Digital Points for Playing Like LuckyJ

Put Placement Before Ukeire

Start with the hanchan, then judge the hand. Tile efficiency tells you which discard gains shape; it does not tell you whether a 1,000-point win is enough, whether 5,200 changes the table, or whether pushing for 8,000 only hands first place to someone else. In South rounds especially, the same two-shanten hand can be a hand to end quietly, a hand to build into mangan, or a hand to abandon the moment one dangerous tile appears.

LuckyJ's results make this practical rather than philosophical. Across over 1,200 games and 13,000 hands, the average placement was about 2.27, with a win rate of about 23% and a deal-in rate of about 10%. The split is the important part: top-half games averaged 3.09 wins and 0.75 deal-ins, while bottom-half games averaged 1.39 wins and 1.48 deal-ins. Strong play was not just winning more hands; it was converting the right hands while refusing the wrong losses.

The leader pattern is real, but do not exaggerate it. Across a little over five thousand leader spots in the mined sample, LuckyJ showed a repeatable lean toward lower-danger choices among comparable tiles, and every showcased example here has at least one NAGA head on LuckyJ's side. Mortal support is mixed, so this is not a license to strangle every leading hand. It means that when two discards are close, the one that leaves a clean next turn and avoids a needless live middle tile deserves extra credit.

The exception is urgency. If you are third or fourth and a 1,000-point hand leaves the order unchanged, overprotecting shape is just a slower way to lose. A discard that keeps riichi, dora, or mangan alive can beat the locally safest tile when the value threshold is real. Conversely, when a cheap win protects first, do not turn it into a 5,200 dream by volunteering danger that the hanchan does not require.

Do Not Choose the Hand Too Early

Do not make a young hand confess too early. In the first row, the cleanest discard by ukeire is often only the cleanest discard for one version of the hand, while LuckyJ keeps tiles that let the same 13 tiles become riichi, an open yaku hand, value through dora, or a controlled fold. This is not hesitation. It is buying the right to decide after the draws and rivers have spoken.

Five-block theory tells you to build five blocks, but the early question is which blocks are allowed to remain alive. Take a constructed shape like [[2m]][[2m]][[3m]][[4m]] plus [[P]][[P]], red [[5pr]], and a weak [[7s]][[9s]] kanchan. Cutting the dragon because it is “just an honor” may increase immediate smoothness, but the [[P]][[P]] pair is also yaku, call permission, value, and a brake if the table turns. The weak kanchan is the part that needs a perfect draw to justify itself.

That is exactly where the data points. Across over 2,300 pair-or-triplet anchor spots and over 600 dora or red-five material spots, LuckyJ’s off-book choices repeatedly protected the tile with a second job. Early decisions also show the widest disagreement with NAGA, about 45%, while the late row drops to about 28%. The opening is where extra futures have value; by the end, most of them have expired.

Of course, ambiguity is not a religion. Once the hand has become clean ryanmen-ryanmen one-shanten, or the anchor no longer makes yaku, value, safety, or a call, cut the sentimental tile and take the efficiency. The mistake is breaking the pair or dumping the dora branch before you know whether the hand wants speed, value, or a safe exit.

Call Only When the Call Changes the Hand

Opening a bad shape is not the same thing as giving up value. The question is whether the call changes the hand from a wish into something that can win, reach tenpai, buy tempo, or end the round. A [[P]] pon that gives yaku and a 1,000-point finish before the dealer's riichi develops can be better than waiting for a closed premium worth about two han that this hand may never earn. A chi that merely turns bad two-shanten into exposed bad two-shanten is just paying value for no compensation.

LuckyJ is not a closed-only machine: it called in about 36% of hands and averaged about 5.65 calls per hanchan. But it also skipped roughly 2,000 calls that NAGA preferred, which is the important half of the story. In top-half games LuckyJ was more active, with 5.80 calls and 2.11 riichi per game, compared with 5.44 calls and 1.67 riichi in bottom-half games. The lesson is not “call more.” It is that passivity and random exposure are both weaker than selective opening.

