Why Studying Discard Problems Rarely Improves Your Results

Most Riichi Mahjong players study the same way. They watch a video, read a book, or run a hand through an AI reviewer, see the explanation, and think “that makes sense.” The logic is clear, the answer feels obvious in hindsight, and they move on to the next problem.

Then a similar hand appears in a real game and they discard the wrong tile anyway.

This is the gap that separates knowledge from skill. If you cannot produce the correct discard at the table, the study session did not change your results — it only changed how the explanation felt while you were reading it. The training method in this article exists to close that gap, and it is deliberately repetitive. It is not exciting. It does work.

Two Discard Problems That Share One Logic

Start with a simple tile efficiency problem. Ignore Dora and turn number for now — they do not change the answer here. In a hand where the Manzu block is 2-2-4-5-6-7, the correct discard is the 2-man.

Most players compare discarding the 2-man against a lone 2-pin elsewhere in the hand. Cutting one 2-man leaves 2-4-5-6-7 in Manzu, which accepts the 3-man as a kanchan (inside wait) on top of the 5-man and 8-man ryanmen (two-sided wait). Discarding the 2-pin instead leaves the Manzu accepting only the 5-man and 8-man. The 2-man discard buys you one extra accepted tile type for free.

This shape appears constantly in beginner-level discard content and is usually called a ryanmen-kanchan: a block that is simultaneously a two-sided wait and an inside wait.

The Same Logic in a Different Costume

Now take a second problem. The Pinzu block is 3-4-4-5-6-8-9 and the Manzu side holds a separate 7-8 penchan (edge wait). The correct discard is the 9-pin.

Cutting the 9-pin leaves 3-4-4-5-6-8 in Pinzu, which accepts the 7-pin in addition to the 2-pin and 5-pin ryanmen. Discarding a Manzu tile instead leaves you with an independent 2-5-pin ryanmen and a separate penchan, and the acceptance count drops by one type.

If you look at the two problems side by side, they are the same problem. Same logic, same conclusion, different surface. The real question is not whether you can solve them when they are presented one after another with the explanation attached. It is whether you would have solved the second one cold, four turns into a live hand, having only ever seen the first.

Four Ways the Same Shape Disguises Itself

Riichi Mahjong makes pattern recognition unusually hard because a single logical shape can appear in many visually different forms. There are four common disguises, and each one is enough to make a player miss a discard they “already know.”

1. The Suit Is Swapped

Manzu, Pinzu, and Souzu are mechanically identical, but they do not feel identical. A hand where the answer is “discard the 2-man over the 2-pin” becomes a hand where the answer is “discard the 2-sou over the 2-pin” the moment the suits rotate. The logic is untouched. Many players still hesitate.

2. The Numbers Are Mirrored

Number tiles are symmetrical around 5. The 1 and the 9 behave the same way; so do the 2 and the 8, the 3 and the 7. If discarding the 2-man is correct in one hand, discarding the 8-man is correct in its mirror image. Study material almost never spells this out, because it assumes you will convert it yourself.

3. The Starting Shape Is Slightly Different

The first example compared two pairs, where breaking one pair created the ryanmen-kanchan. The same conclusion applies when you are comparing two penchan blocks instead: discarding the 1-man from a 1-2 penchan can leave you with the identical remaining structure. The tile you cut is different, but the shape you keep is the same, and that is the part worth memorizing.

4. The Shape Only Appears After a Draw

This is the disguise that costs the most points. Your hand does not start in the textbook shape — it becomes the textbook shape three or four turns in, the moment you draw one tile.

Picture the hand before that draw. You are thinking about several branches at once: draw the 2-pin, 2-man, or 3-man and you can declare Riichi on an 8-5-pin ryanmen; the 3-man version also picks up Pinfu; if the 8-pin or 5-pin arrives first you have to decide between a shanpon (dual pair) Riichi and staying damaten on a 3-man wait, because the 4-5-6-7 Manzu connection is likely to improve into a ryanmen anyway. On top of that you are tracking that the player across from you just discarded an 8-pin and a 3-man.

Then you draw the 6-man, and the hand is suddenly the exact shape you studied. Do you recognize the ryanmen-kanchan and cut the 2-man — or do you think “good, my Manzu is 4-5-6-7 now, I do not need the shanpon,” and reflexively toss the 2-pin? Under live conditions, most players toss the 2-pin.

