Maintaining and running servers is a steady financial drain,. no matter what. Do you think LE would have fared better if it was designed as an offline-only experience?
Not compared to the amount of staff they have, no. Payroll is almost always the biggest single item on a company’s P&L.
I’m not sure. Certain things would certainly been a lot easier (no lag, for example) but other things wouldn’t have been possible (MG/trade). It would have been different, certainly, but better? Not necessarily.
It would’ve worked better as a SP game with MP functionality. Yes.
Especially since EHG was so reluctant to create a proper market environment (which is half-assed at best and the consumable aspect is entirely missing… which would’ve been a important part for the longevity of a market).
And you think nobody needs to sustain those servers? No moderators which are paid? No tech-staff? No monthly contract which drains funds?
Tech staff is almost certainly handled by the company renting the servers.
There don’t seem to be that many mods in the in-game chat either.
There is a montly contract (or even yearly), but the amount is almost certainly negligible compared to the total amount of payroll, which was Llama’s point. Probably even less than 1% of their expenses.
Gaming servers can be quite expensive.
EHG stated that they have an on-demand solution (reasonable, since games like this always have high peaks and then significant drops). Costs scale with active players online players.
This gives a rough idea how much stuff could cost: Dedicated Game Server Hosting - Amazon GameLift Pricing - Amazon Web Services
Even if you pick the most expensive option in their examples (100 player battle royale), it would still be 250k per month.
With over 100 people on their team, that is still less than half their wages cost. And it’s also not likely that they’d need that much. The monthly cost is likely much lower than that, considering that there are never more than 4 players per instance, at most.
Well, the directly displayed prices are only the on-demand costs showcased. There’s a price-checker with a rough estimation based on the respective services run on em as well. Bandwith is the primary factor.
If we take 5k players median (which is more then we currently have but is supposed to make up for the massively higher numbers during the first weeks) and imagine solely the usage of the Bazzar and the stash we can expect a 1 MB packet per usage roughly and with 10 uses per user per hour as example.
Then we got the API service (which is based on packet bandwith cost) to see the monthly price estimation. In that case we would be at 4,6k USD per month.
That’s not the real-time aspect either which is massively higher as well.
I estimate the total server-cost per month at around 150k+ USD for LE.
Which is actually higher in terms of users per server then a ARPG.
While a FPS needs precision for positioning of the players and uses a large amount of immediate bandwith for that a ARPG does a shit-ton of math in the background which needs substantially higher amounts of server capacity then a FPS does. The majority of server capacity for any FPS is synchronization, which is not as severe in ARPGs but still very high on the list in total.
Hence we can expect a use-case of not 30 per but 20 per… which means a 50% price increase.
Hence 375k USD per month.
And if we count 100 people on their team and take the median income (not everyone there is a coder with 5k/month) we can say the median income will be around 3,5k/month.
Which would make around… 350k in wage cost.
So even if we take everyone at a 5k/month (unrealistic) it would still come out to 500k/month.
Hence we can expect the server costs to actually make up over 1/3rd of the total cost of the company.
That is not small, that’s beyond the total expected profit margin for a reasonable product in any sector.
I… uh… I… what?
Maybe you should pick one?
Running the calculator for several different configurations, the value usually comes up a bit below 100k. Even though there are many different factors that can change that (it can even go up to over 1M/m), I think it’s likely that the value is around that.
Not to mention that those costs are always comprised of 2 aspects:
-The monthly cost for whichever configuration you chose (mostly around 25k).
-The monthly cost of on-demand instances, meaning when you go over the expected player count, which is mostly peaks and first month of season (usually around 75k).
So the server costs will actually be a lot lower during 3/4 of the season, only ramping up during launch and the first couple weeks. Since on-demand is on an hourly cost, even that won’t be a simple monthly cost, since it will decrease as players start leaving.
For the last couple of months, I doubt EHG has spent even a total of 100k on that. Especially because steam count also includes true offline players.
The second is based on your example.
I don’t think they’re using the highest amount and will have special conditions since they’re a huge customers and the prices there are set related to the ‘common pleb’ 
And even if we got ‘a bit below 100k’ that’s massive.
And I took the extra costs out related to the on-demand and simply went with a median amount of 5k players overall without changes, for simplicity’s sake (makes up as mentioned for the high peak in the early weeks and the low count afterwards which stays relatively permanent).
I’m not sure what service you were using for the calculations, then. Because I was using 500k peak in the calculator and it was still under 100k total with the on-demand being 75k.
And 100k over 2 months is 50k, or the cost of 10 devs (many devs will actually be paid more, but that will be dependent on which country they’re from, and seniors will certainly get paid a lot more).
The server costs for running 5k players isn’t too high. Certainly less than 100k per month, otherwise Chris’ statement that you can have a healthy game with a playerbase of 10k doesn’t hold up.
When player numbers are down, you’re not likely to be paying more than 25k a month, at most.
GGG formerly used only a predicitive network mode when this comment was made, with the lockstep being introduced freshly then, which rose costs a lot but made the game experience highly more enjoyable, especially since you could choose network mode (predicitve costs a ton less resources but is better for some players actually).
Also GGG’s system for stash and market access is vastly better designed. You can see that with the severe lag happening while interacting with the stash in LE. In PoE the stash is actually saved in local memory and only causes bandwith when switching tabs (for the server to know where the client actually is at the moment) and changes to the items inside.
In LE this sadly is not the case, every change of every query in the Bazaar for example re-queries the respective aspect from the server anew. Same is the case with the stash. This raises database queries massively and leads to a overall higher bandwith usage. One of the major issues they have.
I personally don’t think their code is otherwise up to par either.
Yes. But it indicates that servers are more expensive than 1% suggests.
1 Like