So before you take the tile, look at what becomes true after the call. Does it make a legal yaku, produce immediate tenpai, turn a useless block into real one-shanten, deny the dealer a free draw, or let you take safe keiten later? Those are different gains, but all of them are visible on the next discard. If you cannot choose the post-call discard cleanly, or you cannot name the safe tile you are keeping for the first threat, you probably liked the sound of calling more than the position it creates.

Of course, a good closed hand should charge rent for being opened. Pinfu with dora, good shape, and riichi access is often aiming at 3,900 or more once riichi pressure and ura are included. Opening that into a fragile 1,000-point hand with no safety is not modern aggression; it is impatience. Bad shapes invite calls only when the call solves a real problem.

Keep Brakes on Open Hands

The first call does not remove defense from the hand; it makes defense more expensive. Once you open, your hand is shorter, your closed value is gone, and you have fewer chances to draw a random safe tile later. Keeping a genbutsu or correctly established suji against the live threat is not cowardice. It is the price of continuing to attack without promising to throw the next live [[5p]] through a riichi.

This is especially clear with cheap open hands. A 1,000-point yakuhai hand that is still two-shanten cannot spend every safe tile just to gain a few ukeire, because the future push may be against 3,900, 8,000, or dealer pressure. The same discard looks different when the hand is 3,900 tenpai with a good wait. Value and distance decide whether safety is fuel or clutter.

The open-hand evidence is now scoped to hands LuckyJ had actually opened. Across 6,021 LuckyJ/Nishiki open-hand splits, LuckyJ retained a recognized defensive tile in 3,536, including 1,723 tied to a live threat. These are descriptive choices, not proof that keeping the tile caused a better result.

After every pon or chi, ask who your saved tile answers. Safe to the quiet player is often just clutter; safe to the riichi player, the dealer, or the open honitsu seat is a real resource. Of course, when your open hand is already a strong tenpai or the table is quiet and your value is sufficient, you can spend the spare safety. The mistake is spending it automatically because the hand has become “attack.”

One saved genbutsu buys one turn, not a permanent fold. After you spend it, the next draw creates a new decision. A new genbutsu or low-danger tile that preserves one-shanten lets you continue by mawashi. A draw that produces good tenpai can justify a fresh push. If the only tile that preserves the shape is live against the threat, mawashi has ended: either push that tile on the value of the whole remaining route, or break the hand and fold.

Placement sharpens both sides of that fork. When you are fourth and the riichi dealer is third, your win also ends the dealer turn, takes the riichi stick, and prevents the gap from widening. That makes a strong one-shanten hand more valuable to preserve. Directly dealing into that dealer is also the worst possible transfer, so “stop the dealer” is not permission to volunteer several unsafe tiles. It supports one tolerable continuation when the winning route is credible; it does not rescue a cheap hand that still owes a chain of danger.

Cut the Tile That Will Get Worse

A floater has two prices: what it accepts now, and what it costs to throw later. Weak players count only the first price. LuckyJ often counts the second price while the river is still quiet, cutting a useful-looking simple tile before it becomes the tile you must push through row three.

Take a constructed slow hand with [[2m]][[2m]], [[6p]][[8p]], [[3s]][[4s]], a lone [[5p]], [[P]][[P]], and scattered honors. If [[5p]] is not dora and the pinzu side is not your main value plan, keeping it because it sees [[4p]] and [[6p]] can be fake efficiency. By the time an opponent riichi appears, that same [[5p]] may no longer be flexibility. It may be the only tile in your hand that nobody has made safe.

The broad review gave this point one of the cleanest LuckyJ signals: in over 3,700 lower-danger choice spots, it repeatedly spent tiles that would become more expensive later. But do not defend individual examples as if they were scripture — the second engine splits on these spots, backing some cuts and preferring pure efficiency elsewhere. The lesson is not “copy this exact middle-tile cut.” The lesson is to price future liability when the hand is slow and the tile has no concrete job.

Of course, pre-cutting a middle tile is terrible when that tile is your hand. A red [[5p]], a dora connector, the tile that gives you perfect one-shanten, or the only tile that makes mangan should stay. Cut the future problem when it is merely a floater; keep it when it is the reason the hand is allowed to fight.