The Training Menu: Build Your Own Problem Set

The method is simple to describe and tedious to execute: take every discard problem you got wrong, generate your own variations of it, and drill them until you cannot miss.

Start with problems where the explanation made you think “that was simple.” Those are the ones inside your reach, and they are the ones that will actually appear in your games. Problems whose explanations still feel confusing are above your current level; leave them for later. Then rebuild each one using the four disguises above.

• Rotate the suits. Rewrite the same hand three times, once in each suit.
• Mirror the numbers. Convert 2-2-4-5-6-7 into 3-8-8-4-5-6 style equivalents around the 5-axis.
• Change the starting block. Turn a pair comparison into a penchan comparison that resolves the same way.
• Shift the whole shape. Push 2-2-4-5-6-6-7 up one step into 3-3-5-6-7-7-8 and keep going.

Between those four operations, a single logic can generate several hundred problems. You do not need to write out a thousand. The standard is: if someone handed you a thousand variations of this one logic, would you get a thousand right? Getting 950 means you have not finished. Strong players move on only at 1,000 out of 1,000.

A notebook is enough — write the hands out yourself, suit by suit, with the answers on a separate page. If you want to face each problem cold, use any app that shuffles a set of saved images and shows one at random. Building a small personal discard trainer this way is well within reach of anyone comfortable with basic coding tools.

What to Drill First If You Are Stuck

Players who plateau in the middle ranks almost always have a tile efficiency problem underneath the symptoms. Before adding more theory, close the gaps in the basics. Three areas give the best return:

• Basic shape problems. Block counting, ryanmen-kanchan, and other acceptance comparisons.
• Defensive accuracy. Folding cleanly without leaking a deal-in on the way out.
• Basic push or fold decisions. The routine cases, not the marginal ones.

Build a variation set for each of these three, weighted toward problems you have already missed. Mastering a problem you once got wrong — including in forms you have never seen — removes a weakness permanently rather than papering over it.

Reviewing your own game records is the other classic answer, and it is fundamentally the same practice: over hundreds of reviews you eventually meet every scenario repeatedly. The catch is efficiency. If you are targeting one narrow category, a comparable situation might surface once every few dozen games. A purpose-built problem set delivers the same repetitions in an afternoon.

One caution about relying on published problem books. They are useful, but most are written to showcase as many different theories as possible, because variety is what readers reward. A book with 200 problems that all reduce to 30 patterns looks lazy on the shelf, even though 30 patterns drilled to mastery is exactly what makes a player stronger. That structural incentive is why the repetition has to come from you.

Does Repetition Training Actually Work?

Coaching sessions where a strong player reviews someone's game records are popular, and they are genuinely useful as a way to notice blind spots. But a single two-hour review does not produce a large jump in strength. If it did, nobody would be stuck.

A more revealing comparison comes from structured drilling. In one documented coaching case, a mid-rank player spent six consecutive days doing almost nothing but the same category of discard problems — basic tile efficiency and shape matching, drilled to the point of tedium. Because ranks move slowly and swing with luck, the before-and-after comparison used AI hand analysis instead: the blunder rate across the ten games before and the ten games after.

The blunder rate dropped from roughly 13 percent to roughly 10 percent, driven entirely by cleaning up fundamentals such as five-block thinking. Ten-game samples fluctuate, so treat the exact figures loosely. The qualitative change was harder to argue with: the same player could now apply basic shape logic in live hands, which is precisely the ability that separates a stable rank from one that keeps sliding back.

By contrast, lecture-style sessions covering defense and hand value in a single stream produced far less visible improvement. The pattern matches ordinary studying. Attending a lecture rarely made anyone good at a subject; repeated drills and past papers did. Applied to Riichi Mahjong, that principle produces exactly this training menu.

Quick Reference Summary

Use this table as a checklist when you build your own problem set.

Final Thoughts

This is not a clever shortcut, and it is not enjoyable in the way that watching a strategy video is enjoyable. Writing out the same hand in three suits, mirrored around the 5-axis, shifted up a step, and then answering all of it correctly a week later is genuinely tedious work.

It is also the most reliable tile efficiency practice available. Every problem you convert from “I understood that” to “I cannot get that wrong” is a permanent gain that shows up in real hands, at the moment you draw the tile that turns your hand into a shape you have already mastered.

Pick one discard problem you missed recently. Build ten variations of it tonight. That is the whole method.