[QUOTE]they found out that players in parties of friends will play better than when they’re soloing[/QUOTE]
If I'm honest I do worse with friends in most games because we're all having a laugh and not really concentrating on the game.
[QUOTE=Yahnich;43099353]i hope this won't result in dota2 dickswinging 'i'm in platinum tier scrub'[/QUOTE]
it's separate to regular matchmaking, it won't.
[QUOTE=ThatSprite;43099764]it's separate to regular matchmaking, it won't.[/QUOTE]
so is ranked in league of legends and it's full of that shit
I wonder if this takes actual gameplay skill into account as opposed to just your win-loss ratio.
[QUOTE=ironman17;43100085]I wonder if this takes actual gameplay skill into account as opposed to just your win-loss ratio.[/QUOTE]
...how the hell would you measure that?
[editline]7th December 2013[/editline]
[QUOTE=Kondor;43100027]so is ranked in league of legends and it's full of that shit[/QUOTE]
I think I've yet to see a game where it doesn't happen once the system's in place.
[QUOTE=ironman17;43100085]I wonder if this takes actual gameplay skill into account as opposed to just your win-loss ratio.[/QUOTE]
Yes, if you read the article it says just that.
[QUOTE=Marik Bentusi;43100131]...how the hell would you measure that?[/QUOTE]
My guess is checking the heroes you play, checking their kill-death ration, actions per minute, gold earned, wards placed, missing calls made, things that indicate how well you helped the team, whether through carrying or supporting. To get a more refined measurement it'd need to evaluate what the others did, so that in the event of failure it can determine whether you lost because of your actions or the actions of your fellows, and when you win it would determine whether your team rocked or the enemy was merely sub-par in play.
It'd take a lot of data, but one could probably make a program that does the data processing on Valve servers to study the match and work out what went wrong/right.
[QUOTE=ironman17;43100301]My guess is checking the heroes you play, checking their kill-death ration, actions per minute, gold earned, wards placed, [B]missing calls made[/B][/QUOTE]
BRB, making a script to spam ss in chat.
[QUOTE=ironman17;43100301]My guess is checking the heroes you play, checking their kill-death ration, actions per minute, gold earned, wards placed, missing calls made, things that indicate how well you helped the team, whether through carrying or supporting. To get a more refined measurement it'd need to evaluate what the others did, so that in the event of failure it can determine whether you lost because of your actions or the actions of your fellows, and when you win it would determine whether your team rocked or the enemy was merely sub-par in play.
It'd take a lot of data, but one could probably make a program that does the data processing on Valve servers to study the match and work out what went wrong/right.[/QUOTE]
[QUOTE=ironman17;43100301]checking K/D/A[/quote]
Unreliable because one game someone might carry you and despite doing nothing you rack up kills and assists even as a support, then next game you need to give it your all to not lose horribly, sacrificng yourself for the carry whenever necessary.
Would also encourage fountain farming for as long as possible.
[QUOTE=ironman17;43100301]actions per minute[/quote]
Could be easily be farmed with move commands or ground attacks or buying/selling branches in fountain when the game's almost over.
[QUOTE=ironman17;43100301]calling ss[/quote]
Again something that could easily be farmed, and can't be counted if you do it via voicechat.
[QUOTE=ironman17;43100301]determine whether your team rocked or the enemy was merely sub-par in play.[/quote]
That's what the gold graph analysis is for. If your team snowballed, the graph will just be one big mountain. If the time between the last equilibrium and the end of the match was short however, chances are the game was either close throughout or someone was able to make a comeback at some point.
I honestly don't see what kind of magic analysis you could do instead of increasing and decreasing MMR depending on W/L. Sometimes someone will drag you down, sure, but everyone else will play consistently better while he will keep getting downgraded until he has something close to a 50% win rate.
