General DiscussionA Supplementary Experiment to Swiftending's

Yes. I agree.

I fuck bad bronies thats my fucking problem.

The basic concept here is correct. While you are not forced to win 50%, there are plenty of games that extremely good or extremely bad performance on your part would be required to change the outcome.

A game like this http://dotabuff.com/matches/521046528 is impossible to win. Even though my bloodseeker is really fairly good (60% win and over 3 KDA for 22 games) there was no way to overcome the 0-22-0 warlock who also bought as many couriers as possible and fed those too.

To examine the other side of the question I played a series of 15 solo ranked drow games. I am very bad at drow ranger and I think I played quite badly in all but 2 of the games in which I was ok instead of aweful. Despite failing my lasthitting, failing my orbwalking, missing tons of gusts and not really knowing how to adjust my items to the game situation... despite an average KDA of 1.65 for the 15 games I still won 7 because about half the games were forced wins. So as bad as I am at Drow Ranger... merely failing, but still trying to win is not enough to cause an overall loss of MMR. You have to be EXTREMELY bad at a hero or throw games to consistently lose MMR.

Contra-positively you have to be dramatically superior at a hero to consistently win and gain MMR. This is because you have to turn games your team was going to lose. Doing well only when things were going well is not enough. To gain MMR you have to win the games your team started out feeding and does not know their heroes.

In my opinion depends allot on which heroes are picked.

By example I'm currently ~3500 MMR. If I play only Jakiro / Drow probably I'll get to 3800+. If I start playing only BountyHunter / Dazzle / AM, probably I'll get to 3000 MMR or bellow, and then start winning ~50% of my games with these heroes.

A 3500 player can be a 3.8k+ with some heroes, and 3k- with others. Matchmaking has no idea which hero I'm going to play next.

Sigh...I gotta retype everything huh.

I want to know how many games are played "in the margins." How many games in the sample of 100 are "even" vs how many are lopsided?

Rating system:

1 - complete blowout in favor of other team (a team score of something like 5-30 would be a stomp)
2 - slanted toward other team; not unwinnable, but unlikely for my team to win for whatever reason (bad picks at start, bad supports, etc.)
3 - balanced; both teams have good shot at winning
4 - slanted in favor of my team
5 - complete blowout in favor of my team (basically i could afk and my team would most likely win)

Here's where I need some 5k+ players. Basically I'll post a subset of games (probably 10 games per round, for 10 rounds) and add brief 2-3 sentence descriptions. I will judge them according to my best abilities, but I cannot deny the existence of bias, since I am an active participant. The 5k+ players can review the games (they don't need to watch the replays) and agree/disagree with the rating. I will post their feedback (anonymously if desired) so other people can observe.

If what I postulated earlier were correct, about 40 out of 100 games would be 1's and 2's, about 20 would be 3's, and about 40 would be 4's and 5's. These are numbers that currently have no backing, but I doubt the actual value would deviate significantly.

Basically here's the process:

1) Play 1 game a day. I won't play every day, so basically this will take months.

Reason: Going on a loss streak can negatively affect performance. I want to minimize these negative factors (since I'm not some 5.5k player "stuck" in a lower bracket, there's no reason to suspect I could raise my MMR quickly) and anyone who wants to improve should also adopt a good attitude. Playing 1 game a day means a loss won't affect my performance the next time I play.

2) Hero picks. I will pick to the best of my ability - unlike Swiftending, I won't pick role 1/2 heroes all the time. I'll pick what the team needs, but if I can pick up a hard carry and not incur any conflicts with the team, why not? (I still cringe whenever the orange/brown player goes mid though)

3) I won't flame, and if anyone tries to flame for whatever reason I'll most likely just mute them.

Current MMR: ~4600

Hopefully I can get a few volunteers (5k+)...if not I'll just ask friends.

you're doing it in us west and us east, pick a server with better players (EU, china, SEA)

wave knows it!

i should play on SEA - for some reason i actually win more there (not even joking)

every time i play on eu my team picks like ass, though i guess the players are tolerable

didn't murs say 5.2k on eu is low or something? doesn't this mean games are actually harder on us east?

we'll see, i'll most likely play on all 4 servers, not that it makes a huge difference to me

It should be 11%, not 20% (1/9th) where you are the game decider.
That means in best case you can get ~60% winrate.

