Why sanbokan resists an even split

Sanbokan is not a straightforward citrus. It carries more bitterness in the albedo than a mikan and a sharper acid than a yuzu, and both of those scale unevenly as you add more of it to a blend. Doubling the proportion does not double the bitterness — past a point it jumps.

Pear does the opposite. Maizuru pears are gentle almost to a fault: add more and the blend does not get sweeter so much as vaguer. The two failure modes are not symmetrical, which is exactly why an even split is the wrong starting assumption.

How we ran the trials

Twenty-two ratios across three weeks, in five-point steps from 30/70 through to 70/30, with a few half-steps where things got interesting. Same growers throughout, and where a lot changed mid-run we re-ran the neighbouring ratios rather than compare across lots.

Each ratio was pressed fresh, poured into identical glasses, and tasted blind by three of us at pressing and again at four hours. We wrote one line each before discussing, because talking first turns three opinions into one.

Where 50/50 fell down

The even split was pleasant and instantly forgettable. Nobody disliked it. Nobody could describe it afterwards either — the pear covered the citrus without adding anything of its own, and what was left was sweet, faintly floral, and anonymous.

That was the useful failure of the whole exercise. It told us the question was not how to balance the two but which one was supposed to lead.

The range that worked

Everything worthwhile sat between 50/50 and 60/40 in favour of the pear. Below that the sanbokan took over by the four-hour mark. Above it the blend went flat.

55/45 was where the citrus stayed legible without dominating — you notice it in the finish rather than the first mouthful. That narrow band is also why we re-taste every new lot: a slightly sharper sanbokan can push 55/45 into what 50/50 tasted like the week before.

Three trials worth recording

The 35/65 pear-heavy trial looked promising on the first pass and collapsed on the second. Something in the pear — probably sugars that had not settled after pressing — turned cloying by the four-hour mark. We did not repeat it.

The 70/30 citrus-heavy trial was bracing at pressing and undrinkable by mid-afternoon, which is the bitterness curve doing exactly what we expected.

A 52/48 half-step was, honestly, indistinguishable from 55/45 to all three of us. We kept the rounder number because a ratio you cannot execute reliably at the press is not a real ratio.

What the method cannot tell you

We knew which sample was which on the second tasting, which is a genuine flaw. A properly blind repeat would be better and we have not run one.

We also tasted at different times of day depending on the shift, and palates are not consistent across a day. The finding held across three weeks, which gives me some confidence, but I would not present this as anything more rigorous than careful bar work.

How it ended up on the board

55/45, pear leading, with the ratio re-checked whenever a new lot of either fruit arrives. The number on the board is fixed; the tasting behind it is not.

If the sanbokan from a particular grower comes in sharper than usual, we will shift to 58/42 for that lot and say so on the card rather than pretend the recipe is immutable.