[QUOTE=Marik Bentusi;43100131]...how the hell would you measure that?[/QUOTE]
K/D/A, XPM, GPM, last hits, denies, observer wards bought and used, invis detection bought and used, courier bought/upgraded, items shared, damage to enemy heroes, damage to enemy towers, healing done, level when the game ended, all globally compared to the average of everyone else who has played the hero (possibly in your rating bracket too). If you play a game and do better than this statistical average, it should help your rating either by giving you more "points" per win or, if you lose, less of a loss on your rating than if you did the average or less performance. If you do even or worse, it shouldn't negatively affect your rating, even if you lose, you're just under-preforming and are not going to a benefit just because you picked support. (ex. Bob wins a match as a carry but played below average. Gets +10 points on his MMR rating, a base amount for winning an evenly matched game in his bracket. Bob plays his next match as support and loses, but plays above average, only loses -8 points because of his contributions throughout the game. Bob plays a support in the next game, plays above average, wins, gets +12 points.)
Then again, I don't know the logistics behind setting up a system like this, but if you wanted a rating system that is more of an all encompassing rating you would have to include what I mentioned above and probably more into a players rating.
[QUOTE=Banned?;43101613]K/D/A, XPM, GPM, last hits, denies, observer wards bought and used, invis detection bought and used, courier bought/upgraded, items shared, damage to enemy heroes, damage to enemy towers, healing done, level when the game ended, all globally compared to the average of everyone else who has played the hero (possibly in your rating bracket too). If you play a game and do better than this statistical average, it should help your rating either by giving you more "points" per win or, if you lose, less of a loss on your rating than if you did the average or less performance. If you do even or worse, it shouldn't negatively affect your rating, even if you lose, you're just under-preforming and are not going to a benefit just because you picked support. (ex. Bob wins a match as a carry but played below average. Gets +10 points on his MMR rating, a base amount for winning an evenly matched game in his bracket. Bob plays his next match as support and loses, but plays above average, only loses -8 points because of his contributions throughout the game. Bob plays a support in the next game, plays above average, wins, gets +12 points.)
Then again, I don't know the logistics behind setting up a system like this, but if you wanted a rating system that is more of an all encompassing rating you would have to include what I mentioned above and probably more into a players rating.[/QUOTE]
I talked about K/D/A
XPM, GPM, last hits, denies, items, hero damage, healing can all be farmed (especially once the game's basically over and everyone's fountain diving). XPM/GPM isn't even a relevant stat for supports, just like buying support items isn't for carries. I imagine you'd also have difficulty balancing them since carries rely on their team to get room for farming while support items can be gotten pretty much regardless of how your team's affecting your game.
If you're relatively low on levels and haven't taken out all towers, it can also be an indication of a really good, snowballing game that you managed to win decisively, which could mean display of greater skill than someone who didn't play quite as well and thus took much longer to finish.
I don't really understand why you propose so much tracking player-individual stat, then decide MMR increase/decrease based on W/L anyway with merely the amount being affected by that complex analysis.
If that increase/decrease needs a variable amount at all, I'd tie it to the enemy team's average MMR. Winning against people with on average higher MMR is (if the system works) more difficult than playing against people with on average lower MMR, while losing isn't as unexpected.
[QUOTE=Marik Bentusi;43110163]
XPM, GPM, last hits, denies, items, hero damage, healing can all be farmed (especially once the game's basically over and everyone's fountain diving).
XPM/GPM isn't even a relevant stat for supports, just like buying support items isn't for carries.
I don't really understand why you propose so much tracking player-individual stat, then decide MMR increase/decrease based on W/L anyway with merely the amount being affected by that complex analysis.
If that increase/decrease needs a variable amount at all, I'd tie it to the enemy team's average MMR. Winning against people with on average higher MMR is (if the system works) more difficult than playing against people with on average lower MMR, while losing isn't as unexpected.[/QUOTE]
You say that like it doesn't happen in normal games already, thus affecting the would-be average stats in the first place.
GPM, probably. I'm not talking strictly about supports though. XPM, definitely though because a good support (character depending) will either be ganking and getting XP from heroes kills or stacking/pulling for additional XP outside of the lane.
MMR should always go up and down based on a win/loss. That's part of the point of it. I just gave a small scale, and kinda terrible (not much thought went into it), example. That isn't some kind of end-all-be-all solution.
That should also affect an MMR gain/loss. I know it does in HoN.
Sorry, you need to Log In to post a reply to this thread.
