General Discussion - Page 3359

>Match ID: 246510650

All 10 players in this game were in the 93rd-94th percentile MMR range. The difference in Elo's between the highest and lowest player was 50 Elo points. The "noob" with only 13 wins actually had the highest Elo of any player in the match. (This was a smurf.) They did play poorly in this match, but in the previous match (246456658), in which they played against several 4000 Elo players, they had 13 kills and 1 death.

>Match ID: 246567215

The 14-win player did have the highest MMR in the game. However, it probably wasn't really high enough to push him into smurf territory. So this was not a great match. There was a player who had waited 5 minutes, which is why a match of this relatively low quality was accepted. Thanks for reporting this. It's very helpful.

We're trying to find a way to provide more information about matchmaking, both how it works in general, and also why particular matches were formed. We're looking for a way to do this that does not lead to abusive behaviour now, where any lopsided game is immediately blamed on the person with the lowest win count, as opposed to, for example, somebody playing a hero they aren't familiar with or just having a bad game. This blame is often misplaced, but it's understandable since the win count is the only piece of data currently available.

>I still got 600 win vs 60 win players.

If you post a match ID, I might be able to help explain why it decided to form the match. Either it will help us improve the tuning, or there's probably some piece of data that makes it make sense that we are not exposing.

We are looking for a way to provide information to players to help them understand why matches were formed in a way that doesn't lead to abuse of whoever the "worst" player is (according to whatever metric is preferable to the person giving the abuse). In the meantime, walking through some particular matches will help us better understand where it's still getting things wrong, or educate players about how it's making its decisions.

Just to clarify, there was never any "win rate" calculation. Ever. It is true that a goal of matchmaking is to make even teams, so that the odds of the Radiant winning any given match is 50%. The matchmaker also will raise your Elo and try to put you in players of equivalent skill, which indirectly tries to get the win rate to 50%. However, it has never looked at your historical win rate and, for example, put you in a game where it knew that you were expected to lose, to end a winning streak, or given you a stomp to end a losing streak.

>246572710

The game looks pretty balanced to me. The Elos are all relatively close. Here are the Elo's on the two teams:

R D
3172 3047
2918 2958
2810 2804
2788 2720
2409 2508

Now it's a pretty big spread between 2409 and 3172, and it is a legitimate question to ask why in the world would we put people together with that big of a skill differential. The answer is that we didn't. The Radiant had a 4-stack which covered that range (the highest and lowest Elos on the Radiant were in the same party). We matched them with two 2-stacks on the Dire. One of the two-stacks had the highest and lowest Elos for the Dire, and the other two stack was in 2 of the middle slots. You were a single who was also in the middle, and there was also a single on the Radiant, also in the middle of the Elo range.

So, the average Elos of all the parties were pretty close. And, player-by-player, each team had somebody on the opposing team of roughly equal skill. Given that 4 stack with the big skill spread, I think it's hard to come up with a better way to get them into a game.

Would you mind being a bit more specific about what you think could have been better?

Note: all of the players had sufficient experience such that we rely almost entirely on Elo. (We do not consider a player with 75 wins to be a "noob".) In this match, there was no correlation between the Win count and the Elo. Perhaps 75 is too low of a number? We have some pretty good data that suggests that it's not. But averages and numbers often don't tell the whole story.

"Noob" is a relative term. We don't consider a person with 150 games to be a "noob". We have some good data that by 75 wins (approx 150 games), Elo is pretty accurate, and so we rely on it almost exclusively at around that point. If you are getting matched with those players, it should be because your Elo is approximately the same. Parties can complicate things considerably. I might be able to provide further insight into why it thought the match would be a good one if you provide a MatchID.

Is 75 the wrong number? Should a person with 2000 wins complain about playing with you, since you are a "noob" with only 1000 wins?

Hi Luna, I think I already did: It was not a great match. Your party of two didn't have any inexperienced players in it, and so you shouldn't have been matched with inexperienced players. (The other experienced players WERE parties with newer players, though.) The 14-win player is borderline smurf, and there might need to be a number that needs some tuning. And he was partied with an experienced player (whose Elo was 1000 points lower, BTW), but still, if you are not partied with an inexperienced player, in general you should not play with an inexperienced player.

I think that the reason that it formed the match is that somebody had been waiting a while, and it needed to fill that last slot, and your party was the best it could find. We're still working on how it decides when to give up.

>246581071

Congrats! You don't have a crappy Elo, and can feel free to keep playing Dota! Your Elo was right in the middle in this game.

You were on the Dire. The Dire had a 2-stack that had an Elo spread of 1000 points. The guy with 150 wins was the bottom half of that party, and the other "noob" with "only" 298 wins was the player in this match with the highest Elo. We paired them against a corresponding party with an Elo spread that was approximately the same on the Radiant.

In summary, it is true that you were matched with a player with an Elo lower than yours, because he was in a party with an average Elo approximately equal to yours. This means that you were also matched with a skilled teammate to compensate for the lower skilled player on your team. (But you didn't know that, because win count is the only metric you have to evaluate the skill of the players in your game, and win count is really not useful past a certain number of wins.) Furthermore, the match was fair because team you played against had a very similar situation. Comparing player-by-player, the teams were balanced, and the match was just about as good as could be formed based on Elo, given the fact that we allow parties to be made with players with differing skill level. Parties CAN cause players with differing skill to be put into the same game. There is no way around that, however, in most cases the matchmaker now does a significantly better job of matching up parties to make equitable teams than it did before.

If you want to avoid this situation and not matched with parties, try the solo queue. We've solved several problems that were causing long wait times and low quality matches. (We should have mentioned that in the patch notes.)

If you party with a noob, it is very likely that you will get matched with a noob. (Very likely it will find another party containing a noob and a player of approximately your skill level and put them on the other team.)

If that situation is not ideal, might I ask who you think you should play against?

Yes, we use a Elo-type system. We have data that shows that it correlates well with a number of quality metrics.

Showing players their Elo has proven to cause a lot of unwanted behaviour. (The same problems exist in Chess.)

We know that there really is value in knowing how the match was formed, and even in knowing which players were the best, and also who was partied up with whom. It helps to analyze your own skill and try to understand how to get better. We are actively working to try to figure out what data we CAN provide to players (perhaps after the match) to help them do this analysis, without causing too many of the undesirable effects I mentioned. Right now, if there's a steamroll, people look at the only data available to them, which is totally understandable. Also, I believe that there have been problems in the past where it was not doing a good enough job segregating truly inexperienced players from experienced players. We've improved on that. However, our definition of "new player" probably doesn't match exactly with many players. (Their definition is often: "the other guy has fewer wins that I do.") And we have to deal with smurfs and players coming in from other games in the genre, where win count is definitely totally unrelated to the skill level of the player.

Just one last comment on this. Elo is a TERRIBLE way to give players a sense of "progress." Many (most?) people reach a plateau, and their Elo stabilizes. It is simply not mathematically possible for Elo to keep increasing in general for players indefinitely as more and more games are played.

Given this reality, if players used Elo to measure "progress", we would constantly be reminding them that they are NOT making any. That would be really bad.