Genshin Pull Planning: Two Complete Teams Without Burning Primogems

A Pull Plan Is Not a Wishlist

By the end of this guide you will be able to do three things: convert your account's state into a number of wishes you can actually expect rather than hope for; decide in under a minute whether any banner is a pull, a skip, or a conditional pull using a rule you can defend; and write a stop condition before you start pulling, which is the only thing that protects you from the worst outcome in this game — spending 150 wishes on a character you wanted moderately and having nothing left for the one you needed.

The mental model that ruins accounts is the single-banner model: see a character, decide whether you want them, spend until you get them or run dry, repeat from zero. That process has no memory, and gacha rewards are entirely about memory. The pity counter carries over. The guarantee state carries over. The decision you make now determines your starting position on the next three banners. A single-banner thinker routinely spends a guarantee on a character they like, then cannot afford the character their account needed — and the fix is never more primogems, it is a plan made several banners ahead.

The second half of the model is what you are buying. You are not buying characters, you are buying role coverage. An account needs two teams that handle different chamber types, and each needs an engine, a converter, a driver, and usually sustain. Once you can name which job is unfilled, the pull question stops being aesthetic and becomes arithmetic. That is why two players with identical income end up in different places: one bought four drivers, the other bought two engines and two converters.

Internalise one honest test: if you deleted this character tomorrow, would either of your two teams lose a job that no character you already own can fill? If not, it is a want — and wants get funded only after needs.

How Pity, Banners, and Guarantees Actually Behave

On the limited character banner, the base five-star rate is 0.6 percent per pull. Hard pity guarantees one by the 90th pull, but the important number is soft pity: somewhere around the mid-70s the rate climbs steeply, and the overwhelming majority of five-stars arrive between roughly pull 75 and 82. Because of that cliff the effective average is roughly 62 to 63 wishes per five-star, not the 166 that a flat 0.6 percent implies. Plan with the average; budget with the worst case.

The 50/50 works like this: your first five-star on a limited character banner is 50 percent featured, 50 percent standard-banner. If you lose, your next five-star on a limited character banner is guaranteed featured. Two properties follow. Both the pity count and the guarantee state carry over between banners indefinitely — a guarantee built eight months ago is still a guarantee. And a loss is not wasted value: it converts the next featured character from a coin flip into a certainty, which is the most valuable thing your account can hold.

Write these down, because you will need them: 90 wishes for one five-star at the worst counter, 180 wishes for a guaranteed featured character from zero with no guarantee. At 160 primogems per wish that is 14,400 and 28,800. Anyone who tells you to "save 90" is setting you up to lose the 50/50 and start over.

Four-star structure matters more than people think. Every tenth pull is guaranteed to be a four-star or better, and among four-stars the featured/off-banner split uses the same guarantee logic. But there is no pity for which featured four-star you get, which is why chasing one is a trap: you can easily spend 60 wishes and never see the one you came for.

The weapon banner is a different machine. Hard pity is 80, soft pity arrives earlier, and instead of a 50/50 you get roughly a 75 percent chance that the five-star is one of the two featured weapons. The safety net is the Epitomized Path: pick a weapon, and every five-star that is not it grants a fate point; at two fate points your next five-star weapon is guaranteed to be your choice. Fate points reset when the banner ends or when you get the weapon, so they are not bankable. The worst case for one specific weapon is therefore three five-stars — 200-plus wishes for what is usually a modest gain.

Standard and Chronicled-type banners run on their own pity counters and their own guarantee state; a guarantee built on a limited banner does not transfer to them, and vice versa. Never spend on a banner because you are "close to pity" if that pity is not shared with the banner you actually care about. One newer mercy rule is worth knowing but not planning around: after consecutive lost 50/50s, the game can convert a later loss into a win rather than let the streak run. Treat it as reduced worst-case pain, not a mechanic you can schedule.

What a Patch Cycle Actually Earns You

These are ranges, not promises — the biggest variables are whether the patch adds a region and how much Abyss you can clear. Use the low column for anything you cannot afford to miss. Figures assume a roughly six-week patch.

