A two-dimensional physics model of the break you can shoot, replay, run two hundred times, and share, with every assumption printed beside it.
Interactive appendix, October 2026Calibrated against The Modern Break's cited researchEvery run reproducible from its seed
One break, every ball traced. Schematic.
The break, one variable at a time.
01 What this is
A model you can shoot
The Modern Break organizes what measured studies, rulebooks, and documented champions say about the break. This appendix is the other half: a physics model that lets you change one variable, hold everything else still, and watch what moves.
Place the cue ball, aim, set the speed and the tip, and break. The model plays the break back in slow motion, shows the cue ball's spin magnified, and lists what fell. Then run the same setup fifty or two hundred times: each run jitters the rack, the contact point, and the speed by the amounts you choose, and the lab reports the spread of outcomes: success, scratches, dry breaks, illegal breaks under your ruleset, and where the cue ball ends up.
Two rules govern everything here. First, it is a model: the disclosure panel inside the lab lists what it models and what it leaves out, and no number it prints should be quoted as evidence about real play synthesis. Second, every run is reproducible: the same setup and the same seed play the same break, bit for bit, on any machine, so a result you find is a result you can send.
The experiments load ready-made comparisons into the lab. The reading guide says exactly what each statistic counts, and Seeds and shared links explains how to hand someone your exact break.
02 The simulator
Run the break yourself
Everything in The Modern Break is drawn from documented evidence. This lab is a two-dimensional physics model you can shoot, not a source of evidence. Place the cue ball, aim, set tip position and speed, and break; then run the same setup two hundred times and read the spread of outcomes. Every run carries a seed, so any result can be reproduced exactly.
The model is calibrated for plausibility: simultaneous ball contacts, sliding-to-rolling cloth friction, english-dependent cushion rebounds, and WPA-spec pocket mouths. Read the disclosure panel before trusting any number it prints.
Model disclosure illustrative model
What is modeled
Ball–ball collisions as simultaneous compliant contacts, with a two-ball restitution target of 0.93
Cloth friction with a sliding-to-rolling transition (sliding μ 0.20, the physics behind draw and follow; rolling resistance 0.03) and spin decay
Cushions: normal restitution 0.70 with tangential friction. In this equatorial 2D model, oblique rail hits may increase spin only through friction-bounded transfer from translational energy; horizontal roll is not coupled to the cushion.
Six pockets at WPA mouth specifications (corner 4.5–4.625″, side 5–5.125″, 2.25″ balls) as capture geometry with jaw tips that rattle
Tip-position spin from ω = 2.5vb/R², with combined contact displacement b limited to R/2, an assumed safe non-miscue domain B.8. Selected speed is cue-ball speed, not cue-stick speed. Strong low-tip shots can retain draw at the rack.
Adaptive substeps capped at 10 microseconds near moving ball contacts; travel limited to 0.05″ per coarse step
What is not modeled
Squirt, deflection, and swerve: sidespin here costs no accuracy; on a real table it does
Ball throw and cling; spin transfer between balls; miscues outside the assumed R/2 contact domain
Vertical dynamics: hop, bounce, jump cues
Cloth nap, direction, and speed; humidity; table level; worn or oversized bar-box balls
Pocket shelf bounce-outs and jaw facing angles (142°/104° simplified)
Rack templates versus triangles beyond the tightness slider; the forceful-strike standard and referee judgment
Calibration. Tip efficiency (COR 0.73 leather / 0.81 phenolic / 0.87 hardest phenolic) and measured pro break speeds (22–26 mph, average 24) come from the project's Break Equipment & Tournament Evidence research 161314; break-strategy behavior (the square hit first, the squat hit, wing-ball geometry) follows the Break Strategy research 1112; pocket mouths follow the WPA equipment specification 26; legality formulas follow the WPA, CSI, and APA rules as cited in Rules 242829. Parameters are chosen for plausibility, not fitted to outcomes. The event-based engine. The lab opens on the event-based engine; the lists above describe the classic 2D model, which is one click away. The event-based engine is a port of pooltool 0.6.0 98 that jumps from contact to contact, and it does model squirt and ball-to-ball throw; it is still flat, so hop, bounce and jumps stay out of both. This lab is an illustrative physics model, not predictive of tournament outcomes. It is not calibrated to reproduce tournament result rates, and no number it prints should be quoted as evidence about real play.
