Mini Golf Strategy: Pull Length, Bounce Cost and the Two-Putt Map (2026)

1. The course in numbers: nine holes, par 28, one field

The build ships nine hand-drawn holes on a fixed 640×480 canvas. The playable field is the rectangle x 24–616, y 24–456, the ball is 14 px across (radius 7) and the cup is 28 px across (radius 14). Nothing is generated at runtime, so the whole course can be counted:

Two facts fall out of that table and explain most of the course. The pars sum to 28 — the header, the HUD and courses.js all agree, although the package's NOTICE.md still says 27, a documentation slip with no effect on play. And the longest tee-to-cup distance, 504.8 px, is shorter than one full-power putt, so distance is never the problem on this course. Position is. Seven of the nine direct lines are blocked, two of them by water, and the two open ones are the two holes where the naive shot actually works.

2. The shot model: pull length, speed and roll

The control is a slingshot: press anywhere, drag back, release. The drag vector, not the ball position, sets the line, and its length sets the power. A 190 px pull is full power and launches the ball at 980 px/s; the minimum accepted pull is 5% of that, 49 px/s.

Once the ball is moving, the only thing slowing it is an exponential decay of 1.7 per second on grass, and the ball is declared stopped below 5 px/s. That gives a closed form — roll = v/1.7, minus a couple of pixels of discretisation — and the replica matches it:

Read the second and third columns together and the control becomes predictable: the roll is linear in the pull, so half a pull is half the distance, and the whole 592 px field is reachable at about 85% power. The 50% row is the useful calibration: from the hole-1 tee a half pull ends at x = 401, which is 119 px short of the cup — and the live page agreed to the pixel.

3. A putt dies at the end, which is why short is the expensive miss

The drag is proportional to speed, so the ball loses the most speed where it is fastest — right after the launch — and almost nothing where it is slowest. The last 100 px of a full-power putt take about 1.2 seconds; the first 100 px take 0.11 seconds. Two practical consequences follow.

The end of a putt is forgiving in distance and unforgiving in direction. A ball rolling at 60 px/s moves one pixel per frame, so a line that is 5° off will still be 5° off when it reaches the cup.

Overshooting costs less than you think, undershooting costs a whole stroke. Because speed decays exponentially, adding 6% more pull adds only about 34 px of roll at the top end (75% → 100% is 142 px, but 94% → 100% is only about 30 px). A ball that stops 30 px short leaves you a 30 px tap-in, but the stroke is gone either way. When in doubt on a hole with no water, take the extra 5%.

4. Walls cost 40% per bounce

Every wall is an axis-aligned rectangle. The collision keeps 62% of the speed component perpendicular to the face and then multiplies both components by 0.97, so a head-on bounce keeps 0.62 × 0.97 = 60.1% of the speed. Measured on the hole-3 block: a ball arriving at 952.6 px/s left the wall at 572.9 px/s, exactly 0.601.

That number is the reason bank shots are a last resort rather than a technique. Two bounces leave you with 36% of the launch speed, which after the 1.7 drag is roughly a 78 px roll from a full pull — you have spent a stroke to move the ball nowhere. Banks are worth it on exactly one hole family here: the ones where the second shot has to come from behind a block, which is the 2-shot route on holes 6 and 7 measured in section 6.

The gaps are part of the same arithmetic. The ball is 14 px across, so a 16 px slot between a wall and the field edge is a 2 px corridor: it exists geometrically, it does not exist in practice. Holes 2, 4, 6, 8 and 9 all have such a 16 px slot above a block, and on all five the real opening is elsewhere — 136 px under the hole-2 wall, 146 px on each side of the hole-3 gate, 166 px over the first hole-4 block, 116 px under the hole-6 crossbar, 80 px through the hole-8 gate, 120 px through the hole-9 gate.

5. What the naive shot does on each hole

Here is the measurement that reorganises the whole course: from every tee, aim straight at the cup and pull to 100%. The replica and the live page give the same outcome.

So the course is not nine holes of aiming at a flag. It is two holes where the line is the answer (1 and 8), two where the line is a water hazard (5 and 9), and five where the first shot has to be a position shot around a block.

6. The two-putt map, measured

To turn "position shot" into something you can copy, the harness searched every hole: first shot over 21 aim angles (±50°, 5° steps) × 14 powers (35%–100%, 5% steps), then a second shot from wherever it stopped over 13 angles × 9 powers. It found a plan inside two shots on eight of the nine holes. The routes below are the ones the search found first; they are examples that work, not the only ones.

