Tile Efficiency (牌効率)
8.1 What Is Tile Efficiency?
Tile efficiency (牌効率, hai kouritsu) is the systematic study of which discard choice maximizes the number of tiles that improve your hand. In its purest form, tile efficiency asks: "Given my current 14 tiles, which tile should I discard to maximize ukeire—the number of remaining tiles that bring me closer to tenpai?" This is the single most important mechanical skill in riichi mahjong, and the one that separates beginner play from intermediate play.
Pure tile efficiency considers only speed—how quickly can you reach tenpai? It does not consider hand value, defense, or game context. These factors modify tile efficiency decisions in practice (see Module 27, Efficiency vs. Value), but understanding pure efficiency is the essential foundation. You must know the "correct" efficient answer before you can make informed decisions to deviate from it.
8.2 Core Principles
Principle 1: Keep Tiles That Connect
Tiles that are adjacent or near other tiles in the same suit have more potential to form mentsu. A hand full of connected tiles (sequences in progress, pairs, two-sided partial groups) has higher ukeire than a hand full of isolated tiles. When choosing what to discard, prioritize keeping tiles that are part of or adjacent to partial groups.
Principle 2: Two-Sided Waits Are Superior
A two-tile partial like 4m-5m can become a complete sequence by drawing either 3m or 6m—two different tiles, each with up to 4 copies remaining. A one-sided partial like 1m-2m can only be completed by 3m. An edge pair like 8m-9m can only be completed by 7m. Two-sided connections are approximately twice as flexible as one-sided ones. Keep two-sided partials and discard one-sided ones when possible.
Principle 3: Middle Tiles Are More Flexible Than Edge Tiles
A tile like 5m can participate in three different sequences (3-4-5, 4-5-6, 5-6-7). A tile like 2m can participate in two (1-2-3, 2-3-4). A tile like 1m can only participate in one (1-2-3). This means middle tiles (3-7) are inherently more valuable for hand building than edge tiles (1, 2, 8, 9). Isolated middle tiles should generally be kept over isolated edge tiles.
Principle 4: Pairs Have Hidden Value
Every standard winning hand needs exactly one pair. When your hand has zero pairs, finding a pair becomes critical—any pair-making draw is valuable. When your hand has two or more pairs, one pair is the jantai and the other(s) may become triplets (with pon or a lucky draw) or remain as useful pair components. The optimal number of pairs in a developing hand is typically 1-2.
8.3 Tile Group Rankings
Understanding the relative value of different tile configurations is essential for tile efficiency:
| Configuration | Example | Acceptance (tiles that complete it) | Notes |
|---|---|---|---|
| Complete mentsu | 4m-5m-6m | N/A (already complete) | Done—no action needed |
| Two-sided partial (両面, ryanmen) | 4m-5m | 2 types (3m, 6m) = up to 8 tiles | Best partial group |
| Pair (対子, toitsu) | 4m-4m | 1 type (4m) = up to 2 tiles for triplet | Also serves as jantai |
| Middle gap (嵌張, kanchan) | 4m-6m | 1 type (5m) = up to 4 tiles | Better than penchan |
| Edge partial (辺張, penchan) | 1m-2m or 8m-9m | 1 type (3m or 7m) = up to 4 tiles | Worst sequential partial |
| Isolated middle tile | 5m alone | Potential to connect with many draws | Keep over isolated terminals |
| Isolated terminal | 1m alone | Very limited connectivity | Often first discard candidates |
| Isolated honor | 3z alone | Only pairs with identical tile (3 remaining) | Discard early unless yakuhai |
8.4 Worked Example: Choosing a Discard
Hand (14 tiles, 2-shanten):
Analysis:
Partial groups identified: 4m-5m (ryanmen, excellent), 2p-3p (needs 1p or 4p, one side is penchan), 7p-8p (ryanmen), 7s-8s (ryanmen), 3s-5s (kanchan, needs 4s).
Isolated tiles: 1m, 1z, 5z, 6z.
This hand has too many partial groups and needs to shed isolated tiles. The discard candidates are 1m, 1z, 5z, and 6z.
Best discard: 1z or another non-yakuhai honor. Isolated honor tiles have the least potential to connect with future draws (they can only form triplets with exact matches). Between 5z (haku) and 6z (hatsu), note that these are yakuhai—if you draw a second one, the pair has value as potential yakuhai triplet. But at 2-shanten with multiple good partial groups, speed favors discarding the honor that is least likely to help. In practice, most efficiency-focused players would discard 1z first (non-round, non-seat wind, purely isolated), then the isolated 1m terminal.
8.5 Practice Drills
Tile efficiency improves dramatically with practice. The following drill types are recommended:
What-to-cut drills (何切る, nani kiru): Given a 14-tile hand, determine the optimal discard for maximum ukeire. These drills are available on many Japanese mahjong websites and in published books. A classic resource is the "Nani Kiru" problem format used in Japanese mahjong magazines.
Speed drills: Practice sorting hands and identifying partial groups quickly. On online platforms, play fast games focused purely on hand efficiency, temporarily ignoring value and defense.
Review drills: After each game, review your discards in the early hand and check whether you chose the highest-ukeire option. Use analysis tools (Tenhou's log analyzer, Mahjong Soul's replay system) to verify.
