GenetixBayEST. 2009
Breeding · Advanced · 4 min read

Selection and pheno hunting: sample sizes and scoring rubrics

A pheno hunt is a selection experiment: grow enough seeds to have a real chance of finding what you want, clone every candidate before flowering, score each plant on a fixed weighted rubric, and re-grow the top few as clones before you call a keeper. The plants needed follow 1 - (1 - p)^n: a 1-in-10 type needs 29 plants for a 95% chance of seeing at least one. Sex stability is a gate: any plant that herms is culled, whatever else it scores.

Supported
Draft: awaiting expert review Updated
VegFlowerHarvestCureAny system

Being reviewed. Our editors are double-checking a figure in this article against primary sources. Marked passages may change.

Short answer. Decide what you are looking for before you pop a seed, grow enough plants to have a fair chance of finding it, clone every candidate, score every plant the same way, and confirm the winners in a second run. One run of one plant is a lottery ticket, not a selection.

Breeder's standard. Sex stability is a gate, not a score. Any plant that shows pollen sacs or nanners during the hunt is out, along with its clones, however good it smells. Don't use it as a mother, a reversal donor or a pollen parent. Record the week and conditions, and report the line's herm rate. Cannabis is a dioecious species; a herm in a hunt is a defect in that seed lot, and the light leak that exposed it did you a favour.

What's happening

In a segregating seed lot (F2, polyhybrid, landrace), each seed is a different genotype. You are sampling a distribution. Two forces fight you:

  1. Rarity. The combination you want may be 1 in 10 or 1 in 100.
  2. Environmental noise. A plant in the best corner looks better than its genes deserve [1].

Good hunting raises your odds against the first and filters out the second.

How many plants?

Probability of finding at least one plant of a type with frequency p in n plants: P = 1 - (1 - p)^n, so n = ln(1 - P) / ln(1 - p) [1].

Frequency of target type 80% chance 90% chance 95% chance 99% chance
1 in 4 (25%) 6 9 11 17
1 in 10 (10%) 16 22 29 44
1 in 20 (5%) 32 45 59 90
1 in 50 (2%) 80 114 149 228
1 in 100 (1%) 161 230 299 459

Add germination losses and males (regular seed: double it). Practical rule: fewer than 10 females is a sampling, not a hunt.

Before you start

  • Write the brief. Three to five must-have traits and the weighted score below. Changing criteria mid-hunt makes you pick favourites.
  • Mint a GB-IND per seedling and label pots with it.
  • Uniform conditions. Same pot size, medium, feed; rotate positions weekly.
  • Clone everyone before flipping (2 cuts each). Discard clones of rejected plants later.

Scoring rubric (score 1-5 each, multiply by weight)

Trait Weight How to score When
Vigour and health 2 Growth rate, leaf colour, no deficiencies under standard feed Veg
Structure 1 Internode spacing, branching, support needed Late veg / week 3 flower
Sex stability 3 (gate) Any pollen sacs or nanners = cull plant and clones; note week and any trigger Whole flower
Pest and mould resistance 2 Powdery mildew, botrytis incidence, mite tolerance Whole run
Flowering time 1 Days from flip to harvest-ready Harvest
Yield 2 Dry grams per plant (or per m²) After dry
Resin and trim 1 Trichome coverage, leaf-to-calyx ratio Late flower
Aroma (dry) 2 Intensity, distinctness; blind rated by 2-3 people After cure
Chemistry 2 Lab cannabinoid and terpene panel, same lab, same stage After dry
Smoke / effect report 1 Blind panel notes; reported, not promised After cure

Gate traits (sex stability, serious mould) remove a plant regardless of score. Weights are a Grower practice template; adjust to your brief, but fix them before scoring. Structured descriptors in DUS test guidelines are a good model for consistent scoring language [5].

