Open Field Protocol

Baffle Defeat Test: Measuring the 5-7-9 Rule in Your Own Yard

The 5-7-9 rule is a population average built on how far a typical gray squirrel can jump. This protocol measures the margin your squirrels leave you, by stepping pole height, baffle size and launch clearance one variable at a time and scoring every attempt.

By the ProBirdFeeder research teamMethod: our research methodsUpdated August 30, 20269 min read

The short answer

You test one geometry variable at a time against a fixed reference setup: feeder base at 5 ft (1.5 m), nothing climbable within 7 ft (2 m), no branch within 9 ft (2.7 m) overhead.
A defeat is scored only when a squirrel reaches and eats seed. Reaching the baffle, or sitting on top of it, is a separate score and is not a defeat.
Measure every clearance with a tape before each window. Estimated distances are the single biggest source of junk data in this protocol.
Log snow depth at every winter window. Snow raises the effective ground level and cancels your height clearance, which is the classic winter failure of an otherwise good setup.
One determined individual is not a population rate. Record approaches and individuals separately, and report counts rather than percentages until you have 20 or more logged approaches per configuration.
Never grease, oil or coat a pole, never use sticky products, never trap or harm a squirrel. The whole test is geometry, not chemistry.

Protocol at a glance

Protocol ID: PBF-03
Version: 1.0
Status: Open for replication
Question: At what pole height, baffle diameter and horizontal launch clearance do the squirrels in your yard actually reach the seed?
Runs needed: 6 or more observation windows per configuration, across at least 5 configurations
Season: Any, with one configuration repeated under snow cover where winters bring it
Time per run: 20 minutes of setup and measurement, then a 60 to 90 minute observation window
Cost to run: Nothing if you already own a baffled pole and a tape measure; about $30 to $60 for a second baffle in a different size
Difficulty: Moderate. The measuring is easy, the discipline of changing one variable at a time is not.

The question this answers#

Our squirrel page gives the standard geometry: a feeder base at least 5 ft (1.5 m) up, at least 7 ft (2 m) of horizontal clearance from any launch point, and at least 9 ft (2.7 m) of clear air below an overhanging branch, with 10 ft (3 m) the honest working number where a squirrel can launch downhill from a deck or roofline. Those distances bracket a gray squirrel's jumping ability in general.

They do not describe your squirrels. A fox squirrel is heavier and jumps shorter; a red squirrel is lighter and climbs things a gray squirrel gives up on. The rule gives the population average, this protocol gives your own margin, and the output is a defeat rate per configuration: how many approaches ended with a squirrel eating seed, under a geometry you measured rather than estimated.

Why the existing evidence is thin#

The 5-7-9 figures circulate widely and are rarely sourced. They trace back to a small body of work on gray squirrel jumping, then get repeated as a fixed rule without anyone reporting how often it holds in a real yard.

Three things make that evidence weak. Reporting bias: people post the day the squirrel wins, not the two hundred quiet days. Unrecorded geometry: someone who believes the feeder is 7 ft (2 m) from the fence has usually not measured it. And missing species and age, so a juvenile that slipped past a narrow baffle and a mature fox squirrel that cleared open ground get filed as the same event.

What you need#

A tape measure reaching 25 ft (7.6 m).
A smooth steel or aluminum pole, height-adjustable or marked with tape at each test height.
Baffles in more than one size: a wraparound or torpedo 15 to 18 in (38 to 45 cm) long, a dome of the same diameter, and a 12 in (30 cm) dome as the undersized rung.
One feeder for the whole study, one seed type from one bag. Black oil sunflower is the sensible default.
Four garden flags, set at the measured clearance distance around the pole, so you can see from a window when something creeps inside the radius.
A timer, thermometer, ruler for snow depth, and a printed datasheet.
Optional: a trail camera on the pole base and nearest launch point.

How to set it up#

Build a reference configuration matching the site guidance, then a ladder of variants each differing in one measurement. Everything not named in the table stays fixed: pole, feeder, seed, fill level, observation times.

Config	Variable moved	Setting	Everything else
A (reference)	None	Base 5 ft (1.5 m), clearance 10 ft (3 m), 18 in (45 cm) baffle topped at 4.5 to 5 ft (1.4 to 1.5 m)	Reference
B	Height	Base 4 ft (1.2 m)	As A
C	Height	Base 6 ft (1.8 m)	As A
D	Launch clearance	7 ft (2 m) to nearest launch point	As A
E	Launch clearance	5 ft (1.5 m) to nearest launch point	As A
F	Baffle size	12 in (30 cm) dome	As A
G (presence check)	Baffle	Baffle removed, one window only	As A

Configuration G proves squirrels are present and motivated, since a run of zeros means nothing if no squirrel came. One 30 minute window, then restore the baffle.

