DUST Search - Matchmaking (Analytical and programming perspective, all nerds report in!)

Zeylon Rho wrote:

Fitting could be an self-selected means of match-making and be fairly functional. MLT/STD, ADV, and PRO/no-restriction matches. They could also gate the matches by Meta-level.

The newbies could opt for MLT/STD matches, and while vets could do that too - they wouldn't be able to field fits with any gear better than STD on it.

Vets would still have a passive advantage if they played lower-gear tiers, but newbies wouldn't have to eat the combo of passives + PRO gear.

You could have matchmaking within a gear-gated system, and make an internal player rating based on K/D, WP, SP, and W/L. I'd give further weight to pre-existing squads on a team (current matchmaking ignores whether or not players or squaded I've noticed), and you could probably make some sort of ELO score as well. Like if the OP and I had similar base scores, but in 9/10 matches against him I lost, then my rating drops.

Gear-based provides a simple solution that will work 100% of the time for gear-issues. You can't field better than STD/MLT? Done. It doesn't fix other issues, but unlike most solutions it always fixes the gear disparity.


Author	Thread Statistics \| Show CCP posts - 0 post(s)
lrian Locust BLACK OP SNIPERS General Tso's Alliance 153	Posted - 2013.10.08 01:08:00 - [1] - Quote Onesimus Tarsus wrote: KDR. Then you'll get snipers fighting snipers, and logi's fighting logi's.
lrian Locust BLACK OP SNIPERS General Tso's Alliance 154	Posted - 2013.10.08 01:10:00 - [2] - Quote Zeylon Rho wrote: Fitting could be an self-selected means of match-making and be fairly functional. MLT/STD, ADV, and PRO/no-restriction matches. They could also gate the matches by Meta-level. The newbies could opt for MLT/STD matches, and while vets could do that too - they wouldn't be able to field fits with any gear better than STD on it. Vets would still have a passive advantage if they played lower-gear tiers, but newbies wouldn't have to eat the combo of passives + PRO gear. You could have matchmaking within a gear-gated system, and make an internal player rating based on K/D, WP, SP, and W/L. I'd give further weight to pre-existing squads on a team (current matchmaking ignores whether or not players or squaded I've noticed), and you could probably make some sort of ELO score as well. Like if the OP and I had similar base scores, but in 9/10 matches against him I lost, then my rating drops. Gear-based provides a simple solution that will work 100% of the time for gear-issues. You can't field better than STD/MLT? Done. It doesn't fix other issues, but unlike most solutions it always fixes the gear disparity. Then you'll get routined players skilling in passive only, who still have a huge advantage over newberries.
lrian Locust BLACK OP SNIPERS General Tso's Alliance 155	Posted - 2013.10.08 02:49:00 - [3] - Quote Crash Monster wrote: Create a Bayesian analysis engine to see the impact that various players have on the outcome and mix and match players until the win/loss impact is reasonably balanced. To get more sophisticated ratchet the results based on the last 10 games to compensate for recent player style or equipment changes. Then how can you measure the impact that people have on a game? A pilot might have a huge impact, although you'll hardly see him in the hacking and killing stats. And how about that speed-tanked scout who sacrificed himself at the start of a Domination game, but gave his team a huge advantage by capturing the objective? I like your idea, but perhaps we're making it too difficult than it needs to be.
lrian Locust BLACK OP SNIPERS General Tso's Alliance 155	Posted - 2013.10.08 03:10:00 - [4] - Quote How about the statistic "average total meta level per spawn"? It works like this: 1) add the meta levels of all equipment, dropsuit and weapons per spawn, 2) add all the numbers of all the fittings together, 3) divide it by the amount of spawns. Here's your average. Calculate it over the last 30 days, and/or or 20 games, or whatever, and off Scotty goes! This has the following advantages: - higher level fittings will be matched with opponents with higher average meta levels - a pubstomping squad's team will be filled with newberries to level out the averages, levelling the playing field - Newberries will therefore still be able to earn ISK and SP - AFK'ers use low-quality suits and will count as less of a force in a game - increase and decrease in difficulty will be gradually - the playing field will be leveled when it comes to gear and experience, but skill can still make the difference - it works with any amount of players - it's easy to calculate - also works for new players entering during a game - a players value can be calculated prior to matchmaking, so the matchmaking process is fast To avoid a scenario where vets with maxed passive skills and militia gear will still outperform newberries with similar gear, PPC can choose to add additional meta points for completed passive skills (e.g: a level 3 electronics skill will add 3 meta level points). Appreciate any feedback!
lrian Locust BLACK OP SNIPERS General Tso's Alliance 155	Posted - 2013.10.08 03:17:00 - [5] - Quote Atiim wrote: Match people based on War Points, as I feel that this is the only true test of skill. That really depends on the role you're playing. A sniper will get less WP, although he might pin the enemy down and soften them up for the guys on the ground. On the other hand, logi's can rake in the WP and might have to fight mercs that are relatively much stronger than they are. They'll have to fight an uphill battle and lose on ISK due to the high cost of their fittings.
lrian Locust BLACK OP SNIPERS General Tso's Alliance 155	Posted - 2013.10.08 03:20:00 - [6] - Quote knight of 6 wrote: I'd say on a 5 match average or something. reason being if you are AV specced and nobody is vehicle specced than you don't work to your level. I agree, it has to be an average, or the skill levels will be all over the place. There need to be more games though, and in a limited period (say, 20 games in max. 30 days?).