Buy Value While It Is Cheap

Value planning starts before the hand looks valuable. The first discard is often the cheapest time to keep a red five connection, a dora side shape, or a yakuhai pair, because the hand still has several ways to reorganize. If you wait until iishanten to ask for 5,200, the price is usually a block, a wait, or a dangerous tile. Modern efficiency is not low value; it is knowing when the efficient hand is too small.

Take the ordinary comparison. A fast 1,000 or 2,000 can be perfect when it ends a parent turn or protects first, but it is a poor answer when South 3 requires a mangan swing. Keeping red-five access or dora use may turn the same skeleton into 5,200 or 8,000, and that changes whether a future riichi, call, or push is worth making. Dealer value sharpens this because dealer hands pay one and a half times; a hand that is merely acceptable as child can become table control as dealer.

The data supports the narrow version of this point, not the reckless one. In over six hundred dora and red-five retention spots, LuckyJ showed a clear tendency to keep value material when it still belonged to the hand. The broader story, behind means buy risk, did not get the same support: on the showcased value trades a NAGA head sides with LuckyJ each time, while Mortal frequently prefers the plainer efficiency line. Copy the red and dora discipline, not the fantasy that any expensive tile becomes correct because you are behind.

The exception is fake value. An isolated dora that does not connect, a yakuhai singleton you cannot use, or a honitsu idea that asks you to throw live middle tiles is not value planning. If the hand's realistic future is still 1,000 after you keep the shiny tile, cut it and build the best hand you actually have. Donating a dealer mangan while dreaming of your own is not courage; it is bad accounting.

Make Riichi Tax the Table

You cannot discuss modern riichi without discussing riichi as pressure. The declaration is not merely one extra han; it adds uradora chances, locks your hand, and often pushes opponents into betaori. A quiet 2,000-point damaten and a riichi hand that threatens 3,900 to 8,000 are different animals. The question is not "can I reach?" but "does reaching make the table pay?"

LuckyJ's volume is active but not mindless. Overall it averaged about 1.9 riichi per hanchan, and top-half games used more riichi than bottom-half games, about 2.11 compared with 1.67. That difference points in the right direction, but the lesson is conversion quality, not button-mashing. The sampled pressure-riichi cases receive unusually broad outside-engine support. Mortal usually supports declaring riichi in them, but that verdict covers the declaration only—not necessarily LuckyJ's declaration discard or wait.

Good riichi changes opponent behavior. First-to-act pressure against quiet rivers can make a leader throw away a 1,000-point open hand, freeze a dealer's two-shanten, or buy you the fold equity that a dama hand never receives. It also changes your own threshold: a bad 1,300 reach into two live threats is often just a 1,000-point stick and several forced danger tiles, while a 5,200-class reach with a live wait can be worth the lock because winning or making everyone fold both have value.

The exception is when riichi stops solving the actual problem. If dama already passes a rival in All Last, or if declaring lets opponents fold while you needed direct points from one seat, stay quiet. If two opponents are already live and your wait is poor, the pressure may be mostly on you. Riichi is a weapon, but a weapon pointed at the wrong table is just exposure.

Reprice the Whole Push Every Draw

A push is not the tile in your fingers. It is the whole string of discards you are promising to make before the hand wins. Reprice that promise after every draw, call, riichi, and new safe tile. The common leak is deciding this is my pushing hand and then treating the next discard as if the table has not changed.

LuckyJ's review splits punish that laziness. Across over 3,700 lower-danger choices against Nishiki, the review looked much cleaner; across over 5,000 higher-danger choices, the review got harsher. That does not mean every safe discard is correct. It means the danger meter is not decorative, and the second or third dangerous tile in the chain often matters more than the first tile you are staring at.

The extreme-danger sample here is nondealer-only. After removing every hand where LuckyJ was dealer and every locked discard after LuckyJ's own riichi, the Tokujou sample contains 46 chosen discards at 90% or more on the maximum NAGA danger proxy. All 46 faced at least one riichi. Thirty-three were tenpai, 12 were one-shanten, and only one was farther away; 20 came from open hands, and at least two of the three NAGA heads supported 30. Repeated 99.99% readings behave like a capped alarm state, not a literal deal-in probability.

