Wednesday, February 20, 2008

The PvP Equations

All the buzz about WAR has inevitably led to discussions about it’s PvP and RvR components. The one game mechanic that has not been mentioned in all the PvP talk is instanced PvP. The success of games like Counter Strike: Source, Doom, and even WoW’s battleground PvP seem to indicate that no matter how much sandbox PvP, or RvR perpetual PvP zones, are desired, some players still want to be able to log in and frag a few opponents without considering the ramifications of where, when, why and with whom.

How important is the multi-player instanced PvP option today? Most single player games are incorporating multi-player PvP options (see Halo), and some are even designed as multi-player PvP games with the solo play thrown in at the N’th hour (arguably, also Halo). I think the demand for instanced PvP is evident; the problem lays in the game mechanics that shape the end product.

One of the biggest obstacles is how to reward, or rank, the combatants. Segregation by experience levels, power scores based on rank, scores created by win/loss records, etc. have all been tried; yet all seem to be lacking in some way or another.

When considering how to create a ranking system (independent of gear, experience, or talent rewards) what should matter? What should separate a top ranked player from the lower masses? Do you weight the formula on dedication and time spent playing or purely on wins and losses? What about kills vs. deaths, damage dealt or Killing Blows?

From the protected secrecy of Credit Score Formulas, to the arcane (and some would say arbitrary) BCS rankings in college football, we are surrounded by complex formulas that matter in our real lives. In a world where complex esoteric formulas are fundamentally important, should not games begin attempt to also create elegant solutions, even if complex, rather than just hitting the ‘easy button’ by reducing complexity and thus representation?


will said...

I think that as a ranking system becomes more complex, it naturally becomes closer to a true ranking. Just win:loss isn't enough, so we should toss in ahhh, how about teams played? You can't have 2 10:0 teams, one playing 10 tough / close games against equally skilled opponents and the other having just asswhipped 10 teams of a lesser skill. Right?
Well I believe in all fairness you might as well take that idea and run with it. Lets add team roster strength, player abilities, home/away games, weather?, latency, stats, gear, level...

How many different variables describe a team relative to another in terms of "who's better?" I think that holes as deep as you want to dig it.

One obvious problem I'm seeing with computer games is data. How much data is a system able to record and parse? We are getting crazy with solid state hard drives, quadcore server's(... eights on the way), and all this lovely technology, but how much information is enough?

Can we make a ranking System too large to be parsed up?

How huge can a system's ranking algorithm be before we start leaving variables out?

If some system could keep track of all the possible data, and have a beautifully articulate and unbiased algorithm.... would there be a FirstPlace?

Todd Bursztyn said...

Algorithms...PvP...Unbiased ranking systems...cannot compute. This post went way over my head, although I couldn't resist a "hell yeah!" when Halo popped up.

brandon j said...

Despite technical advances in hardware and bandwidth which allows for more complex algorithms using large variables, the actual weight each variable is given has to be done by Humans. So if we ignore hardware/bandwidth limitations, where does that leave us? What variables, and at which weight, should be used in calculating PvP rankings?

In order to clarify some of these ambiguities, let us examine a hypothetical multiplayer Capture-The-Flag (CTF) scenario. Two teams of five players each, no time limit, simple respawn points and a scenario victory condition of three captures.

1a) Scenario Victory – This variable is needed, otherwise we could scrap the CTF scenario, and just place the two teams in a sandbox death-match.

1b) Flag Captures/Returns – Should specific captures/returns be weighted, or is overall scenario victory enough? If we do reward individual Captures/Returns, is that reward given to the specific player who accomplished the action, or the team as a whole?

2) Opponents Rank – There is a need to weight team vs. team competition, and comparing existing ranks is an old standby. But how should this be used in CTF? Average out the entire team and compare? Use player specific rank in determining kill/deaths? How important in a CTF example is the actual rank of a player, when trying to create a balanced algorithm for rank? This is needed, but reminds me of using a word in its own definition…

3a) Kills/Deaths – The most common variable used in PvP rankings, but is it acceptable just because it is common? In a CTF example, how much importance should be placed on winning the scenario compared to gaining a favorable Kill-to-Death ratio?

3b) Damage/Healing Given/Taken – These four variables seem to me a better measurement in weighting combat than Kills/Deaths. The arbitrary nature of scoring Killing Blows leads me to dislike it as an important variable. In the CTF example, a dedicated flag-runner should not be penalized for being the main target of the opponent, while delivering relatively few Killing Blows. The problem is in how to weight the four components. Or should Kills/Deaths and Damage/Healing be ignored completely in a CTF scenario?

4) Time to Completion – Should fast victories be rewarded? This variable could limit honor farming (or Kill farming), but it could also encourage pre-made groups scouting for (and only fighting) easily beatable PUG’s. I like the concept of rewarding fast victories in a CTF scenario, but how important is it?

5) Gear – How should gear affect the algorithm? Should clear gear imbalances be acknowledged between teams, what about between individual players? Basically, should gear imbalances be prevented in a CTF scenario, or weighted as a variable and allowed to exist?

6) Team Class Make-up – If different classes are offered, then certain five-man combinations will be ‘optimum’. Unless the team combinations are locked, or the classes offered were balanced with CTF in mind, there will be an imbalance of class combinations with everything else being equal. Should unusual class combinations be rewarded, or is the min/max use of optimum class combinations a goal?

What I am after here is not a working mathematical algorithm. Rather I am interested what factors we believe are important to PvP rankings, and why. From the above examples, I would argue that 1b, 2, 3b, 4, 5 and 6 would be my choices (in that basic order of importance).

What is missing from that list, and what is there that should not be?

R.C. Price said...

I think I like having the option of blowing everything up in sight ro playing well with others. Halo is one of the few video games I play. Granted, I accidentally blow myself up than any opponent, but there is always hope of improvement.