You mean 1/10? I thought about that as well, but the game is decided by which team performs better. You're part of 1 team, not both teams, so I "guessed" 1/5.

We'll see what the numbers actually are.

EU West is a lot more competitive than US East. It's not surprising since there are more DotA players in Europe than in the US.

Alright so I'll just use this template and post the games onto some document uploader in the future. This is just an example:

Game 1: http://dotabuff.com/matches/530008552
W/L: W
Score: 3
Summary: My team had a bad start, but at no point did I feel either team had any significant advantage/disadvantage until the other guys kept losing team fights (bad positioning, not casting spells properly, etc.) like 3x in a row. By the way, this was one of those games where we didn't have courier at start; usually I would moan and QQ but I decided against that, as stated in one of the conditions for performing the trials.
I don't even play Mirana at all (4 games on this account, maybe 10 on my main), so I think if I played a hero I usually do that could solo against WR (considering how atrocious my last hitting was), I might have tilted the game to a 4, but that's all right.

Notes: (Here's where the other feedback by 5k+ players would go, once I ask them to assess the score rating)

99 games left to go, let's see if I can hit 5k.

can you do a tl;dr please? :D

tl;dr: he wants to know how many of the games are "forced" and closed, and if the forced loses and wins are the same %.
he believes 40% really hard to lose, 20% close and 40% really hard to win games, wants to test it with his own games, needs players to judge.

yea pretty much

one way to think about it analogously is flipping a coin. if you flip a coin 100 times, the expected value is 50. except the coin is weird. sometimes the chance of getting heads is 60%, 90%, 15%, etc. The expected value will still be 50 over a large number of games, unless your skill level goes up/down (in which case you might get 52 instead of 50).

is it possible to still get 50/100 even if ur actual skill goes up? yes, but the more you improve, the less likely that will be.

it's not particularly complicated, but dota has a pretty steep learning curve the higher you climb, so i can understand why people might think MM is "forcing losses" onto them at some point with terrible allies. newer players with 2000 mmr (and i guess those who just stay terrible after several years) will also feel that way. you just have to keep playing consistently at a higher level.

all of my friends with > 5.5k have played competitively at some point or another, even if some of them aren't well known. playing inhouses and scrims with better players will provide lots of insight; things you normally miss will surface. i don't think spamming pub games is particularly helpful unless you're practicing mechanics, which is another reason the 1 game/day is in effect

Well, if this is relevant, I have a sample size of 40 games with Terrorblade played nearly back to back. I'm currently 25-15 on him (one game my CPU burned out while raxxing and I got an abandon), with the highest GPM on my team in nearly every game. Just sitting slightly above 60% win rate.

According to your theory it would mean that I'm winning all the games I'm supposed to win and all the games that are slanted towards the other team. However, evaluating my losses, I don't find this to be the case.

By order of how recent the games are:

http://dotabuff.com/matches/529236166

I single handedly throw the game through a couple of bad decisions late game after getting two lanes of rax and being too complacent. My teammates weren't really useful but it was a relatively easy win which I threw away.

http://dotabuff.com/matches/525975955

The enemies made the good choice of putting an extremely strong offensive trilane (Bane, Potm, Lion) and even with trying to use illusions to cancel nightmare on me I could practically get no farm out of the lane. Naturally with some levels catchup farm isn't too hard, but they pushed early and finished us. It didn't feel like the skill level favoured them particularly, just that they had better laning decisions.

http://dotabuff.com/matches/524805549

Mostly out-matched, with Storm dying four times to rotations early and underestimating the mana-draining power of a QW Invoker. This let Invoker snowball to unimaginable proportions raping us early.

http://dotabuff.com/matches/519062652

Partly a throw and partly an outplay; the Invoker got mad when I refused to get a BKB because I was only having trouble with Double RP+March, and deliberately threw the game. Otherwise it would have been an easy win through slow siege.