Nothing goes on the pole but a baffle

Never apply grease, oil, petroleum jelly, cooking spray or any slippery coating to a pole. It mats the interlocking barbs that make a feather waterproof and insulating, a bird cannot preen it out, and a bird that cannot insulate itself in cold or wet weather often dies. Never use sticky pastes or glue products, which trap and maim birds, bats and squirrels alike. Never trap, relocate, poison or harm a squirrel, and do not use capsaicin dust, which blows into eyes. If a configuration is being defeated, move the feeder, do not treat the pole.

The procedure#

Run configuration G once, then rebuild configuration A.
Before every window, tape all three clearances again: base to ground, horizontal distance to the nearest launch point, and feeder to the lowest overhead structure. Never carry forward yesterday's numbers.
Walk a circle at 12 ft (3.7 m) and note every surface a squirrel could launch from, including things you do not think of as furniture: a hanging basket, a soil bag, a sapling.
Fill to the same marked level with the same seed. A near empty feeder changes the reward, and the effort a squirrel will spend.
Observe for a fixed 60 to 90 minute window in one of two fixed slots: the first two hours after sunrise, or mid afternoon. Keep the slots constant all study.
Log every approach separately. An approach is any squirrel coming within 3 ft (0.9 m) of the pole base or launch point with the feeder as its apparent target.
Score each approach on the 0 to 4 scale below, and note its route.
Complete 6 windows before changing anything.
Change one variable, re-measure, repeat, working the ladder in shuffled order.
Repeat configuration A at the end. If its numbers drifted, your squirrels learned something and the middle configurations need re-reading.
Repeat configuration A under snow cover, measuring snow depth at the pole base first.

Score the outcome, not the drama

Most arguments about whether a baffle "works" are arguments about what counts. Use this scale only. 0: no approach. 1: approached, did not climb or jump. 2: climbed and was stopped under the baffle, or jumped and missed. 3: got past the baffle or landed on the feeder but did not eat. 4: reached seed and ate. Only a 4 is a defeat. A squirrel sitting on top of the baffle looking annoyed is a 3, and a 3 is the baffle working.

What you record#

One row per observation window:

Geometry as measured: feeder base height, baffle type, diameter, length and top height, clearance to the nearest launch point and that point's height, and overhead clearance, in ft and m.
Window: date, start and end time, minutes, observer, direct watching or camera.
Animals: species (eastern gray, fox, red, Douglas or unknown), age class if you can tell, and distinguishable individuals.
Behavior: approaches, count stopped at the baffle (score 2), count of defeats (score 4), approach route (pole climb, jump from ground, fence or deck, drop from branch or roof), and minutes to the first defeat.
Conditions: temperature, wind, precipitation, and snow depth at the pole base with the resulting rise in effective ground level.
Constants: seed type, fill level, feeder model.

Controls, and the ways this goes wrong#

One individual is not a rate. A single squirrel that has solved your setup generates a long run of score 4 events and makes a configuration look catastrophic. Tell individuals apart where you can, by tail damage, coat or size, and report both the approach-level defeat rate and the number of distinct individuals that ever scored a 4. Those are different findings.

Learning and order effects. Squirrels test a new arrangement hard for two weeks, then give up on ones that never pay. Run the easiest configuration first and you have trained the population before reaching the hard ones. Shuffle the order, and repeat the reference at the end as a drift check.

Snow. The classic winter failure. A 14 in (36 cm) drift under the pole turns a 5 ft (1.5 m) feeder into a 3 ft 10 in (1.2 m) feeder, and a plowed bank becomes a launch point that did not exist in October. Measure snow depth every winter window, in its own column.

Your own presence and hidden changes. Squirrels adjust to a person in the yard, so watch from indoors or a camera at the same vantage point every time. Moving the pole for clearance also changes sun, wind and distance to cover, so note anything that moved besides the named variable.

Bird welfare comes first. Do not empty a feeder or leave a station down in freezing weather to make a cleaner test. Keep a second, properly baffled feeder running throughout. Put nothing on or in the seed, and do not mark, catch or handle any bird or squirrel.

How many runs before the number means anything#

A defeat rate is a proportion, and proportions need events, not hours. Six 60 minute windows per configuration is the working minimum, but precision follows the count of logged approaches. Below about 20 approaches, report raw counts (2 defeats in 11) rather than a percentage: one extra event moves that percentage by nine points. To separate a low rate from a moderate one, aim for 30 or more approaches, which in a quiet yard can take weeks.

Zero defeats is not proof of zero risk. With no defeats in 20 approaches the honest upper bound is roughly 3 divided by 20, about 15 percent, by the rule of three. Report it that way, not as "squirrel-proof".

Submitting your results#

Download the datasheet for this protocol, keep the column names as they are, and fill one row per observation window. Leave a cell blank rather than guess.

Send the file, with a photograph of each configuration taken from the nearest launch point, through the Feeder Log. Negative results matter as much as defeats: a configuration that held for 40 approaches is the kind of data nobody posts. Include your region, habitat and squirrel species, because a defeat rate with no species attached cannot be pooled.