Gakumas Memories (メモリー): What They Store, How to Use and Filter Them

When a produce run ends, the game saves the result as a memory. That record holds:

• The idol you raised and her final stats (Vocal / Dance / Visual, stamina).
• The skill cards set on the memory — this is the part you reuse in a later run.
• The P-items picked up during that run.

The community simulation engine gakumas-tools models a memory in exactly this shape: one P-idol, a stat block, up to 4 P-items and up to 6 skill cards (plus each card's customisations). A memory is therefore not "one card" — it is a whole dossier, and its value comes from the combination inside it.

1. Contest. Contest is asynchronous PvP and the units you field are memories, not your live idol. A memory's contest power comes from its stats, stamina, P-items and skill cards — the formula is broken down in the contest guide, and you can price any configuration yourself in the Contest Simulator.

2. Bringing it into a new produce run. When a memory is slotted into a produce line-up, its skill cards are added to your deck deterministically. That is the only way to "order" a specific card instead of praying for the draw — so if your plan has a couple of core cards, aim to build (or keep) memories that carry exactly those.

Many memories also come with parameter bonuses for the run, much like support cards: some add a percentage, others a flat value. In game, a % mark at the top-right of the parameter icon tells you it is the percentage kind. Rule of thumb: put the % memory on the idol's main parameter and the flat one on the parameter you care about least.

There is no "objectively best memory" — only memories that fit what you are about to play. The practical method: write down the skill cards and P-items you want to bring into the next run (for the plan you are running), then see which memory matches the most of them.

The gakumas-tools engine does exactly this with a search score: a memory that holds the exact card you asked for gets full credit; a memory that only holds the un-upgraded version when you wanted the upgraded one (or vice versa) gets partial credit; and cards higher up your wish list weigh more than the ones below. You do not need the tool to use the principle: prioritise the memory that matches your most important card, and accept near-matches on the rest.

Do not dismiss a memory whose skills look mediocre but that carries many parameter bonuses — in produce, stats are what multiply into your evaluation score, so that kind of memory is still worth a slot.

Rent memories. Having no good memories early on is normal. The quick patch: follow strong players and rent their memory. Rentals are capped (Japanese community wiki lists a maximum of 3 per day), but a single borrowed card can outclass everything you own.

The "deliberate plan mismatch" trick. Memories push cards into your deck — which sounds great, until the memory only holds weak R cards. Then it is diluting your deck: every junk card is a wasted draw. While you still lack decent memories, you can deliberately slot a memory whose plan does not match in order to reduce how many cards get added — a thin clean deck beats a thick junk one. This is an early-game trick, for when good memories are still scarce.

Good memories do not fall out of the sky — they are the output of a produce run you planned. Three habits:

• Decide the loot before you start. Before hitting produce, ask: which skill card or P-item am I trying to take home this run? A run played on autopilot produces a memory that is good for nothing in particular.
• Raise stats by raising difficulty. Stats are the heaviest term in a memory's contest power, and they are what a memory hands forward most reliably. Higher difficulty → higher stat ceiling → better memory.
• Price it in a tool before you burn a run. Use the Contest Simulator to see how much power a swapped P-item or card is actually worth — far cheaper than replaying a whole produce.

To see which support cards feed better stats and cards into a run, browse the support card list.