http://dotabuff.com/matches/519004975

Got shut down by a Wisp-Troll-Skywrath tri-lane early. Nothing exceptional about the enemies, just had the stronger laning stage. My last pub picking a jungler resulted in us having nothing to contest the tri-lane with.

http://dotabuff.com/matches/517617387

Can't really remember much, sniper basically snowballed alot.

http://dotabuff.com/matches/514047499

Dc-ed while raxxing and got an abandon.

http://dotabuff.com/matches/512785857

Earth Spirit picked ES without knowing about the patch's changes, ended up feeding alot. Also partially a throw on my part through bad decisions.

http://dotabuff.com/matches/512730398

Can't remember, got outplayed mostly.

http://dotabuff.com/matches/511984517

Lost to a Weaver+Wisp offlane, Alch has zero lane presence and fed first blood to them, resulting in a level advantage that allowed them to tower dive by 5 minutes. Probably the game I was the most frustrated at my teammates.

http://dotabuff.com/matches/510605736

Can't remember much, got outplayed.

http://dotabuff.com/matches/509402907

Can't remember much, got outplayed.

http://dotabuff.com/matches/508376916

Bad decision making late game again.

The rest I just started playing TB and was experimenting around with some (relatively) bad builds.

TL;DR

Most of my losses are not "forced" in the sense that my team was noticably worse than the opponent's. I threw a handful, lost by greedy jungling picks against strong offensive tri/duos in a few, and got simply outplayed in around 3-4 matches out of 40.

So yeah, the closest to "forced losses" I've come across is about 10% of my total matches or less, and there could be alot more reasons for that loss which I simply wasn't in a position to observe.

" there was no way to overcome the 0-22-0 warlock who also bought as many couriers as possible and fed those too."

This is what i mean, HOW THE FUCK swiftendings never gets those guys in his team, i mean the ones who buy courrier and feed and suicide every time they are alive, because there is a good amount of those people on the 3k mmr.

It makes no sense at all.

Well, you can avoid some of them by dominating the game very quickly and decisively. This warlock started boots only, came to my lane and autoattacked up the lane until he could feed FB at the enemy tower. Then he immediately started to try to provoke me and said that he was going to feed and throw the game because I did not deserve to win games... etc. until it became clear I was not paying attention to him. Then he kept feeding while trying to get into a fight with the razor who was taking the bait... thus making the warlock feel important. And so Warlock spent the rest of the game saying he was feeding because the Razor did not deserve to win. Everyone knows he just planned to do this before the game started.

However, many throwers decide to throw only after something goes wrong. If you can get FB and get something good going quickly before the hidden thrower decides to lose the game then you succeed in dodging the thrower. Swiftending is so vastly superior to his lane opponents its very likely he will outplay them so much that they die at early levels. So he will avoid all but the most determined throwers. This warlock... no one could have prevented that, but a thrower like that is not really so common.

I do think that maybe 2 games out of 5 are determined by the "game is lost so I give up" sort of thrower. 1 game will be a free win, 1 game will be a free loss. But you have some chance to make it a win by making the quitter on the other team quit first. Demoralizing the other team is important. This is also a big factor in why 5 stacks win more... because they are far less likely to quit when they start badly. Lots of dota games that start badly are winnable. But because 1 guy in 5 is a quitter, once it goes badly they start throwing... by afk farming, by avoiding deaths, by just mentally checking out, by raging, etc.

[QUOTE] "Well, you can avoid some of them by dominating the game very quickly and decisively. This warlock started boots only, came to my lane and autoattacked up the lane until he could feed FB at the enemy tower. Then he immediately started to try to provoke me and said that he was going to feed and throw the game because I did not deserve to win games... etc. until it became clear I was not paying attention to him. Then he kept feeding while trying to get into a fight with the razor who was taking the bait... thus making the warlock feel important. And so Warlock spent the rest of the game saying he was feeding because the Razor did not deserve to win. Everyone knows he just planned to do this before the game started." [/QUOTE]

Apart the fact that playing the hero you picked better than the average as you suggested above might increase your MMR, there is still just so much you can do with certain heroes.