Step by step

  1. Germinate the full lot together; record germination rate (it is QC data for the batch).
  2. Veg cull only for gross faults (weak, deformed, sick). Don't cull on looks too early.
  3. Clone all survivors, label clones with the parent GB-IND.
  4. Flower and score using the rubric, with dated photos at weeks 3, 6 and harvest.
  5. Shortlist the top 10-20% by total score.
  6. Re-run shortlisted clones together (2-4 each). This is where most "keepers" fall out.
  7. Select the keeper(s). Record why. For breeding, also keep the best males or note which females pass traits to offspring (progeny testing).

Selecting males

You can't smoke a male. Use: - vigour, structure, resistance and sex stability (no female flowers; a male that makes female flowers is intersex and is culled); - stem rub aroma (Grower practice, weak evidence); - progeny testing: cross each candidate to the same female and compare offspring. This is the only strong evidence of a male's breeding value.

Common mistakes

  • Popping 5 seeds and calling the best one "the keeper".
  • Not cloning before flower, then losing the winner at harvest.
  • Scoring THC from one sample at one lab and treating it as fixed [4].
  • Letting a strong-smelling but intersex plant through "because the herm was stress". Stress reveals susceptible genetics; sex stability is a gate, not a bonus.

Evidence

  • Sample-size probability and selection theory: Established [1][2].
  • Cannabinoid ratios more selectable than absolute amounts: Supported [3].
  • Lab variability in potency: Supported [4].
  • Rubric weights and male stem-rub: Grower practice.
  • Sex stability as a hard gate (cull herm-prone plants, never breed from them): GenetixBay breeder's standard; see diagnosing hermaphrodites.

Questions this article answers

How many seeds should I pop for a pheno hunt?

A pheno hunt is a selection experiment: grow enough seeds to have a real chance of finding what you want, clone every candidate before flowering, score each plant on a fixed weighted rubric, and re-grow the top few as clones before you call a keeper. The plants needed follow 1 - (1 - p)^n: a 1-in-10 type needs 29 plants for a 95% chance of seeing at least one. Sex stability is a gate: any plant that herms is culled, whatever else it scores.

How do I pick a keeper?

A pheno hunt is a selection experiment: grow enough seeds to have a real chance of finding what you want, clone every candidate before flowering, score each plant on a fixed weighted rubric, and re-grow the top few as clones before you call a keeper. The plants needed follow 1 - (1 - p)^n: a 1-in-10 type needs 29 plants for a 95% chance of seeing at least one. Sex stability is a gate: any plant that herms is culled, whatever else it scores.

What should I score in a pheno hunt?

A pheno hunt is a selection experiment: grow enough seeds to have a real chance of finding what you want, clone every candidate before flowering, score each plant on a fixed weighted rubric, and re-grow the top few as clones before you call a keeper. The plants needed follow 1 - (1 - p)^n: a 1-in-10 type needs 29 plants for a 95% chance of seeing at least one. Sex stability is a gate: any plant that herms is culled, whatever else it scores.

Should I clone before flowering?

A pheno hunt is a selection experiment: grow enough seeds to have a real chance of finding what you want, clone every candidate before flowering, score each plant on a fixed weighted rubric, and re-grow the top few as clones before you call a keeper. The plants needed follow 1 - (1 - p)^n: a 1-in-10 type needs 29 plants for a 95% chance of seeing at least one. Sex stability is a gate: any plant that herms is culled, whatever else it scores.

How many times should I run a cut before deciding?

A pheno hunt is a selection experiment: grow enough seeds to have a real chance of finding what you want, clone every candidate before flowering, score each plant on a fixed weighted rubric, and re-grow the top few as clones before you call a keeper. The plants needed follow 1 - (1 - p)^n: a 1-in-10 type needs 29 plants for a 95% chance of seeing at least one. Sex stability is a gate: any plant that herms is culled, whatever else it scores.

