fewer remaining copies
Compare two isolated incomplete groups only: 4–5 dots and 1–2 bamboo. Ignore the rest of the rack, scoring, danger and future shape changes for this small exercise. If no completion tiles are visible anywhere, 4–5 dots has up to eight unaccounted-for immediate completion copies: four 3 dots and four 6 dots. The 1–2 bamboo edge has up to four: the four 3 bamboo.
Now suppose you can see three 3 dots and four 6 dots outside those groups, but no 3 bamboo. The first group has only one unaccounted-for immediate completion copy, while the edge group has four. The broader-looking shape is no longer ahead on this one measure.
“Unaccounted for” does not mean “definitely waiting in the wall.” Other players may hold those tiles, and replacement areas or other game state can make them unavailable. Counting visible copies narrows what is possible; it does not reveal hidden information. Nor does this isolated comparison identify a best discard for a complete rack we have not specified.
Use a decision note instead of a hunch
| Question | Useful answer | Unsupported shortcut |
|---|---|---|
| Shape | “Keeping this 4–5 preserves 3 or 6 as completion types.” | “Middle tiles are always best.” |
| Value | “This keeps the dragon pair that might become my open-hand yaku.” | “A dragon pair already gives me a yaku.” |
| Risk | “This tile is in the riichi player’s own discards, so they cannot ron it.” | “One opponent discarded it, so nobody can win on it.” |
| Uncertainty | “An open call or a new riichi declaration would change my plan.” | “It survived once, so it must always be safe.” |
You do not need a numerical danger score to make a disciplined decision. Be precise about the reason you know, then name the uncertainty that remains.
Keep the scoring plan attached to the shape
In Hong Kong, ask how the proposed hand reaches this table’s three fan. Opening a set removes the concealed-hand fan. Breaking a promising dragon pair may remove a future scoring route, while stubbornly keeping every honor can leave too little flexibility.
In Riichi, name a yaku before calling. If all simples is your route, a retained 1, 9, wind or dragon prevents it. If riichi is your route, opening the hand removes it. A faster shape with no qualifying yaku is not a successful shortcut. The yaku and dora guide explains why bonus value cannot fix that problem.
Reassess when an opponent threatens
An opponent’s riichi declaration tells you that their hand is ready under the declaration rules. Your decision now includes the cost of dealing into that hand. A distant, low-value hand may justify defending; a strong ready hand may justify continued attack. You are comparing your own prospects with risk, not following an unconditional rule to push or fold.
In Riichi, a tile appearing in a particular opponent’s own discards is safe from that opponent’s ron because of furiten. It is not automatically safe against the other two players. Check each relevant opponent separately. Do not transfer that guarantee to a real-world Hong Kong table. This app currently also applies an own-discard and passed-win restriction in its Hong Kong mode; that is an implementation difference, not a universal Hong Kong rule.
Treat clues as clues
Suji reasoning and visible-copy counts can help assess danger, but they do not make every related tile safe. A discarded 4 does not prove that a 1 cannot complete a different kind of wait. Even all four visible copies of a numbered tile do not universally prevent that tile from completing someone’s sequence; they only rule out some holdings.
When no guaranteed safe discard is available, compare the evidence and accept that the result is uncertain. A sensible decision can lose, and a careless one can survive. Review the information available at the time rather than judging the move only by the outcome.
Use a five-question practice routine
- Which two discards am I comparing?
- What useful draws does each preserve?
- What scoring route remains?
- Who might win from the discard?
- What new information would change my choice?
Choose a move, say one reason, and continue. After the hand, revisit a single decision. This is a manageable way to build judgment without pretending a short beginner guide can calculate the perfect move for every table.
MAKE THE DECISION, THEN EXPLAIN IT
Three table situations.
Choose an answer, then compare your reasoning. You can retry any situation or open its full explanation.
0 of 3 situations solved in this visit
SITUATION 1
Count copies, not only tile types
Compare immediate sequence completion only; ignore the rest of the hand, value and danger. For 4–5 dots, you can see three 3 dots and all four 6 dots elsewhere. For 1–2 bamboo, no 3 bamboo is visible. Which group has more unaccounted-for completion copies?
Choose the move you would make.
Read the answer and compare every option
A. 4–5 dots, because two-sided groups always win the comparison — Why this misses the point
Here its eight theoretical copies are reduced by seven visible ones, leaving just one. A shape’s label does not override the visible count.
B. 1–2 bamboo: four copies versus one — Best answer
Correct for this limited measure. Four 3 bamboo are unaccounted for, compared with one 3 dots. Hidden copies are not guaranteed to be in the live wall, and this does not settle a full-hand discard.
C. They are equal because both are incomplete sequences — Why this misses the point
Count the actual completion types and their four-copy limits. The stated visible tiles leave different numbers of possible completions.
Take it to the table: Tile visibility can reverse a simple shape comparison; keep the conclusion limited to what you counted.
SITUATION 2
Safe against whom?
In Riichi, opponent A and opponent B have both declared riichi. Your candidate 7 dots appears in A’s own discards but not B’s. Assume you have no other relevant safety information. What can you conclude?
Choose the move you would make.
Read the answer and compare every option
A. Neither opponent can ron the 7 dots — Why this misses the point
A’s discard history establishes a restriction for A only. It does not tell you whether B can win on 7 dots.
B. It is guaranteed dangerous against B — Why this misses the point
The absence of this tile from B’s discards does not prove B is waiting for it. B’s status remains uncertain from these facts.
C. A cannot ron it; B’s ability to ron is not established — Best answer
Exactly. Own-discard safety is opponent-specific. Evaluate B separately instead of extending A’s guarantee to the entire table.
Take it to the table: Always attach a safety claim to the opponent it actually covers.
SITUATION 3
A faster shape is not the whole decision
You have an open Riichi hand. A proposed change would remove its only possible yaku route, while creating a shape with more immediate completion tiles. Assume no other yaku or special circumstance can qualify the resulting hand. Does the extra speed alone justify calling it a winning plan?
Choose the move you would make.
Read the answer and compare every option
A. Yes, reaching four sets and a pair is enough — Why this misses the point
Under the stated assumptions the new shape has no yaku. More completion tiles cannot make an ineligible hand into a legal winner.
B. No; the plan must retain a qualifying yaku — Best answer
Correct. Tile efficiency is useful only within a legal scoring plan. A real hand might have alternative yaku, but this scenario explicitly removes them.
C. Yes, as long as the hand contains dora — Why this misses the point
Dora is bonus value and cannot meet the missing yaku requirement. It does not repair the eligibility problem.
Take it to the table: Optimize progress toward a legal hand, not merely toward a complete-looking arrangement.
YOUR NEXT HAND HAS A PURPOSE
Record one reason, then review one decision
The game starts a new random hand. The worked hand above is a demonstration, so look for the same idea in the tiles you actually receive.
- Open a Riichi bot table with its fresh random deal and choose two candidate discards before making one move.
- For each candidate, name one shape or pair it preserves and the tiles that would improve it.
- State your yaku route and whether either discard damages it.
- If a player has declared riichi, check the candidate against that player’s discards and consider the other opponents separately.
- After the hand, review that single choice using only the information you had at the time.
Success looks like this: You can explain a choice in terms of shape, value and risk, and identify one new fact that would have changed it. A lucky result alone is not the success criterion.
Rules & further reading
These guides distinguish our game’s rules from the rules used at other tables. Check your group’s agreed rulebook before a real match.