The clearest supported push was closed monster tenpai: [[3p]] registered 99.99%, every NAGA head chose it, and the hand later won 10 han 60 fu for +17,000. A separate third-place hand pushed [[1m]] at 95.74% to declare riichi into an existing riichi; two heads supported it, and the hand later won +5,900. These are the clean stories: real tenpai, enough reward, and broad model support before the result.

Placement urgency produces harder cases. From fourth with only 2,900 points, LuckyJ pushed [[5s]] at 99.99% while still one-shanten. It kept 13 effective tiles instead of nine after Nishiki's safer [[5m]], but only one NAGA head supported the push. LuckyJ later won +6,500; that outcome shows the purchase, not proof that the price was correct.

Calls create a different kind of extreme push: an inventory trap. With three open melds and only five concealed tiles, LuckyJ pushed [[6s]] at 99.99% to keep tenpai. All three heads supported the line, but LuckyJ dealt in for -2,900. The hand was not locked by riichi rules, yet earlier calls had removed most practical exits. This is why the safety calculation belongs before the second and third call.

The strongest warning came from first place with 41,000 and five wall tiles left. LuckyJ pushed [[5m]] at 99.99% to preserve open tenpai while all three heads preferred [[7s]] at 2.97%, accepting one-shanten instead. The result was a -12,000 deal-in. Maintaining tenpai explained the move; the lead, short wall, cheap open route, and excellent alternative made it an overpush rather than a model to copy.

The outcomes across all 46 decisions were mixed: 12 self-wins, 11 direct deal-ins, 14 wins by another player, and nine draws. Results cannot adjudicate the discard. Use the state before the tile was cut: value, distance, placement, remaining wall, safe alternatives, number of future dangerous discards, and whether independent heads support the purchase.

The point-value comparison is blunt. Taking a -1500 draw payment can be cheaper than buying another live turn that ends -4200, while a hand with a realistic +8900 finish can afford a much larger toll. You are not asking, "Is this one tile passable?" You are asking whether the hand's reward pays for every unsafe discard still ahead.

Of course, there are nondealer hands where the answer stays push. High-value tenpai, a riichi whose added pressure and value matter, or All Last score needs can justify a dangerous tile even after the table worsens. But that is a fresh decision, not momentum from last turn; if the new draw added another live middle tile to the future, the old push no longer exists.

The Third Row Is a Counting Drill

The late row is not early strategy with fewer tiles. It is a separate skill: count the winning tenpai, count the safe tenpai, and count the fold. LuckyJ disagreed with NAGA on about 45% of early decisions, but only about 28% late. That drop is the point. By row three, most grand plans have expired, and the good decisions are smaller, colder, and more exact.

Keiten is the easiest place to see it. In the recorded wall-exhausted draws, LuckyJ was tenpai in 823 of 1,761 (46.7%). That is 823 of all 13,054 hands, so keep the denominator straight. Noten payments are real value, but they do not justify sending a live middle tile through two threats with no winning upside.

Take a constructed third-row spot: you can cut a live [[5s]] for pinfu tenpai on [[3p]][[6p]], cut genbutsu [[9s]] and remain one-away, or follow with [[E]] and fold. The correct question is not which tile is prettiest. Ask what the win changes, what tenpai at draw is worth, how many safe tiles remain after this discard, and whether the next tile is also dangerous. If the answer to the next-turn question is blank, the current push is usually fake.

Of course, late precision does not mean cowardice. If the hand is dealer mangan tenpai, the wait is alive, and the only discard is a reasonable suji, folding because “it is late” gives away too much. The discipline is to separate real attack from vague hope: win, take safe tenpai, or fold cleanly.

Treat Keiten as Attack

A drawn hand is not a failed hand. Keiten is a small attack that happens after winning has become unlikely, and the payments are large enough to plan for: 1,000 when several players are tenpai, up to 3,000 when you are alone. That can be the same practical swing as a small win, especially in South rounds where 1,500 or 3,000 changes who must chase next.

LuckyJ reached tenpai in 46.7% of genuine exhaustive draws. That does not mean 46.7% of all hands: abortive draws are excluded, and the all-hand share is about 6.3%. The last row is where safe tenpai, noten acceptance, and final danger pricing decide placements; keep a path to tenpai only when it can be traveled with harmless tiles.