While playing a support, you can do everything, you can ward, deward, buy dust, be jesus christ on earth.
Yet still there are games where your carries simply cannot win the game.

to be honest, to me (and this is my personal opinion) the system valve is using is:

1. non productive
2. not stimulating
3. doesn't promote a community
4. lacks of a proper ranking (especially in ranked games, where all you've got is a number, but there's no way to know with precision where you are compared to the rest of the community, maybe even on the specific server you are playing in).

[QUOTE] "However, many throwers decide to throw only after something goes wrong. If you can get FB and get something good going quickly before the hidden thrower decides to lose the game then you succeed in dodging the thrower" [/QUOTE]

The problem here is we cannot predict human beings, and that's also why this system Valve is using fails.
I've had plenty of games with mistakes at the beginning, FB's and other stuff, yet still managing to win because the team is "friendly" not to mention "smart" and understands mistakes can happen.

All this to say that in my opinion, people playing as support or heroes that aren't capable of turning the tides alone late game especially, are more influenced by this than other heroes.

Then again of course, there are just games you play bad, nobody is perfect, I've had horrible streaks of questioning myself why am i even playing at this moment? should take a break and come back more focused.

Honestly all i wish is a more balanced system with competitive games, i don't even like to get easy wins, makes no sense since its a competitive game

speak of the devil, and he shall appear

http://dotabuff.com/matches/531699028

hehe ^

@ zenoth

some of those losses are indeed based on games heavily slanted against you (i.e. ur basically not playing with a full deck). in others, you might have made bad decisions. that being said, there are 2 "laboratory conditions" i'm imposing for my experiment

1) i'm not exclusively picking a carry every game. since i'm not > 5k elo, i'm not going to be the highest elo player so it's not necessary to try and "carry all the bads," so to speak. there are 2 other reasons:

a) a lot of times, by picking a hard carry, ur early game will suffer. as you are well aware of, games spiral out of control early esp. in the current dota meta. if you pick something like mirana/clinkz instead, the game might be a 3 instead of a 2. terrorblade is strong early but has no escape/needs to use the unreliable sunder to survive ganks

b) your teammates who want to play carry (or only know how to play carry well) might get upset, and not support properly. as a "jack of all trades player," i am fine playing support as long as my carry doesn't miss excessive cs in free farm or does something stupid like dive t2 towers with no vision and nobody to back him up.

the downside is that i don't know which players are good/bad at carrying, but that's essentially why i am performing the experiment. if i get a player who instalocks carry and fails miserably, then the game is going to be a 1 or 2. this pattern should be systematic and not random, if my hypothesis is correct.

2) i'm not playing whenever i'm upset, tired, angry, etc. so basically it's almost as if i took a vacation to maui, and decided to play dota sparingly. i'm essentially in a low/non-stress environment whenever i play.

i don't believe most players are using those conditions, but if they want to improve their MMR consistently, in addition to improving their overall skill levels by critically thinking about every play/mistake, they should do something along those lines

to respond to your initial complaint, i do believe it is possible to lose games that are 4's and 5's. if u have a major advantage and throw it away, then of course you should blame yourself, but i would still rate some of those games 4's.

by a cursory glance of your first game, your team was dying heavily, but you had quite a bit of kills and farm, so that would be...a 3? it's not entirely your fault, seeing as your team fed double digit kills across the board, and basically the burden of carrying fell entirely upon you

----------------------------

tl;dr what you said is completely reasonable. about 10% of games are 1's. that fits perfectly with what i "guessed" earlier (that means ~30% are 2's, and those are difficult but not impossible to win)

the only roadblock i've come across is how to rate games properly. if i can't do this, then i have to rely on intuition, which means the experiment isn't based on logical reasoning. i'll try and solve this dilemma ASAP and not wait till my 50th game.

relentless is there some way to come up with a formula to ascertain whether games are 3's instead of 4's, etc.?

if u look at swiftending's example that's clearly a 1, with the feeding alchemist. there are cases that might be 3's, but i might rate them 4's, and i want to minimize the margin of error and reduce uncertainty

ultimately if i rate a game "1" when it should be a "2," it doesn't matter cuz they're categorized in the same group. 1 + 2 according to my hypothesis should compose of about 40% of your games. if i rate something a "4" when it should be a "3," then there's a problem.

i'm basing the score off game results, not how the participant played.

http://dotabuff.com/matches/531754588 How would you judge this game I just played?