Two conclusions fall out directly. A free player who logs in daily, clears every event, and finishes the Abyss earns roughly 50 to 70 wishes per patch — call it 60 for planning. And a guaranteed featured five-star at the 180-wish worst case costs about three patches of total discipline, or a bit under two with Welkin. That number governs everything: you can afford roughly one guaranteed character every two to three patches, and everything else is average luck or a four-star.

There is also a quiet rebate: pulling generates stardust and starglitter, and starglitter buys fates. Over a long saving period that refund effectively discounts every wish by something in the neighbourhood of ten percent — but only if you spend it on fates rather than convenience materials.

Three Pull Plans, With Explicit Stop Conditions

Scenario A — the near-empty roster. Few or no limited five-stars, no two coherent teams, and a lot of unclaimed overworld content. Your advantage is enormous front-loaded income from exploration, quests, and achievements; your risk is spending it on four attractive carries instead of two functional teams. Sequence: first, identify the two archetypes your four-stars can almost support. Second, pull only for characters who complete those archetypes, ordered by whichever job no four-star can cover. Third, set the stop condition as "obtain the featured character or reach 180 wishes, whichever comes first," and if you hit the character early, stop and bank. No weapons, no constellations. One more functional character beats any damage increase, because an account with one team cannot farm the content that pays for the second.

Scenario B — one strong team, one empty half. The most common mid-game state and the most expensive place to make mistakes, because you have enough power to clear most content and therefore no pressure to fix the gap. Sequence: list your existing team's four jobs and mark the load-bearing characters. Then ask what the opposite team needs — not a second carry, but a second engine. If your first team is a Pyro amplification team, your second should not be another Pyro carry competing for the same supports; it should use entirely different supports so both teams can be fielded at once. The rule that saves this scenario: never pull a character who competes with your existing team for the same support slot. A new carry who needs the same off-field buffer your current carry needs is a sidegrade no matter how strong they are alone. Stop condition: pull until the second engine is acquired, then skip everything else in the patch, including upgrades to your first team.

Scenario C — two banners you both want, budget for one. The tiebreak is irreplaceability: for each character ask whether anyone you own fills that job at, say, 80 percent effectiveness. The one with no cheap substitute wins, regardless of which you like more. If genuinely tied, use the second criterion — which one unlocks content that pays for itself, since a character who lets you clear more Abyss stages is worth more than one who speeds up content you already clear. The execution rule is the important part: decide the order, pull the first to completion, stop, then re-evaluate. Never split wishes across two banners. Splitting is how you end up with two 50/50 losses and no characters, which is both the statistically worst outcome and the emotionally most common one.

Two notes that apply to all three. Pulling primarily for a featured four-star is gambling on a slot machine with no guarantee — acceptable only as a side effect of wanting the five-star, never as the reason. And if you are sitting on a guarantee, remember that spending it spends your most valuable asset: the correct price for a guarantee is a job that needs filling, not curiosity.

The Pull Priority Matrix

Read the third column as the heart of the table. Four-stars cover an enormous amount of this game: healers, shielders, off-field applicators, and respectable Anemo crowd control. What they cover poorly is high-strength off-field application with long uptime, universal resistance shred paired with grouping, and buffers whose numbers are a tier above. That is precisely what primogems should be reserved for.

The tradeoff players get backwards is weapon versus character. A signature weapon is usually worth somewhere around 10 to 18 percent more damage than a strong alternative — for one character. A new support that enables a second full team is worth content you previously could not clear at all. Until both teams exist, the character wins almost every time, even when the weapon banner looks unusually good.

Advanced Edge Cases That Decide Close Calls

Constellation versus new character. A carry's early constellation is usually a 10 to 20 percent personal improvement, occasionally with quality of life that makes a rotation feel better. A new enabler is a new team. The new character wins unless the constellation fixes an energy or uptime problem you are actively suffering from, or removes a dependency and effectively frees a team slot.