Note what the plans have in common: on six of the seven multi-shot holes the first shot is aimed away from the cup — up and to the left on 2, 3, 5, 7 and 9. That is the opposite of the instinct the drag control invites. The first shot is a placement, the second shot is the putt.

7. Water: two strokes for nothing

The water rule is the most expensive thing in the build. A ball that enters a pond gets one penalty stroke and is returned to the spot the shot was played from. Since the shot itself was already counted, a water ball costs two strokes and gains zero distance. Measured on hole 5: from the tee, the straight full-power putt reached x = 262 in 0.17 s, played the penalty and put the ball back at (110, 240) with the stroke counter at 2.

The geometry makes this a trap you choose, not one you fall into. Hole 5 is a 420 px hole with a 120×160 pond sitting exactly between the tee and the cup; hole 9 is a 420 px hole with a 70×120 pond and a wall gate. In the 915-shot sweep per hole, 421 shots (46%) on hole 5 and 146 (16%) on hole 9 ended in water — and every one of them was aimed within ±60° of the cup line. The pond is not an obstacle you skirt by a little; it is the middle of the hole. On both holes the route is over the top (hole 5 at y ≈ 90, hole 9 at y ≈ 32) or under the bottom, and the two-putt plans above do exactly that.

One more arithmetic point: because a water ball replays from the same spot, retrying costs the same as the first attempt. There is no reason to "play it safe" on the second try — the safe shot is the miss that finds water.

8. Sand: four and a half times the drag

Sand is not a wall, it is a different drag constant: 7.5 per second instead of 1.7. Same launch, same pull, 4.6× less roll — a full-power putt rolls 122.1 px in sand against 565.5 px on grass, and it stops in 0.72 s instead of 3.12 s. The ratio holds at every pull tested, from 5% to 100%, so the sand does not "eat" a shot in a fixed way: it shrinks the whole shot to about a fifth.

That makes sand a distance tax with a shape. A ball that touches sand for the last 40 px of its run loses only a few px; a ball that starts in sand loses 80% of the shot immediately. In this course book the sand never sits between the tee and the cup — it sits to the side (hole 2 at x 430–540 below the cup line, hole 4 at x 300–370, hole 6 at x 380–480, hole 7 at x 120–230, hole 9 at x 360–440) — so the practical rule is simple: if the line you want clips a bunker, pick the other line, because a clipped bunker turns a 300 px putt into a 60 px putt and leaves you a longer second shot than the one you were trying to avoid.

9. What the cup accepts

The cup is 28 px across, but the ball does not have to hit its centre. The build drops the ball when it is within 12 px of the cup centre and moving slower than 520 px/s. Above that speed the ball skates over the hole. Between 26 px and 12 px there is a magnet: below 260 px/s the cup pulls the ball toward the centre at 260 px/s², which is why slow balls curve in.

Measured from 60 px out, the boundary is sharp: the fastest ball that dropped came in at 517 px/s, and the first one that skated past came in at 521 px/s. So the gate is real and it sits exactly where the source says.

Two consequences:

Aim error is cheaper than power error near the cup. Any ball that arrives slower than 520 px/s within 12 px drops, and the magnet widens the effective target to 26 px at the low end.

But from a tee, the power window is wide open on top. On hole 1 the straight putt sinks at every pull from 67% to 100% — 34 of the 51 tested powers. There is no upper edge to hit; the whole risk is being short. That is a property of the drag, not of the cup: over 400 px the ball loses 680 px/s of speed, so by the time it arrives it is always slow enough.

Aim is the tight axis. The same sweep with 2° of aim error closes hole 1's window at 83% (17 of 51 powers) and hole 8's at 86%, and at 4° off the line neither hole sinks at any power. A 12 px target 400 px away subtends 1.7°, so on the long holes the line matters more than the pull — the opposite of what the wide power window suggests.

10. Par, score and the round record

Par is 28 over nine holes, and the pars are 2, 3, 3, 4, 3, 3, 3, 3, 4. Against the measured two-shot map that reads generously: eight of the nine holes fall in two shots with a coarse grid, which is at or under par everywhere except hole 4 (par 4), where the search did not find a two-shot route and the second block runs to the bottom edge. A clean round is therefore 18–20 strokes, and a round in the low 20s is a normal good result rather than a fluke.

