Server side recording

[Candyx]

It's been said before but I'll say it again: Well done on the development, keep it up!

With that in mind, how much of the game play is recorded server side? Can you option to record more for a short space of time? If you can record enough then you could save alot of time on amun, and a lot of back and forth with the balance updates.

So can you option to write all received packets to a drive for a few hours? You've clearly got the decryption algorithm so you could then post process the data. Then you're just a hop-skip-and jump from having a reasonably accurate regnum model that you can run balance simulations on. This would also make testing proposed changes easier and allow you to cut the bad ideas before having to code!

[Robasiewicz]

while on small scales this could work, it'd be completely broken on larger scales.

additionally, analysing the same scenario over and over will make it impossible to find a bug that exists outside of it

[Candyx]

Saying it would be completely broken on a larger scale doesn't make a great deal of sense. If properly implemented I can't see why it would fail. It seems like you've assumed I'm talking about Monte-Carlo simulation sets; these types of simulations are specifically for the large scale, they're used for prediction of environment, space debris growth, and a whole host of other 'large scale' tasks.

The purpose of this wasn't to find all possible bugs; it was to reduce the time spent testing changes for balance.

[TurboLover]

Well if you're talking about something like a GoldSrc game recorder (something a little different - every client can record the gameplay) for the server side, I can't imagine how
people would analyze all that data. The best I can think of is replay like a movie or
detecting places where people get stuck in the terrain.

Also, I don't suppose any of the traffic is encrypted, if you were talking about network packets. What simulations are you talking about?

[Candyx]

What I'm really talking about is developing a system where unique entities are simulated during a battle. This could show that some proposed changes are going to have not enough effect/too much effect - a simple example is the energy barrier for mages. The increase from 1000 - 1400 could have been simulated to (potentially) show that it may make very little difference in large battles, but a greater difference in 1v1. This might then suggest that the energy barrier value at top level could be tied to the number of allies in a location. This is just a very simple example and there's a lot this type of simulation wouldn't be able to tell you; like the increase in number of people who use energy barrier - I don't on my conj but if it's increased more maybe I will do.

These types of simulations use some quite complicated topics, but there are hundreds of sources out on the internet that allow you to surpass a lot of the "how's this work" stuff and get on with the simulations and results. At the very least it would be a useful tool in the development process.

[Robasiewicz]

Originally Posted by Candyx

Saying it would be completely broken on a larger scale doesn't make a great deal of sense. If properly implemented I can't see why it would fail. It seems like you've assumed I'm talking about Monte-Carlo simulation sets; these types of simulations are specifically for the large scale, they're used for prediction of environment, space debris growth, and a whole host of other 'large scale' tasks.

The purpose of this wasn't to find all possible bugs; it was to reduce the time spent testing changes for balance.

lets say that we want to test (among others) power X, for this we record everything happening both in the WZ and inner realms (we can disregard noob areas)

the recorded data shows how people use power X with its original stats. lets say it is 100% in the most simple scenario it would be damage.

let us follow what did 3 players - 2 greens and 1 red.
in the original scenario 1 green dies and so does the red guy. 2nd green is left barely alive.

if power X were to be boosted up to 200% the green player would die even faster, thus not providing heals, buffs or whatever. this is the optimistic version if things went south he'd be running around with negative HP possibly breaking many mechanisms in the process.

if we lowered power X to 50% the green player would survive longer. no visit to altar, no going elsewhere. if he did something important at another battle site: though luck.

there are just too many variables to account for. if we wanted to test things based on recorded data, someone would need to browse thru all of it and probably adjust for each and every simulation, consuming way too much effort. this is why NGD asks for user feedback.

If they made a separate server and made bots to play the game, it would be possible to test a lot of things.
the problem is a lack of decent AI. Adrian said that dragons have a terrible AI and they worked quite hard to even make it work. imagine writing an AI capable of using every skill in the game. imagine how much would it cost. once again: this is why NGD asks for user feedback

[Candyx]

Originally Posted by Robasiewicz

lets say that we want to test (among others) power X, for this we record everything happening both in the WZ and inner realms (we can disregard noob areas)

the recorded data shows how people use power X with its original stats. lets say it is 100% in the most simple scenario it would be damage.

let us follow what did 3 players - 2 greens and 1 red.
in the original scenario 1 green dies and so does the red guy. 2nd green is left barely alive.

if power X were to be boosted up to 200% the green player would die even faster, thus not providing heals, buffs or whatever. this is the optimistic version if things went south he'd be running around with negative HP possibly breaking many mechanisms in the process.

if we lowered power X to 50% the green player would survive longer. no visit to altar, no going elsewhere. if he did something important at another battle site: though luck.

there are just too many variables to account for. if we wanted to test things based on recorded data, someone would need to browse thru all of it and probably adjust for each and every simulation, consuming way too much effort. this is why NGD asks for user feedback.

If they made a separate server and made bots to play the game, it would be possible to test a lot of things.
the problem is a lack of decent AI. Adrian said that dragons have a terrible AI and they worked quite hard to even make it work. imagine writing an AI capable of using every skill in the game. imagine how much would it cost. once again: this is why NGD asks for user feedback

Your example is somewhat limited by your assumption that battles will be 'replayed' using the different values. Making an AI that copies what people do is relatively easy compared to making one that does something new (dragons). If a player's health goes to below zero in a simulation then they would die, and go alter. If this were a Syrtis player trying to defend efe castle there would be a chance they had saved at efe save before the battle, and there would be a chance they forget to go to the wall (if they hadn't saved as efe). All these things are easy to find through data mining.

To run through a more accurate simulation: Lets say recharged arrow is boosted to 100% weapon damage bonus (clearly too overpowered). If people who current use it stack it with lethal strike then there will be marksmen hitting 2-3k damage with on hit, potentially killing players who aren't warmasters in a single hit. It would quickly become obvious from simulation that the makrsmen don't die as frequently, and the side with the most marksmen win the majority of battles. This would suggest that the spell is too strong and needs to be altered. Concisely one of the attributes of the simulation would be the inclusion of the game mechanics (hp less than 0 = death, mana too low to cast = can not cast, etc.). I had thought this was somewhat obvious from the context

[ieti]

As i see your idea - you propose a tool which to help NGD to balance the game easier. I think they can do this pretty easy without any recording. They just need to make some data mining on setups and spells which active players use.

Recording this amount of traffic on production system is a NO-NO. This is usually done only in scenarios when some show-stopper bug can not be simulated on development system. In this case you sacrifice the system performance and make your customers angry to track a problem. Doing this for long periods is unacceptable.

You can get what spells are most used and ask players why you use this and that and in what cases. You will get same information without all this resource consuming stuff. I think they do some recording on Amun and if we participate in the tests they can get the picture easier.