Pulling while behind on pity. Pity is not a reason to pull on a banner you do not want, but it is a reason to be careful which banner you build it on. Since pity carries between limited character banners, building pity is only correct if you would accept the featured character as the outcome. If you would be unhappy to receive them, every wish spent improving the counter carries a real chance of a bad result.

The near-miss trap. You are at 70 wishes with 20 left and the banner ends tomorrow. Stop, and carry the pity to a banner you actually want. Arriving at a loved banner with 70 banked pity is far better than spending 20 wishes for a small chance and arriving with zero.

Banked fates versus primogems. Converting changes nothing about value, but the monthly shop reset costs fates rather than primogems, so keep a modest buffer of both instead of converting everything.

Rerun and region banners. Banners that mix older limited characters run on separate counters with their own guarantee logic, and fate points reset when the banner ends. They are only good value when you want a specific older character and can afford that banner's worst case specifically.

Guarantee as an asset. The moment you lose a 50/50, your next limited five-star is certain. That certainty is worth roughly 90 wishes of removed risk, so any plan that spends a guarantee on a low-priority character has burned the most valuable currency in the game.

Track the counter properly. Use the in-game history, not memory: count from your last five-star and record whether you are on a guarantee. Players routinely misremember by ten or more wishes in the optimistic direction, and ten wishes is most of an event reward.

Common Mistakes and the Correction

Saving to 90 instead of 180. Ninety wishes buys one five-star, not a character. If you cannot afford 180 and hold no guarantee, you are gambling — enter it knowingly, with a written answer to "what do I do if I lose."

Spending across two banners at once. Splitting 90 wishes into two attempts of 45 produces two near-zero-probability outcomes instead of one decent one. Pick one and finish it.

Pulling because the four-stars look good. There is no guarantee for a specific four-star, and the five-star you did not want is not consolation. If you would be unhappy with the five-star, do not pull.

Buying a weapon before two teams exist. Weapons are percentage increases; teams are capability increases. Order matters.

Treating a lost 50/50 as wasted wishes. A loss is the moment your next target becomes certain. Let it dictate your next pull instead of tilting into an unplanned banner.

Reactive pulling on day one. Every banner looks irresistible on release day. The fix is structural: never pull before writing down the job the character fills and confirming no owned character fills it.

Forgetting that early income is front-loaded. Exploration primogems are temporary. Spend them on permanent role coverage, then let slower patch income fund wants.

A Two-Patch Execution Plan, With Self-Diagnosis

Before the patch: write down your two intended teams with four named characters each and mark every slot filled by a substitute. Count your current wishes and record whether you hold a guarantee. Write one sentence naming the job you are shopping for. That sentence is the whole plan.

Patch one: spend nothing in the first week unless the opening banner is that job. Log every source of primogems you collect so you can replace the ranges in the income table with your own numbers — most players discover they earn less than they assumed, and knowing early is worth more than any pull. Pull only when the named job appears, and stop the moment it is filled.

Between patches: re-audit. Which jobs remain unfilled? Has a substitute performed well enough to demote the target? Update the sentence.

Patch two: execute the second priority. If both named jobs are filled, stop pulling and bank. Banking is a legitimate and often optimal play, because a large stockpile converts every future banner from a gamble into a decision.

Closing: decide what job you are shopping for, fund the worst case before starting, spend the guarantee only on a job you cannot otherwise fill, and stop the moment it is filled. Do that for two or three patches and the account builds itself.

Closing Synthesis

Pull planning is not restraint for its own sake; it is converting a fixed income into maximum role coverage. The mechanics are on your side once you understand them — pity carries over, guarantees never expire, and the effective average cost of a five-star is far below the advertised worst case. What punishes players is not the 0.6 percent rate; it is making twelve independent one-banner decisions instead of one three-banner plan. Write the sentence naming the job. Budget to the worst case, not the average. Treat the guarantee as an asset worth roughly ninety wishes. And remember the goal was never to own every strong character — it was two teams that can handle any chamber, which is a surprisingly short shopping list if you buy in the right order.