This game was impossible for me to win and probably impossible for anyone to win. The TA on my team was throwing. Nothing negative was said by me or any other player to the TA. And TA said nothing the entire game. He simply fed throughout the early and mid game and then when it became clear we still had some chance to win anyway he DC and immediately abandoned ensuring that we lose.

Besides this Rubic and bounty hunter did very badly often walking into the enemy team who was visible on the minimap. BH attacked groups of enemies solo without checking for true sight. They knew this was a mistake, they were even using mics and would express their frustration with their failure. So I guess they could not have played better.

I get the highest hero dmg, highest tower dmg, highest KDA, highest xpm, highest gpm... all on a Shadow Shaman. But still I lose.

I was as positive and encouraging to the team as I could be, but when TA destroyed his items and quit they also lost the will to fight.

if u have a thrower, there is nothing you can do about it. based on game results it would likely be a 1. small chance of being a 2.

their slark is pretty bad, but that doesn't offset the ta on your team.

edit: remember that i'm ultimately interested in game results, not how it "might have played out." it is indeed probable to win games that are 2's (i.e. ur team feeds heavily off the bat and they take a bunch of towers, but you make a comeback later) and lose games that are 4's (your team does very well for the first 25 min, and then they start throwing/feeding kills to the other team's carry)

Tempest I think a lot of games have a thrower that determines the loss. Some of them are obvious from the start, like the first game I posted. This last game was not so obvious, the TA just quietly played extremely badly. If he had not destroyed items at the end you might not even be able to tell from stats that he was the cause of the loss.

It might be required for you to watch the replays to decided if you had a thrower or if the other team did. It's not always obvious during the game. I watch a lot of replays of pub games and there are a lot of people throwing games intentionally... but not in the obvious "mass couriers and feed them" fashion. I would estimate 1 game in 5 has an intentional loser. That's basically saying that 2.5% of the dota population are throwing. 1 Player in 50. If you think about your experience in school its easy to see how likely that is. I'm sure you had 1 person in your class who would vandalize things or cause people hurt if they did not get caught... and sometimes even if they did.

"a lot" seems to be an overstatement. i do get throwers maybe 1/10 to 1/15 games (6 - 10%), but that might be because of different mmr brackets.

throwing can be subtle or blatant. someone who feeds courier right off the bat is blatantly throwing.

http://dotabuff.com/matches/531731793

in my most recent game, i drafted a lineup that is far superior to theirs. pugna counters a lot of their heroes (tinker, lion, even gyro late game once bkb drops to 4 seconds and i can perma-decrepify + drain him, etc.) and ck/sf can easily carry late game. however, the rubick adamantly wanted to play sf and he went mid for the first 5 or so minutes because he was upset he didn't get to play mid. i didn't flame or say anything.

was he throwing? i would say that fits the definition of throwing since he was not playing his hero role properly (note how much gold he has from farming most of the game as a support hero, which turned out not to be a bad thing hilariously enough) but he was still throwing during the lane phase.

that game is a 4, due to better hero picks + plays.

throwing doesn't automatically mean a "1" game. i'm sure at some point or another, almost everyone here played in a "throwing" fashion due to shit teammates/being in a bad mood

Well what I am saying is even 1 player in 50 throwing causes 20% of games to have a thrower. You have to know about that to judge the games.

You get throwers a lot more than you know. If you watch the replay and watch all the players you would see them. Part of Valve's "forced 50% win" reputation is that the community really does have plenty of jerks who cause the 50% winrate themselves independent of valve.

So in the case you present the Rubic tries to throw the game early, but then decides later that he wants to win. He doesn't follow through on the throw.

correct, but the chances of that player being in your game is quite small. furthermore the thrower might be on the other team.

(or if you want to be more technical about it: the player becoming a thrower during the game is quite small)

if u want, i can note throwers in my games, but i doubt it's going to happen in "a lot" of games

The odds really are not that small. As I said 1 thrower in 50 players makes 20% of games decided by a thrower.

