How Many Seeds Do You Need to Find a Keeper?

Watch the episode · 4:35

Breeders grow hundreds of plants to keep one. You've got a pack of five. So what are your real odds of finding a keeper? This is the simple maths behind the pheno hunt: why ten times more seeds won't give you a plant that is ten times better, and why the best plant in your tent might not be the best at all.

The short answer ▶ 0:30

If the plant you want turns up about one time in ten, you need about 30 seeds to be 95% sure of finding it. A 5-pack gives you about a 41% chance. And whatever you grow, take a cutting from every plant before it flowers.

Now, why.

What's a keeper, and your odds ▶ 0:47

First, what is a keeper? It is the plant you'd grow again: the one that beats its brothers and sisters on whatever you care about. Its vigour, its yield, its smell, its finish time.

Think of each seed as a roll of a ten-sided die, and you need a ten. Five seeds, five rolls: a 41% chance of at least one ten. Ten seeds: 65%. Twenty: 88%. Thirty: 96%.

Plant breeders have long worked out how many plants they need to be fairly sure of finding a trait, and a handy rule falls out of the maths: to be about 95% sure, grow three times the odds. One in 10: 30 seeds. One in 20: 60. One in 100: 300.

Source · Sedcole, Crop Sci. (1977)

Those numbers are for female plants. With regular seeds, about half come up male, so you need roughly twice as many.

So how rare is a keeper? That depends on the pack. Seeds from stable, inbred parents come up much more alike, so most plants will be close to the best. Seeds from a cross of crosses scatter in every direction, as the phenotypes article shows. And as we saw in how a new strain is made, wanting three simple traits at once can mean one plant in 64.

Source · de Meijer et al., Genetics (2003)

More seeds, smaller gains ▶ 1:55

Here is the part nobody tells you about big hunts. Picture the best plant in a group, against the average. With five plants, the best is typically about 1.2 spreads above average, where a spread is how much plants usually differ. Best of ten: 1.5. Best of a hundred: 2.5. Best of a thousand: only about 3.2. Ten times the seeds, nowhere near ten times better.

Source · Falconer & Mackay, Introduction to Quantitative Genetics (1996)

That is why breeders grow plants by the hundred. Every small step up the ladder costs far more plants than the last. In one breeding run we read about, 25 candidate fathers were started, and just two were kept.

Luck or genetics? ▶ 2:37

And your best plant? It might not be the best at all.

In a 2021 study, Australian researchers grew 121 cannabis lines in one carefully controlled growing facility and worked out how much of the difference between plants came from genetics. Height: about half. Time to finish: about half. Yield: only about a third.

Source · Naim-Feil et al., BMC Plant Biol. (2021)

So about two thirds of the yield differences came from everything else: a slightly better spot under the light, a warmer corner, a pot that drained better. Even in a research facility. In a home tent, your heaviest plant may simply have had the best seat.

Breeders deal with this by growing their best candidates again, as cuttings, side by side, before they decide. If a plant still wins the second time, the win is far more likely to be in its genetics.

Your own hunt ▶ 3:29

So how should you run your own hunt? Know your odds: about 30 seeds for a one-in-ten plant. Take a cutting of every plant before it flowers, because you only find your keeper after harvest. Give every plant the same conditions, and swap their positions now and then, as researchers do. And grow your winner again before you call it a keeper.

One catch with autos: a cutting is as old as the plant it came from, so you can't keep an auto as a mother. An auto keeper only lives on through its seeds.

About 30 seeds for a one-in-ten plant, roughly double with regular seeds, and more seeds bring smaller and smaller gains. Your best plant may just be lucky, so grow it again before you call it a keeper.