The HUD tracks hole, par, strokes and the running score against par, and the score label flips to red when you are over par. The round record is stored in localStorage under playnesta:mini-golf:best as total strokes and relation to par; it updates only when the total improves. Restart (button or the R key) clears the round and starts again at hole 1 — it does not clear the best round.

11. A one-minute checklist

1. Look at whether the straight line is open. On this course it is open on two holes.
2. On a blocked hole, pick the placement shot first and say out loud where it should stop.
3. For a putt under 200 px, pull 35–45%; for 200–300 px, 50–60%; for 300–420 px, 65–75%; only the long diagonals need 85%+.
4. Never aim within 20° of a pond's centre line unless the plan is over the top.
5. If the line clips sand, choose another line; a clipped bunker costs about 80% of the shot.
6. Take the extra 5% of pull on any hole without water.
7. After a wall bounce, recompute: you have 60% less speed than you think.

FAQ

How do you play Mini Golf?

Press anywhere on the course, drag back from the ball and release. The drag vector is the shot: its direction is the line and its length is the power, up to a 190 px pull. The ball rolls with friction, bounces off the walls at 62% of the perpendicular speed, slows down four and a half times faster in sand, and a ball that stops in water costs a penalty stroke and replays from the same spot. Nine holes, par 28, one round against par.

How far does a full-power putt roll?

565 px on grass, which takes 3.1 seconds. The field is 592 px wide, so full power covers the whole course and about 85% is enough to cross it. In sand the same pull rolls 122 px and stops in 0.7 s.

Why does my ball keep stopping short?

Because the drag is exponential: half a pull is exactly half the distance, and the last third of a putt looks slower than it is. The measured table in section 2 is the correction — 50% pull is 281 px, not 400. If you are consistently short, add 10%, not 30%.

Is it better to aim or to power the shot?

Both matter, but they fail differently. A directional miss of 5° over 400 px is 35 px wide at the cup — outside the 26 px magnet. A power miss of 5% at the top end is about 28 px short, which is a tap-in. So on this course direction is the expensive error, which is why the placement shot matters more than the perfect pull.

Can I get a hole in one?

On two holes, yes, and not by luck: hole 1 from the tee sinks at any pull from 67% to 100% on the exact line, and hole 8 sinks through its 80 px gate from 72% to 100%. The window is narrow in the other axis — 2° of aim error on hole 1 caps the pull at 83%, and 4° off the line sinks nothing. On the other seven holes the grid found no one-shot line at all, because the block or the pond sits on the line.

Why did my ball stop right at the edge of the cup?

The cup accepts a ball within 12 px of its centre moving slower than 520 px/s. Above that it skates over. Between 12 and 26 px there is a magnet that pulls slow balls in, so a ball that stops at the lip was probably above 520 px/s when it crossed, or crossed 13–26 px wide with too much speed to be caught.

What happens if I hit the water?

One penalty stroke is added and the ball returns to where you played the shot from. The shot you played is also counted, so a water ball costs two strokes for no distance. On hole 5 that happens 0.17 s after the launch if you aim at the cup.

Does sand cost me a stroke?

No, sand costs distance. The drag is 7.5/s against 1.7/s on grass: a full-power shot rolls 122 px instead of 565 px. You keep the stroke, you lose about 80% of the shot, and you now have a longer second putt than the one you were trying to play.

How do the walls work?

Each wall keeps 62% of the perpendicular speed and then both components are multiplied by 0.97, so a head-on bounce keeps 60.1% of the speed (measured: 952.6 px/s in, 572.9 px/s out). Two bounces leave you at 36%, which is why banks are a route only when the direct line is blocked outright.

What is par, and what is a good round?

Par is 28 across the nine holes: 2, 3, 3, 4, 3, 3, 3, 3, 4. The measured two-shot map beats or matches par on eight of nine holes, so a clean round is around 18–20 strokes. Anything in the low 20s is a normal good round; the best round is saved in the browser.

Can I restart a hole?

There is no per-hole restart: the Restart button (or the R key) starts a new round at hole 1 and clears the current score. After a sunk hole the Next hole button appears, and the next hole also advances by itself after 2.2 seconds.

Is Mini Golf the same as putt-putt?

Yes — mini golf, minigolf and putt-putt all describe the same format: a short putting course with walls, bunkers and water instead of fairways. This build is a nine-hole version with an arcade control (slingshot pull) rather than a mouse-follow stroke, so the practice that transfers from real mini golf is the reading of angles and the placement shot, not the swing.