Sourdough, as a dataset.

Five doughs I bake regularly, in baker's percentages, with the starter's flour and water counted where they belong. Set the sliders, read the weights off the scale.

5Formulas

60–91%Hydration range

8.3–22.2%Inoculation range

−1.37Fitted exponent

Calculator

Weigh it out

Pick a formula, then move any slider off it. Everything else recalculates.

—

Dough weight—

Hydration—

Inoculation—

Warm phase at 21 °C—

Fridge temperature—

Ingredient	g	% flour

TIME IN THE FRIDGE

—

Convention

Half the starter is flour.

The starter runs at 100% hydration, so every gram of it splits 50/50 into the flour and water totals. Percentages that ignore this are off by a few points, and more at 20% inoculation.

Two ways to quote the starter. Inoculation is starter over total flour, the slider. Prefermented flour (PFF) is the share of flour already fermented, exactly half of that for a 100% starter. PFF drives the rate; at a fixed starter hydration the two are interchangeable. A stiff starter breaks that: 200 g at 50% carries 133 g of flour, not 100 g.

Flour type of the starter

Mine is a spelt starter; the rye loaf is fed from a separate rye starter. That determines which flour row the starter's contribution lands in, which is why the spelt formulas read 100% spelt rather than 92%.

The model

Three hours warm ≈ thirteen hours cold

Fermentation is a rate, so a dough needs a dose, not a duration, and the two phases buy that dose at very different prices. At Q₁₀ = 2.5, an hour at 21 °C does what 4.3 hours in the fridge do. The 3-hour bulk is already worth half a day of cold.

Charging the warm phase first and fitting the rest against the three spelt bakes gives dose = 1378 · i⁻¹·³⁷ fridge-hours, R² = 0.97. With three points the exponent lands anywhere from −1.1 to −1.5 depending on Q₁₀ and warm-phase length, so the page refits it live rather than hardcoding it.

Which is why the chart carries a band, not a line. Every bake is an interval (‘6–24 h’ is a fourfold range on its own), so the same points support a line through the window lows and one through the highs. Swept across a defensible Q₁₀ range, those two are the shaded region: ±50% at 8% inoculation, a factor of two at 13%, an order of magnitude by 22%.

Formula	Inoc.	PFF	Salt	Hydr.	Fridge	In fit

Three caveats. Salt is not constant in the series: 3.14, 2.94, 2.86% of flour, falling as inoculation rises, and salt slows fermentation, so some of the exponent is salt. Temperature is not in the data; Q₁₀ = 2.5 is textbook. The other three bakes share one 6–24 h bracket and cluster at 17.9–22.2%, so they are excluded, though fitting all six gives 965 · i⁻¹·²² at R² = 0.95, which agrees.

Identical for all five

Method

Feed the starter, hold it above 21 °C. Next to the coffee machine is a real answer.6–12 h before
Combine everything to the weights above.+0:00
Rest. Autolyse: the flour takes up the water.15 min
Mix again to full incorporation.2 min
Bulk at room temperature. Stretch and fold every 30 min. Optional, but it buys structure in the wet formulas.3 h
Second rest at room temperature, or into the fridge for the window on the chart. The cold phase is where the acidity develops.3–72 h
Lid on, 275 °C. Trapped steam does the oven spring.20 min
Lid off, 180–200 °C, bake to the colour you want. That last call is yours, not the model’s.15–20 min

Building a starter from scratch, 12 days OpenClose

One jar, one feed a day, 21 °C minimum. The stop condition is behavioural, not the calendar: keep going until it rises and falls predictably between feeds.