The key word is harmless. If you can discard genbutsu, dead honors, or visibly exhausted tiles while keeping tenpai, the 1,000 to 3,000 draw value belongs in your attack calculation. If the price is a live middle tile into dealer riichi, you are no longer buying keiten; you are gambling a possible dealer mangan loss to save a small payment. A Tenhou player who never chases draw points leaves money on the table, but a player who chases them through danger is just calling noten fear by a prettier name.

The exception is forced danger. When the only tenpai path asks you to pass two unsafe tiles, or when an open hand's value has become clear, fold and accept noten. You can still defend actively by choosing the safe discard that leaves the most future safe draws. Draw points are attack value only while the defense remains intact.

Make Engine Splits Prove Their Case

An engine disagreement is not a verdict. It is a review prompt. LuckyJ matches NAGA's top discard about 73% of the time, so the split hands are exactly where the book becomes interesting: not because LuckyJ is always right, but because the difference forces you to write down what the move is buying.

The honest framing matters. The full review holds over 36,000 LuckyJ/Nishiki mismatches from about 135,000 split-head decisions, and most of them are not giant alarms. The NAGA heads are not a single judge either; in the samples below, some NAGA head backs LuckyJ. Mortal often backs Nishiki or chooses a third line, so you should not pretend a disagreement is engine consensus.

Use three buckets. First, blunder: the move fails basic value, safety, or yaku accounting, and you fix it. Second, style trade-off: LuckyJ is choosing safety, pressure, or placement texture that you can explain but do not copy automatically. Third, genuine insight: the split reveals a conditional habit you can reproduce, such as valuing a future safe exit over a little ukeire.

Point values keep the review honest. Blindly copying a split can become a -3900 deal-in, while flattening every LuckyJ deviation back to NAGA can make you miss a +2900 win or a +1500 draw-tenpai result. The number is not proof by itself. It is a sanity check: what did the move risk, what did it buy, and was that purchase available to a human at the table?

Of course, if your explanation is weak and the stronger engines all point away, do not cosplay as an AI. Mark the hand uncertain and move on. The goal of review is not to raise your match rate; it is to copy only the decisions you can explain before the result is known.

Price the Missing Yakuhai

When an opponent opens and you cannot see their yaku, a live dragon, round wind, or seat wind stops being an ordinary honor. It may be the tile that makes their hand legal. Discarding it gives them the chance to pon; holding it denies that chance for now. The decision is therefore timing, not automatic cleanup.

There are two deliberate lines. Release the singleton while no yaku or clear advancement is visible, accepting today's pon risk so you do not carry a worse live tile into a later turn. Once advancement becomes visible—more calls, a late turn, or a tenpai-like river—choke the tile unless your own value, safety, or placement gives a concrete reason to feed it.

The observed medians are only descriptive. Tiles eventually cut against an unknown open hand left around turn 8, versus about turn 10 against tanyao-shaped opens and turn 7 when the yaku was already visible. Those groups condition on the tile surviving until that context; they do not prove that the opponent's call caused LuckyJ to wait.

Do not copy this point as a blanket rule. The selected examples are engine-contested, and Mortal usually favors Nishiki or another line. Treat each as a stress test: compare the pon risk now with the danger of carrying the tile, then keep it when it is your pair, value, only safe tile, or a real honitsu/chiitoi component.

Name Who Your Safe Tile Defends

A safe tile is not safe in the abstract. It is safe against a person. Genbutsu to a quiet child does not solve dealer riichi, and suji to an old river is only useful if that player is the one who can punish you. LuckyJ's stronger defensive habit is not hoarding; it is keeping a tile because you can name the opponent it answers.

The empirical anchor is LuckyJ's whole profile. Across about 13,000 hands it won 22.9% of hands while dealing in only 10.1%. The top-half games averaged 0.75 deal-ins, while the bottom-half games averaged 1.48. That gap is not created by mystical reading; it is created by keeping a brake for the player who actually matters.

Think in prices. Cutting a slightly useful shape tile now may cost speed, but if the kept genbutsu prevents even one -4000 turn, the exchange has paid for itself many times over. Conversely, a -500 nuisance loss is not a reason to spend the only real answer to the dealer. Genbutsu is cash, suji is a smaller bill, and both must be labeled by opponent before you put them in your wallet.

