Early physics based concept
designs.
New game ideas, built far enough to play and judge. Unfinished on purpose, so they are cheap to change or bin.
Ricochet.
One button, one steel ball, and a machine.
Play it here
Space to fire · drag to look around · D for debug
You press FIRE. A heavy chrome ball gets launched into one section of an enormous working machine. Gears flick it, a hammer flattens it, a lift catches it and fires it back across. Two to four seconds later it lands in a multiplier.
The result is picked before the ball moves. The machine then plays a real physics route that ends in that exact spot, so it's showing you the answer the long way round. Ready for provably fair seeds when CoinPoker wants them.
- 100×Trips a lever, rides a rail behind the flywheel and opens the vault. The camera leans in. Rare.
- 10×The high hatch. Every route there goes under the drop hammer first.
- 2×The boiler hatch. Usually off the great gear, sometimes after a ride up the lift.
- 1×The glass tube at the bottom. Money back.
- 0×The scrap hopper. The house thanks you.
Why I don't think it's worth developing.
I've spent a fair bit of time on this one and I'm glad I built it, but I don't think it should go any further as a game. Three reasons.
You press one button and watch. There's no decision to make, so when it lands in the 10× it doesn't feel like you did anything. The win just happens to you, and I think that takes most of the buzz out of it.
The first ten shots are great. By about the thirtieth you've seen most of the routes, and you're waiting three seconds for a result Limbo or Dice would give you instantly.
It feels more like a tech demo or a screensaver than something people would sit and play for an hour. Nice to look at, but I can't see it holding anyone.
- Result first, physics second works. The result is picked before the ball moves and the ball still takes a real physics route to it. None of the few hundred test rounds I ran failed, so a physics game can still be provably fair.
- The world does a lot of the work. Going from a grey test box to the machine made the exact same mechanic far more watchable. That's worth remembering for every game we do.
- The browser can take it. A 3D world with moving parts and physics running 480 times a second plays smoothly on desktop. I haven't tried it on a real phone yet.
- It costs time. Every route is worked out in advance, so any change to the machine means recalculating them all. The big moving parts, like the hammer, sometimes have to nudge the ball to keep it on its route.
- Use physics for the reveal. On top of something where the player actually decides things, like Blocks, a Plinko upgrade or a bonus round, I think it adds a lot.
- Anything that drops or fires into a result could use this setup. Plinko, pachinko, coin pushers, claw machines. The result-first system carries straight over.
- Build one great section. A single dressed section like this one is where the effort pays off. Full 360° sets get expensive fast.
- Keep rounds short. Two to three seconds is about the limit. Any longer and people are just waiting.
Launch Test.
A car, a wall, and one very brave dummy.
Play it here
Space or CRASH to play · D for debug
You press CRASH. A barrier shoots up out of the road, the car hits it at about 130 km/h and the crash test dummy goes straight through the windscreen. The further he flies, bounces and slides, the bigger the multiplier. Really big hits can knock an arm or a leg off.
Same trick as Ricochet. The result is picked before you press anything, and the game finds a real physics route that ends in that zone, so pressing early or late makes no difference. I ran it 100 times and every round landed where it was meant to.
This is a very early concept. I built it quickly to find out if watching the dummy go is fun enough to do over and over. Pretty much everything you see is placeholder. The car, the dummy, the track and the sounds all need proper art and a lot of polish, and the maths isn't done.
- 10×Bounces off the airbag, flies off the ramp or rips through the billboard, then the giant fan blows him the rest of the way. Rare.
- 5×Clears the crate tower. Usually ploughs into the barrel pyramid.
- 2×Clips the big road sign or smashes into the crate tower.
- 1×Clears the cones and often gets stopped by the tyre wall. Money back.
- 0×Flops onto the bonnet or skids to a stop in the cones.
Early days. It needs a lot of work.
I think the idea has legs (sorry). He's funny to watch, and the bit at the end of each round where you're wondering if he's done yet is where the tension is. But it's rough, so please judge the idea rather than the look.
- Proper art for the car and the dummy. Right now they're drawn in code as stand-ins.
- A much busier track. More set pieces, moving machinery and things to smash, plus the higher zones built out.
- Bigger crash moments. A proper windscreen smash, a crumpling bonnet, more dust and debris.
- Better camera work. Slow motion on the big hits, and tighter framing so he doesn't go tiny at the top of the big arcs.
- Real sound. The beeps and thuds in there are placeholders.
- The maths. The multipliers and odds are placeholders. No real RTP yet.
- 25× to 1000×. You can see the signs for them, but those zones aren't built yet.
- Near misses. Sliding to a stop just short of the next zone only happens now and then. I want more of those.
- Phones. The layout works on a phone, but I haven't tested it on a real phone yet.
Human Cannonball.
Aim the cannon, fire the man, flatten the castle.
Play it here
Drag to aim · Space or FIRE to play · D for debug
You set the cannon's angle and a dotted line shows where it's pointing. Press FIRE and the crash test dummy flies across the sea, maybe off a gull or a crow, and slams into three linked towers made of stone, wood and ice, with rope bridges and hanging buckets. Everything after FIRE is the physics engine. Nothing steers him and nothing steers the collapse.
Coins only pay if he hits them himself. Debris knocking them about doesn't count, so he can bring half the castle down and still win nothing. That keeps the smash and the money separate: he reaches the towers on about 9 shots in 10, but medium risk pays on 4 in 10.
Under the hood it's the Blocks system. I simulated 200 shots in advance for every angle. When you fire, the game picks one for your angle. The bits you can't see are the exact powder charge, the spin and where the birds are. Then the physics plays it out. Every angle pays 96.5%, so aiming changes the show, not the odds. Before each shot your device replays it to check it lands the same coins.
This is an early concept. The background is the painted plate, but the towers, dummy, cannon and sounds are still first pass.
- TicketThe crates with a golden ticket. 3×, 8× or 30×.
- GoldThe coins he reaches least often, up high and deep inside. 2×, 4× or 12×.
- SilverThe middle floors. 0.8×, 0.7× or 0.5×.
- CopperThe ones he hits most. 0.4×, 0.3× or 0.1×.
- BirdsWhite gulls bounce him on. Black crows knock him into the sea.
Every shot is a smash. Most of them lose.
The physics side is doing what we wanted. He hits the towers on 90 to 95% of shots at every angle, and something always comes down. Low, medium and high risk pay out on 55%, 40% and 24% of shots, with top wins of 4.6×, 9.2× and 30.4×. Every angle lands on 96.5% exactly.
- Proper art for the towers. They're drawn in code. They need painting to match the background plate.
- A real character. The dummy and cannon are stand-ins.
- Bigger collapse moments. More dust and debris, splashes, and slow motion when a tower goes.
- Real sound. The booms, squawks and splashes are placeholders.
- Angles below 30°. Down there he only reaches the cheap coins on the first tower, so a fair 96.5% isn't possible. The cannon is limited to 30° to 65° for now.
- Bigger top wins. Medium tops out at 9.2×. More ticket crates deep in the towers would push that up.
- Rounds run long. About 7 seconds including the collapse. Could be trimmed.
- Phones and browsers. Tested in Chrome only. Safari and real phones still need checking.
More coming.
Next concepts land here as they're built.