Was there another thrower on the opposing team who threw harder? Look at that Anonymous Tinker.

no they didn't have any throwers - they just had worse players (or maybe they played worse due to the draft being favorable to my team)

as i mentioned earlier, the thrower was on my team, yet we won and i'm pretty sure the game should be rated a "4" due to me not having to do any heavy lifting.

Well I'll watch the game, but it looks very much like the anonymous player decided "fuck this game I want to try out tinker with 4 rockets".

the last game where i had a blatant thrower on my team (qpad.select, no less) was quite a while ago, so i don't have the thrower problem as much as you do. it's possible the distribution of throwers is not homogeneous - you might have a greater % of throwers in your bracket (3.3k - 3.7k) than in mine (4.6k - 5k)

the tinker got ganked a lot, but i didn't watch most of the map for the game - i just focused on pushing towers and helping allies.

Ok well you should pick out some data defined way to determine how hard the game was for you. Maybe you could base it on your % contribution of certain stats.

Or you could just go with "I feel like it was hard/easy"... however seriously you want to take it.

i'll think of something. didn't valve go on and on about that "gold difference" crap a while back? since people now have no incentive to drag the game out longer, it might actually be a useful metric.

the tinker lost to the dual mid, so despite rubick throwing it was actually a very good thing for us....ROFL. tinker didn't throw the game - just got outplayed. in fact he said in all chat he was going to get to 4800 mmr if he won, so he was ~4775 and i doubt he would want to lose the game intentionally

dear god i watched 2 minutes of that and my last hitting is atrocious lol, i need to work on that for the 5k prize

Well I gotta say a Tinker that dies 13 times to that lineup is either throwing or doesn't have any experience with the hero, since there aren't any heroes in Radiant who can start him from far away. Also going Force and not Dagger in this game was poor decision making by him I guess.

Hmm, words are one thing... but rubic actually helps SF mid a lot. Rubic makes this tinker have a terrible start and secures the FB kill as well which was not a ks... tinker blinded the CK. Are you sure Rubic + SF are not friends in a party and trying to pysch out the tinker. They play like they know and trust each other.... despite the words. I'll keep watching and see what Tinker does.

yes he did play poorly, that's what i'm trying to say. is it possible he rarely/never plays tinker and just sucked on that hero? certainly, but there's no way to tell since he's anonymous. i didn't think he was throwing at any point and the reason he sucked was because he kept dying to sf/rubick mid early on

10:30 he tries to participate in a teamfight near their t1 bot tower and died to nether ward + rubick bolt - 3 deaths in 10 minutes

12:30 dies to nether ward but he was low on hp anyways

whenever i put my cursor on tinker, he's not farming - probably too scared to die to ganks or something with just boots/bottle/soul ring

relentless i don't recall any parties in the end screen except the furion + lion.

string of bad decisions early on = shit score. he might have been throwing later but he certainly wasn't at the beginning

Well I guess with CD he was forced into it perhaps. But this anonymous tinker plays like he has only the vaguest idea of how to play tinker. He gets a soul ring and bottle as if he is going to use a lot of mana (ie early rearm) but then he doesn't even skill it. He doesn't seem to know the range of his spells. He maxes laser-rockets when he has a bad start and so he clearly can't gank.
He always casts rockets before laser... often ensuring the he can't cast laser because the target moves out of range. He even tries to jungle with level 1 march and no rearm while his team is getting pushed in and he could be marching the waves at the tower.

He has either never played tinker or is throwing. As of mid game I don't see how it could be anything but a 5... I guess it gets harder later somehow.

a lot of people in the pink/blue spot usually assume they are the highest mmr / "most capable" so he probably felt obligated to go mid. that's true in cases where your mmr is an outlier (something like > 5.3k)

can we conclude he wasn't throwing but just incompetent?

sometimes i meet people who got brown spot or orange spot also play much better then pink/blue. So i think blue/pink more competent just a assumption.

could be a 5...but they actually got 1 set of rax cuz nobody tp'ed. they had the potential to play rat dota but didn't do it well

Yeah, a thrower slips up and shows their real skill now and then when pretending to be bad. This tinker is just uniformly clueless the entire game. I really have a hard time believing he is anywhere near 4.8k MMR, but who knows. He is anonymous so maybe he played all games of one hero to get there... but he is not even good at clicking it just doesn't make sense. Maybe he was drunk or maybe he just lied and said 4.8k when he meant 2.8k. Maybe he is someone's little brother sneaking in a game while they are away. This tinker does everything wrong. He ports out with half mana because he forgets to use soul ring and bottle to recharge the mana faster... then ports anyway. He attempts ancient stacks 4 times in the first 30 min and misses the timing 3 of the 4 attempts.