Of course, the label can be fake. If the table is quiet, the hand is valuable and fast, or the tile is only weak suji to someone who is not threatening you, do not tax your hand for decorative safety. Safety that answers nobody is just a worse floating tile.

Spend Safety When Its Opponent Disappears

A tile can be genbutsu and still be the correct discard. The question is not whether it used to be safe; the question is whether it protects you from the danger now. Once another player becomes the real threat, or once the hand is close enough that the tile blocks tenpai, yesterday's safety has an expiration date.

This is the necessary counterpart to keeping named safety. In the model-proxy audit, threat-specific keeps and spending stale or off-target safety were less often severe Nishiki disagreements, while generic safe-tile retention was more often rejected. That is not outcome EV; it is a narrower review signal. The tile still has to prove its current job.

The economics are easy to see. If keeping an old safe tile kills a live 2,000-point tenpai chance, you have paid real points for insurance that may not pay out. If the alternative is drifting into a -1,500 no-ten draw, the cost is visible even without dealing in. A stale tile that answers a quiet player is not defense; it is shape damage wearing a helmet.

Of course, do not throw away your last real parachute. If the tile still covers the live riichi, your hand is far, or the next draw may leave you with no safe discard, keep it. Spending stale safety is disciplined only after you can say why that tile no longer answers the dangerous player.

The Edge Tile Can Be the Attack

Do not call every terminal cut defensive. Sometimes the outside tile is the price of keeping the hand’s center alive. A late [[1p]] or [[9s]] discard can be the attacking choice because it protects the middle connector that actually reaches tenpai, while also lowering the immediate danger of the discard.

Take a constructed excess shape like [[1p]][[3p]][[4p]][[6p]][[7p]] when the rest of the hand already has value. Cutting [[1p]] leaves [[3p]][[4p]] and [[6p]][[7p]], two normal ryanmen cores. Cutting [[4p]] because [[1p]] looks safer may preserve a dead-looking outside tile while damaging the part of the hand that wins. In five-block terms, the terminal is not a block; the middle tiles are.

This pattern was not rare. Across over 5,000 spots LuckyJ cut outside material while keeping the inside line, and the reverse pattern appeared in over 10,000 spots. Those are descriptive model-split counts, not proof that the outside cut is better. Mortal support in the showcase is mixed, so copy the structure only when the outside tile has a weaker job than the middle connector it protects.

Of course, terminals and outside tiles are not garbage. A [[1m]] pair can be chiitoi material, a [[9p]] can be dora, and an outside tile may be your only genbutsu against dealer riichi. When the outside tile is value, pair, yaku, or target-specific safety, it is no longer the cheap discard. The principle applies when the edge is truly the cost of preserving the winning center.

Label Every Honor Before You Move It

“Honor tile” is not a strategic category. A dragon pair in your hand, a live dragon for an open opponent, a double-visible wind, and a genbutsu honor against riichi are four different objects. If you review them all as “honors,” you will cut value tiles too early and keep dangerous tiles too long. The first step is to label what the tile is doing right now.

The LuckyJ timing split shows why the label matters. Loose self-yakuhai with no open hand around were cleaned at a median turn 4, but in tanyao-shaped open contexts they were delayed toward turn 10. That is not a contradiction; it is a role change. The same [[C]] can be a bad early singleton, a future 1-han pair, the opponent's missing yaku, or your last honest fold tile.

Use four main labels. Self value: your yakuhai pair, honitsu material, chiitoi pair, or dora-connected plan. Opponent condition: live dragons and value winds that complete an unclear open hand. Dead tile: an honor with little value and enough visible copies that nobody can use it. Defense: genbutsu, multi-player safety, or a low-friction tile reserved for the first riichi. A self-value dragon can make 1,000 by itself and 3,900 with dora; an opponent-condition dragon can cost you the round.

Of course, labels are not permanent. A lone guest wind is pleasant early, but it grows worse against chanta, kokushi, or toitoi shadows late. A dead honor becomes precious if it is your only safe tile against dealer riichi. Re-label after every call, riichi, visible copy, and score change; then cut or keep the honor for the role it has now, not the role it had on turn one.