Setup
Event-based is the default: the engine ported from pooltool, which jumps from contact to contact instead of stepping, meets the cushion at its nose height, and adds ball-to-ball throw and squirt. Classic 2D is the model this page was written with; links made before the switch still open on it. The same seed racks and strikes the same way in both. Lab values use this page's cloth, ball and cushion constants; pooltool defaults roll farther and bounce livelier.
Ball flight runs pooltool's unreleased 3D physics: an elevated cue hops the cue ball, balls land and bounce, meet each other and the cushions in the air, and can leave the table. A ball that leaves the table is out of play; a cue ball off the table counts as a scratch. Its cushion is only the rubber nose, with no rail top above it, and it doesn't yet catch a flying ball on a pocket jaw, so it likely overstates balls flying over rails.
Placement & aim
Drag the cue ball inside the kitchen (break box only). Drag anywhere else to aim.
aim 0.0°
Starting points to test; free drag stays available. Left/right is from the breaker's view. Half-diamond placements sit slightly off center 3. Side-rail placements sit near the head string and aim at the 1 111. Exact rail offsets are unverified 19; WNT placements move inside the break box.
Stroke
ball speed24.0 mph
cue elevation0.0°
tip position centered contact offset, % of ball radius
The dashed circle limits combined tip contact to half the ball radius. This is an assumed safe model domain, not a universal miscue threshold. Miscues are not simulated B.8. The faint inner ring is a quarter of the radius: an average draw, spin about 0.6× natural roll BD. Spin off the tip is the ideal strike's 2.5 × offset; Dr. Dave's high-speed video measured up to 1.37× on a clean hit near the limit BD.
The 3D view needs WebGL and didn't get it here. The 2D table is the same simulation.
Cue spin · magnified
Six dark triangles track modeled spin. Fast spin uses a soft, weighted 1/30-second modeled-time exposure along the current spin axis, independent of playback speed. Blur shrinks as spin slows; rpm remains the modeled value. Cloth or rail friction can turn travel into rotation, so spin can rise while total energy falls. Roll magnitude includes draw; side is signed. Playback ends at rest or 12 seconds.
Pocketed this break model outcome
No balls pocketed yet.
Pocket events only, not a scoring or winner decision.
Place, aim, and break. The rack you see is the rack this seed produces.
Monte Carlo
The lab draws balls at true 2.25″ scale. The Modern Break's diagrams draw them slightly smaller so labels fit. Rack “gap” is the looseness slider: zero is a template-tight rack, higher values are a house rack on its best behavior. The cue ball sinks into the pockets; everything else is the model described in the disclosure panel synthesis.
03 Experiments
Six comparisons to run
Each experiment changes one thing. Load A, run 200 breaks, note the numbers; load B and run again. Both arms use the same seed, so the racks and the random jitter are the same. The only difference is the variable under test.
The experiments deliberately print no results. The point is to run them, and anything the model reports is a model outcome to test on a table, not a finding synthesis.
Speed: 20 mph or 26?
The measured pro range is 22–26 mph, and Onoda found speed was not strongly correlated with making a ball13. Does the model agree, and what does the extra speed cost?
Read: Success rate against scratch rate, and how wide the cue-ball scatter spreads.
Head spot or side rail?
In 9-ball the documented placement is near the side rail, cutting the 1 so the wing ball runs to the corner11. Break from the head spot and from the rail with the same seed.
Read: The wing-ball-to-corner rate and the side-pocket rate; then the per-pocket bars.
Center or a hair above
The instruction is center ball or a hair above: the rack bounces the cue ball back, and slight follow stops it near the center of the table instead of following through1. Compare center ball with a tip 15% above.
Read: The cue-ball scatter: where it parks, and how often it scratches.
A tight rack or a loose one
The rack gap slider is the looseness of the rack: zero is template-tight, the top of the range is a house rack on its best behavior. Same break, both racks.
Read: The balls-per-break histogram and the dry-break rate.
The bar box
The same 24 mph break on a 7-foot table. No instrumented bar-box dataset exists, so the bar-box advice is reasoned, not measured unverified11. Here is what the model does with it.
Read: Scratch rate and mean balls per break; the per-pocket bars shift too.
Inside the break box
The World Nineball Tour's break box keeps the cue ball off the side rail (see Rules in The Modern Break); the lab moves a rail placement inside the box. Compare the same rail break under WPA rules and under WNT rules.
Read: The illegal-break rate and the wing-ball-to-corner rate.