The question then becomes how do you decide if this is a 4 game or a 5? Do you want to base it on just the valve "integral of gold graph" or team Kill scores or towers or your game impact vs your team with KDA, hero dmg, tower dmg? Or for support heroes you could look at KDA/cs taken to judge efficiency.

there's no way he's 2.8k because my queue time was around 1 minute and i'm currently 4649. in order for a 2.8k player to get in there, i'd have to wait something like 7 minutes and queue some retard mode like limited heroes pool/least played (in unranked), if getting a 2.8k player was even possible.

zenoth explained in many mmr "qq" threads that people might be playing a lot worse than they normally do, for many reasons. one or more of those reasons is probably the best explanation.

whlie i do appreciate the discussion, i will probably limit the feedback to 5-7 minutes on steam chat next time so i don't try and delineate every minor detail (i.e. "which team had a thrower," etc.)

i don't mind putting the game down as a 5, but i will consult 1-2 others players about it

I'm just wondering if you want to establish a consistent way to decide games score (1,2,3,4,5). If you do I can help with it. Otherwise you can just go by opinion. It honestly won't be that different than a mathematical method but some people would take it more seriously.

i do - and we can test that for 10 games to see how much the method varies from opinion (including my own...after all i'm the participant)

Personally I haven't met people with the intention to deliberately throw unless we are 99% sure to lose and they throw away that 1% of hope.

Your approach is valid but doesn't reveal anything about matchmaking. It will only show that the 1s and 2s are simply people picking roles that they are unable to play sufficiently well for the average rating of that match, or bad-mannered players throwing on purpose. Determination to throw or mismatch of roles are not part of matchmaking's job. From a personal perspective it may be interesting though.

Anyway the game I mentioned was probably a 3 in my opinion. My team played terribly but I compensated enough to keep the game balanced. In such a state any misplays by me will immediately swing the game heavily in their favor.

There are several aspects to judging the balance of a game we might consider. Kill score, Tower Score, Gold, Experience, and Time all might be important.

Easy to win ------------------------ Easy to lose
Far ahead on kills ---------- Far behind on kills
Far more tower dmg ------ Far less tower dmg
Far higher gold ------------------ Far lower gold
Far higher XP ---------------------- Far lower XP
Short win - Long win - Long Loss - Short Loss

Then we could make a composite score to show the balance of the game.

5 ------------------------ 4 -------------- 3 ------------- 2 --------------------- 1
(+20 kills) -------- (+10 kills) ------ even ------ (-10 kills) ------ (-20 kills) Kill Score
(+8k dmg) ------- (+4k dmg) ----- even ------ (-4k dmg) ----- (-8k dmg) Tower Score
(+1k GPM) ----- (+500 GPM) ----- even ------ (-500 GPM) ----- (-1k GPM) Gold Score
(+1k XPM) ----- (+500 XPM) ----- even ------ (-500 XPM) ----- (-1k XPM) Experience Score
(Win < 15 m) -- (Win < 25 m) ------------- (Loss < 25 m) -- (Loss < 15 m)

This would give a total of 25 points per game. The easiest game would be close to 25 and the hardest close to 5 points. You could then divide by 5 and get a composite score.

http://dotabuff.com/matches/531754588 this game would be
1 Kill Score
2 Tower Score
3 Gold Score
2 Exp Score
3 Time Score

Composite score 2.2. Even though I had a thrower on my team, the other players attempted to win and we were actually ahead early.

http://dotabuff.com/matches/521046528 this game would be
1 Kill Score
2 Tower Score
2 Gold Score
1 Exp Score
2 Time Score