Sources

  1. Suza W, Lamkey K (eds). 2023. Quantitative Genetics for Plant Breeding. Iowa State University Digital Press. iastate.pressbooks.pub/quantitativegenetics/
  2. Suza W, Lamkey K (eds). 2023. Quantitative Methods for Plant Breeding. Iowa State University Digital Press. iastate.pressbooks.pub/quantitativeplantbreeding/
  3. Naim-Feil E, Elkins AC, Malmberg MM, et al. 2023. The Cannabis plant as a complex system: interrelationships between cannabinoid compositions, morphological, physiological and phenological traits. Plants 12(3):493. doi.org/10.3390/plants12030493
  4. Jikomes N, Zoorob M. 2018. The cannabinoid content of legal cannabis in Washington State varies systematically across testing facilities and popular consumer products. Scientific Reports 8:4519. doi.org/10.1038/s41598-018-22755-2
  5. UPOV. 2023. Test guidelines for hemp (Cannabis sativa L.), TG/276/1. upov.int/edocs/tgdocs/en/tg276_01.pdf

Related

DiagnosisSupportedIntermediate · Draft

Hermaphrodite cannabis (bananas and nanners): what to do

Hermaphroditic flowers are anthers (often banana-shaped) or pollen sacs forming on a female plant. Cannabis is a dioecious species, and a well-bred female should stay female; herm-prone plants are mainly the product of bad breeding (repeated selfing, careless feminising, untested 'pollen chucker' seed). Stress such as light leaks or heat reveals that weakness rather than causing it in sound genetics. Fix the environment, remove the plant if it sheds pollen, and never keep it as a mother or breed from it.

EstablishedBreeder · Draft

Quantitative traits and heritability: why most good traits are hard to fix

Yield, potency, terpene totals, vigour and flowering time are controlled by many genes plus environment. Heritability (the share of variation due to genes in a given population and setting) predicts how much of a selected plant's advantage passes to its offspring: response = heritability x selection differential. Cannabinoid ratios are highly heritable; absolute THC % and yield much less so.

EstablishedIntermediate · Draft

Mendelian genetics for growers, with Punnett squares

Traits controlled by one gene follow Mendel's rules: each parent passes one of its two alleles at random, so a cross of two heterozygotes gives a 1:2:1 genotype ratio and, for a dominant trait, a 3:1 look. In cannabis the clearest single-gene examples are autoflowering (recessive) and the THC/CBD chemotype switch (close to codominant).

EstablishedIntermediate · Draft

F1, F2, F3 explained: what filial generations really mean

An F1 is the first generation from two parents; it is uniform only if both parents were true-breeding (inbred). An F2 comes from crossing F1 siblings and shows the most variation. Each later generation of sib-mating with selection increases uniformity slowly. In cannabis, most 'F1 hybrids' are crosses of heterozygous clones, so they behave like segregating F2s.

SupportedBreeder · Draft

Breeder record-keeping and naming ethics

Keep a dated record of every plant, cross, batch and test, linked by permanent IDs, so any seed can be traced to its parents. Name lines honestly: a new, unique name for new genetics, the original name only for the original genetics, clear labels (F2, BX, S1, feminised), and never a famous name on something it isn't. Names are claims; records are the evidence. Sex stability is part of the ethics: cull herm-prone plants, never breed from them, and publish every batch's herm rate.

SupportedAdvanced · Draft

Clone-only cuts and how to preserve them

A clone-only is one selected genotype kept alive by cuttings. It can be lost to pests, disease, hop latent viroid (HLVd) or accident, and it accumulates rare somatic mutations, which are not the same thing as a loss. Cuts from different parts of a plant can also behave differently (topophysis), so take them from consistent locations. Preserve it with clean, tested mothers in at least two locations, regular fresh cuttings, a tissue-culture backup where possible, and seed (S1s or crosses) as genetic insurance, not as a copy.

Was this helpful?

Tell us by joining the waitlist: Founding Members shape what the Professor writes next.

Yes, keep them comingNot quite: I want more
Professor Greenleaf

Grow with the Professor.

The GenetixBay community is coming: honest grow journals, permanent GB-IDs for genetics, and a Professor who answers with citations. Founding Members get first access and the lowest numbers in the archive.

Join the waitlistFund the Bay