04 Reading the numbers
What each statistic counts
A Monte Carlo run is the same setup broken many times, each with its own seed, so the rack gap, the contact point, and the speed vary by the amounts on the Chaos & seed sliders. The panel reports what fraction of those breaks did each thing.
Ball pocketed (success)
Breaks where at least one object ball fell. The cue ball never counts.
Mean balls / break
Object balls pocketed, averaged over the completed breaks.
Side-pocket rate
Breaks where any object ball fell in either side pocket.
Wing ball → breaker corner
9-ball only: the wing ball the lab tracks, in the rack's second row, fell in the near foot corner, the drop the side-rail break aims for.
8 pocketed on break
8-ball only: the 8 fell on the break.
Scratch rate
Breaks where the cue ball was pocketed.
Dry break
Breaks where no object ball fell.
Illegal
Breaks the ruleset's legality check fails. Legal is a pocketed object ball, or four object balls driven to a rail, or (under WPA/CSI and WNT, not APA) three object balls driven across the head string. The forceful-strike standard and referee judgment are not modeled.
3 across head string
Breaks where three or more object balls crossed the head string at any point.
The charts. “Balls pocketed per break” is the histogram behind the mean; a setup that averages one ball can be steady ones or a mix of dry breaks and big runs, and only the histogram tells you which. “Per-pocket counts” totals every object ball by the pocket it fell in, with the near foot corner (the wing ball's pocket) drawn dark. “Cue-ball final position” plots where each break left the cue ball; scratched breaks are counted but not plotted.
Sample size. Fifty breaks show the shape; two hundred steady it. With two hundred runs, a rate near 50% still moves by several points from one batch of seeds to the next, so treat a difference of a few points between two setups as noise until it survives a second seed.
The time window. Each break is simulated for at most twelve seconds of table time. A break still rolling at the cutoff is reported as it stands, and the status line says so.
What these numbers are not
They are not tournament rates. The best-documented professional sample, 422 streamed games at the 2023 US Open, logged 73% successful breaks 36. The model is not fitted to that number or any other outcome, so agreement with it is coincidence and disagreement is not news. Use the lab to compare setups against each other, never against the world.
05 Seeds and shared links
The same break, anywhere
Nothing in the simulator is random in the way a real break is. Every bit of variation (the rack's gaps, the error in the contact point, the error in the speed) is drawn from a generator started by the seed, so a setup and a seed always play the same break.
Seeds
A single break uses the seed in the box as it stands. A Monte Carlo run of N breaks uses the seed with each run's number appended: with the seed break-1, run 0 is break-1:0, run 1 is break-1:1, and so on. The results panel prints the range. To watch any one of those runs, type its seed into the box and press Break: it is the same break the batch counted.
Changing the seed changes the rack and the jitter but nothing else. Keeping the seed and changing one control is how the experiments isolate a variable: both arms see the same racks.
Shared links
Copy setup link, in the lab's Shoot & playback card, puts the whole setup in the page address: game, table, ruleset, speed, tip, the three chaos sliders, the seed, the cue ball's exact position, and the exact aim. Opening the link loads that setup into the lab, so the reader sees the rack you saw and gets the break you got. Positions and angles are carried at full precision, not rounded, so the reader's break matches yours exactly rather than approximately.
What a link does not carry: the cloth color, the playback speed, and the Monte Carlo batch size, which change how the lab looks or how long it takes but not what happens. The link also carries no results; the reader's browser runs the break again. A shared link is checked before it loads: values outside the controls' ranges are pulled back inside them, and anything the lab does not recognize is ignored.
06 Sources
Where the parameters come from
The sources behind the simulator's calibration and its starting points. Numbers follow The Modern Break's research corpus, so a citation here and the same citation there are the same source; gaps in the numbering are intentional.
Instructional & physics: Dr. Dave Alciatore, PhD (drdavepoolinfo.com)
13sfbilliards.com/Misc/onoda_all_txt.pdf · Onoda instrumented study: pro break speeds ~24 mph men (22–26), ~20 mph women; speed not strongly correlated with making a ball measured
14breakrak.com/book-mentions/ · BreakRak book mentions: Kaltofen/Kasten dataset, 24.9 mph male / 19.3 mph female pros measured
98doi.org/10.21105/joss.07301 · Evan Kiefl, “Pooltool: A Python package for realistic billiards simulation,” Journal of Open Source Software 9(101), 7301 (2024); source at github.com/ekiefl/pooltool. The event-based engine is a port of pooltool 0.6.0 (Apache-2.0) physics model