Composite score 1.6 for the game the warlock intentionally threw from the start.

http://dotabuff.com/matches/530008552 your first game you wanted to use, which you judged a 3
3 Kill Score
5 Tower Score
4 Gold Score
4 Exp Score
3 Time Score

Composite score 3.8. The game was maybe a bit more in your favor than you thought. You are up 14.8k to 1.2k on tower damage. They never took any map control.

http://dotabuff.com/matches/531731793 The much discussed tinker game
4 Kill Score
4 Tower Score
3 Gold Score
3 Exp Score
3 Time Score

Composite score 3.4, overall it became a close game by the end. Your team did not push the advantage and fed 5 or 6 kills mid game giving them a chance to catch up. Late game NP was a real threat to your base and gyrocopter became fairly powerful.

http://dotabuff.com/matches/448078670 a game I gave as an example of a forced win
5 Kill Score
5 Tower Score
4 Gold Score
5 Exp Score
3 Time Score

Composite score 4.4, an extremely easy game. My KDA on Viper for this game is 26... absurd; I'm not even good at Viper.

I would definitely change the time factor, or remove it altogether. Most of these games you used (despite having vastly different scores) are 3's on the time scale, which means it's skewing the distribution.

I have rarely, even in games lost horribly, come across games that last fewer than 15 minutes, unless it's an abandon. Getting a game under 15 is almost impossible unless at least 1 player is throwing as hard as possible.

I'll list some games that I think should be "1's" or "5's" to prove my point.

http://dotabuff.com/matches/521083944

I ended up abandoning this early on because my centaur went top to lane with me, and I played like a retard. 22 minute game, my team was playing 4v5.

http://dotabuff.com/matches/521046906

I played like a retard in this game; I was pretty pissed off from the previous 2 losses so I instalocked morph - bad idea. 28 minute game.

http://dotabuff.com/matches/520994403

Luna obviously bought the account or someone else not rated at the appropriate MMR is playing on it. Either way he played like a 2.5K MMR player, not getting luna ult at 6 and dying to a dual lane in a trilane 3 times in less than 8 minutes. 28 minute game.

http://dotabuff.com/matches/509570083

Bottom lane for dire destroyed safe lane for radiant, resulting in an absurdly skewed 40-6 (nobody was throwing - they just got outpicked/outlaned hard). Despite this the game was 27 minutes long.

Using the metrics you prescribed, none of these games except 1 would be classified as a 4 (rest would be 3's), and almost no games in general would be classified as a 5 on the time scale.

< 25 minutes is a reasonable metric for a 1 or 5 game, and < 35 minutes for 2's and 4's. It's just unlikely, no matter how bad one team performs, that a team will throne in 15 minutes. Killing the t3 towers and everything beyond is difficult for most heroes, even pushing ones such as pugna/lycan because an underfarmed support can mess with you and force you back.

Despite this, ending a game at 25 minutes can still imply the game was extremely skewed. Why? Because if you kill their heroes over and over, they will simply turtle (for fear of dying), and teams that turtle can prolong the game by 5-10 minutes if they do it properly. I could pick sniper, go 0-10 in 8 minutes, and still ensure the game lasts 25 minutes or longer by spamming shrapnel whenever they want to push high ground.

The time factor needs to be normalized, so it remains consistent with the other metrics and doesn't skew the distribution. The other numbers look fine to me.

Since people may not be trying to win the game quickly time may not be the best indication of the balance. Some games can drag on with fountain farming or just plain farming instead of pushing. No matter how far ahead you are it can be difficult to get past T3 and T4 towers even with no one defending until you have at least some core items done.

If we take out the time factor is there another factor that might also be important instead? I had initially though we might use teamwork measure it by assists. But then it seemed to put too much emphasis on killing heroes when I put in the numbers so I removed it.

the time factor is useful, but should be modified. i recommended:

1/5: < 25 min
2/4: < 35 min

basically i've had a bunch of almost unwinnable games that lasted longer than 20 min (but tended to end at around 25-28) and games that i was unlikely to win end at slightly over 30 min. in ur own examples, the lopsided games lasted a lot longer than what the time factor would have predicted.