At higher levels, pure tile efficiency is modified by three factors: (1) hand value—sometimes a slightly less efficient discard leads to a much more valuable hand, (2) defense—some tiles are safer to hold for later defensive use, and (3) information—tiles that opponents have discarded or called change the effective ukeire because some tiles are now impossible or less likely to appear. Module 27 covers the efficiency-vs-value tradeoff in depth.
QUIZ — Question 8.1
Which partial group has the highest tile acceptance (most tiles that complete it)?
Answer: A. The ryanmen (two-sided) partial 4p-5p can be completed by 3p or 6p—two different tile types, each with up to 4 copies = up to 8 tiles. Kanchan 4p-6p needs only 5p (up to 4 tiles). Penchan 8p-9p needs only 7p (up to 4 tiles). Pair 4p-4p needs another 4p for a triplet (up to 2 tiles). Ryanmen is clearly superior for speed.
8.6 The "Five-Block Theory" (五ブロック理論)
One of the most important frameworks in Japanese tile efficiency teaching is the "five-block theory" (五ブロック理論, go-burokku riron). A standard winning hand needs 4 mentsu + 1 pair = 5 "blocks." When evaluating a 13-tile hand, you should identify the 5 most promising blocks (which may be complete mentsu, partial groups, or pairs) and discard tiles that fall outside these five blocks.
This framework, popularized through Kindai Mahjong (近代麻雀) magazine articles and formalized in instructional texts by professional players, provides a practical mental model for discard decisions. If your hand has 6 or 7 partial blocks competing for 5 slots, you must choose which to keep and which to abandon. The answer usually comes from ukeire comparison: keep the blocks with the highest acceptance and discard the weakest.
Five-block analysis:
Identifying blocks: Block 1: [2m-3m] ryanmen partial. Block 2: [5m-6m] ryanmen partial. Block 3: [3p-4p] ryanmen partial. Block 4: [7p-8p] ryanmen partial. Block 5: [5s-6s-7s] complete shuntsu. Block 6: [3z-3z] pair.
That is 6 blocks for 5 slots. One block must be abandoned. The pair (3z-3z) is important — every hand needs a pair. The complete shuntsu [5s-6s-7s] is obviously kept. Among the four ryanmen partials, all are equally strong. But you can only keep 3 of them (plus the pair and complete shuntsu = 5 blocks). Which to drop?
Discard: 3z. Wait — discarding a pair tile? Actually, reconsider: any of the four ryanmen could become the pair if you draw a duplicate. Keeping all four ryanmen and discarding one 3z gives you maximum flexibility — you have 4 ryanmen partials and if any of the 8 tiles that complete them arrives, you progress. Drawing a duplicate of any held tile creates your pair. This is more flexible than locking in the 3z pair and dropping a ryanmen. The correct discard is one of the 3z tiles (keeping just one for potential pair flexibility) or committing to the pair and dropping the weakest-positioned ryanmen.
In practice, at this early stage, discarding 3z is likely optimal because the four ryanmen partials collectively have much higher ukeire than the honor pair. This type of analysis is exactly what nani-kiru (何切る) drills train.
8.7 Nani Kiru — The Gold Standard of Efficiency Training
The "nani kiru" (何切る, "what to cut") format is the primary drill type for tile efficiency in Japanese mahjong education. The format is simple: given a 14-tile hand (after drawing), determine the single best discard. Problems range from beginner-level (obvious isolated honor vs. connected tiles) to expert-level (marginal differences of 1-2 ukeire between complex options).
Major Japanese sources for nani-kiru problems include:
Fukuchi Makoto (福地誠), 『麻雀 定石「何切る」301選』(Takeshobo): A collection of 301 problems organized by difficulty, considered a standard training resource. The book covers beginner through advanced scenarios with detailed explanations.
Kindai Mahjong magazine (近代麻雀): Regular nani-kiru columns have appeared in this magazine for decades, often authored by professional players. These problems reflect real game situations.
Online nani-kiru sites: Several Japanese websites offer interactive nani-kiru problems with automatic ukeire calculation. The site "天鳳牌効率" (Tenhou Hai Kouritsu) provides problems derived from actual Tenhou game positions.
Mahjong Soul and Tenhou replay analysis: Reviewing your own games and checking each discard against the optimal ukeire choice is effectively a personalized nani-kiru drill. This method is recommended by virtually all Japanese mahjong coaches because it addresses your specific weaknesses rather than generic problems.
8.8 When Pure Efficiency Is Wrong
A critical nuance emphasized by Kobayashi Gō (小林剛) and other analytically-minded professionals: pure tile efficiency (maximizing ukeire without considering hand value or defense) is the correct approach for approximately the first 6-8 discards of a hand. After that, as your hand approaches tenpai and opponents begin showing threat signals, the calculation shifts. The "correct" discard becomes a function of ukeire × hand value × safety, not ukeire alone. Module 27 covers this tradeoff in full detail.
Source notes: Tile efficiency theory (牌効率) is extensively documented in Japanese mahjong literature. The five-block theory (五ブロック理論) is a standard teaching framework that appears across multiple Japanese instructional texts and magazine articles. Fukuchi Makoto (福地誠), 『麻雀 定石「何切る」301選』, 竹書房 (Takeshobo) — the standard nani-kiru problem collection. Kindai Mahjong (近代麻雀), published by Takeshobo, has featured nani-kiru columns for decades. Kobayashi Gō (小林剛), a professional player known for data-driven analysis, has discussed the limits of pure efficiency in his published works and M-League commentary. The ranking of partial groups by acceptance count is universally agreed upon across all Japanese sources.