All Startups Unprofitable?

[Luciatrum]

Is it just me, or is currently every startup in the game unprofitable? I've never made a single bit of revenue off of the startups since I got the game a couple of days back and they constantly go into bankrupcy by the end of the week, even when the server cost is at $2 per second...

[Halem c:]

Well when my server for Gamazon costed around 300$/Sec i made over 4k a week.

[GustavFrings]

I just had a Setup for Gamazon with ~100$/Sec and 1600 Users. It was bankrupt in 2 days. Same with the first and second startups but not so may users.

I don't know how exactly the startup system works but I think at the start the lower the cost the better. If the startup then fails I don't know how this works. But if this works at the start and you make a couple of days positive revenue, maybe then create a patch that gets more which should result in higher costs which can be paid by the startup.

It should be like in real life. If you have Startup with like 20k $ start and cant find many users and the belance between Costs of Employees, Server, Ads, etc and income doesnt matches then you go brankrupt.

I will try some trial and error and check if I can approve this.

[GustavFrings]

I'm back from my first trial and error session with the first Startup.

I started a new game at #3, made one job to unlock first startup. Developed a basic scheme with capacity for max 215 users (three "decision color tree"). Later upgraded that to max 245 users which had no effeckt on the startups current users.

The overload seems to be a big factor, if you can keep the overload on 0% your startup should run a specific time. In reallife the lifespan of a company goes like this:

http://www.bbc.co.uk/staticarchive/d1ae6d46953382f759225334d9552da3a5b4d2a3.gif

The same happens with the startup in game. Starts low, get high and finally fails. How long this process takes ingame? I don't know.

Becasue we are just a programmer siting in a dark cellar with a cat tha wears a helmet, we can't save the startup once it's dying. We can only save it from overloaded servers.
Overload means: The servers cant handle all requests by the users which then leads to loss of users cause they tell everyone they know: "Don't use it, its damn slow and you have to wait a long time before u can use it."

For more information and data and my analysis for the first startup checkout my spreadsheet:

https://docs.google.com/spreadsheets/d/1so3L5CtVLibq3gQbE2ZZh5IFXu4CY3ys3GtwGYvNd0Y/edit?usp=sharing

[ShivaFang]

Honestly don't worry about overload. And don't use more than one Load Balancer. Even if your users goes up the server costs will eat you alive.

The first one will become unprofitable after about a week or so just because low number of users. I use 2 decision trees to solve it and then replace them with RGB sort once you get that. Try to buy all cost upgrades. Sell that startup as soon as it is in the red twice in a row (there's some randomness - a single red day isn't enough to panic) You buy in for $100 and sell for about $188 (honestly - barely worth it) I can't think of anyway to optimize that any more to make it more profitable.

All the ones after that are MASSIVELY profitable for me. Use a single load balancer (or QuickSort RGB) If you aren't turning a profit on them then your soloution is inefficient.

Hint: Always do Colour before Shapes - because Colour nodes are usually faster. Basically do your fastest stuff first because that will split the load and then you can do the slower stuff at the end.

One trick is to remake 'Shape Divide' after you beat it to become a slow CST sort. Entrance->Sift_CircleAny->Sift_TriangleAny (you can leave one exit empty) Then you can use that in the second and third soloutions to optimize your soloution (The next two startups use genetics and perceptrons and after that you get a Decision Tree for shapes which is way better than Sifts)

Oh yeah... also there's one problem where the input has 50 green triangles and green triangles aren't used in the soloution. I just trash all greens. It is simply not worth processing time to weed out all those triangles. I cannot find a soloution to that that doesn't cause a major backlog since green triangles skew the balance so much.

You get paid per solved shape I think, with a penalty for incorrect output (except when errors are okay) so trashing good soloutions does lower your profits, but sometimes it's not worth solving for pennies. Something like 1/6 greens are good (5 Square and 5 Circles vs 50 Triangles IIRC) and it's just not worth it. There's no other penalty for trashing a good shape.

[GustavFrings]

Well, that overload and the maximum users your scheme can handle doesn't impact the startups life is kinda bad. Whoith your setup @ShivaFang i was able to run the first and second startup for 5 weeks and made some good money out of it. Tahnks for you experience with the game!

I hope someday the overload and the maximum users your scheme can handle will infact the game experience.

[ShivaFang]

Cool beans. I'm glad I could help. I was worried I wasn't explaining things properly.

I think the problem is that server costs scale worse than they actually benefit you. So for example adding a second load balancer only works on half your soloution (because it's after your first one), so you need 2 load balancers to double your speed, but that quadruples your costs (or something... not entirely sure how load balancers scale costs)

If I'm right - it totally explains why people go broke trying to deal with overload.

One thing you can do to help with overload is to use a 'soft splitter'. A decision tree set to any/any works on some levels. It's faster than a sift and so can be used to shunt the output into two sifts. A high error genetic solver or perceptron can be used in levels without decision trees to split things into 3 (actually it'll typically split things into two because the 'right' answer is rarely chosen and by the time you are here you've already sorted for colour). It's important that the soft splitter be faster than what you are splitting for otherwise it doesn't really help anything..