Reverse Engineering Crazy Taxi, Part 2
wretched.computer
wretched.computer
Then the PC/PSX game us free roam games such as Road Rash 3D and Driver. Crazy Taxi gave us a whole bigass city full of details (even trainways and tunnels among a subway) instead of a countryside with few buildings here and there. Or, contrry to Driver, CT had cities without being restricted to less than 10 buildings copied and pasted across the map rendering the same store again and again.
Yes, I'm aware of MUDs, text had no restrictions. And games like that low poly free roaming game in the Amiga and that weird city driver in the ZX Spectrum, but the Amiga one was barely a 3D demo with no textures at all and the game looked pretty empty.
Take me to the Fila store!
We had it on the Dreamcast, and it was one of our most-played games. I remember for YEARS trying to 100% the Crazy Box... but because we didn't know how to do the boost thing, it was just impossible. Specifically the Crazy Jump, there's just no way to even come close without doing the boost. Literal definition of insanity, us swapping the controller back and forth for hours trying to wiggle the car into the corner of the launch platform to get a little extra run-up and still failing.
The Crazy license ($20k) is an interesting benchmark. In one sense it's the end of the game; there's nothing left to do but chase even higher scores. But in a competitive sense it's just the beginning, as some players managed to score over $100k.
There's a point around $30k or $40k, I forget exactly, where you start running out of customers near popular destinations. So you spend more and more time driving without a passenger, which drains your clock and quickly kills your run. Simply driving better isn't enough to get past that hump.
So when I plateaued there, I went online looking for resources and discovered that the customers aren't random, they're mostly deterministic, and that hardcore players reverse engineered the patterns to be able to predict where customers are going.
The city is essentially one big loop, even though it feels like an open map. So the optimal strategy is to keep moving "forward" in the loop, which allows you to spread out your customer pickups. You can't do this unless you can predict where customers are going and (mostly) only take the ones wanting to go forward; if you just pick up whoever's nearby, you fall into a trap of going back and forth to the same few locations over and over, eventually creating dead zones in the map with no more customers while leaving other locations untapped.
Had you told me upfront that memorizing customer patterns was necessary for high level play, I'd have noped out hard because that sounds absolutely miserable, haha. But by the time I found out about it, I was already familiar enough with the map that it felt more like an a-ha moment than a chore.
I think the biggest pitfall is the highway customers, going from the baseball stadium or the station to downtown and back. Yeah, those are valuable customers and going back and forth will rack up your score in the short term, but doing that will kill your run once those locations dry up since you'll end up with a huge dead zone. That's what I did before I knew better. There actually does reach a point where driving passengerless across the highway becomes necessary, but that's for high level play (maybe $80k+), otherwise it's to be avoided like the plague.
The site with all the notes is still up! I could have sworn it went down at some point.
Are you not wasting 8 bits using a Qn.m of 8.8? If you values are only 0-1?