Guild Wars 2

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

YB already got locked on T2…

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

wow, I just looked at this week’s files for the first time myself. NA T1 and T2 are both completely locked now.

-ken

Snowreap.5174:

wow, I just looked at this week’s files for the first time myself. NA T1 and T2 are both completely locked now.

-ken

Nothing says you’re in a rut like another run of “Infinity bottles of beer on the wall, infinity bottles of beer…”

YB got a number of xfers during and after seasons – well a lot for us anyway. At times, I think we outnumber SOS / fa.

Liston.9708:

YB got a number of xfers during and after seasons – well a lot for us anyway. At times, I think we outnumber SOS / fa.

Isn’t the score enough to prove that? We all know ppt doesn’t come from superior skills but from superior coverage.

Snowreap.5174:

wow, I just looked at this week’s files for the first time myself. NA T1 and T2 are both completely locked now.

-ken

That doesn’t need your simulation, that’s clearly visible at http://mos.millenium.org/na/matchups
Rating gap is above 2*(10 + 0.45* dev) ~ 152 in case of t2/3

Rimmy.9217:

Nothing says you’re in a rut like another run of “Infinity bottles of beer on the wall, infinity bottles of beer…”

“Ninety Nine barrels of Ale on the wall, Ninety Nine barrels of ale…”

Let’s play some keg toss! Oh wait… Anet made some stupid rotation for activities now… We can’t play it when we want.

Snowreap.5174:

wow, I just looked at this week’s files for the first time myself. NA T1 and T2 are both completely locked now.

-ken

Yup, pretty much in T1 its always going to be Jade Quarry, server B, server C. If any server, I’d wager Blackgate be the next server to get whooped off T1, its spent enough time and that servers stability isn’t that great either, its comprised of a lot of other guilds unlike a true T1 server like Jade Quarry, the true server full of T1 quality guilds and players that didn’t come from other servers.

But yeah, T1 always is locked, its always 2 worlds full of wannabee T1 guilds going up against TRUE T1 quality guilds from Jade Quarry.
Back when it was JQ SoS SoR…. T1 was locked. Got unlocked by server getting a bunch of wannabees… Than if I recall went from JQ SoR BG… BG was the T1 wannabee guild server. Than JQ BG TC, with TC being the guild wannabee server. If anyone does drop from T1 it’d defiantly be BG.

Because Jade Quarry never, ever, received transfers, right?

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

Give me that 8% chance of GoM/DH! SoR ded sevr, we don’t want them anymore…

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

May RNGeesus smile upon Eredon Terrace this time… I’ll die if I have another T7 or T8 week.

And we got the 1,6% match with HoD/Mag.

Still better than SF/FC. Should be interesting…

It’s certainly been an interesting night for us at HoD. We’ve been stuck at T5 with BP and no other 3rd server we can match, and now we’re up against Mag and ET. Weird stuff. You guys fielded some impressive numbers, and that was quite a tier leap to get here. I almost wonder if the matching was done manually this week. CD got NSP and DR and SBI somehow ended up with DB for the 3rd week in a row, and IoJ. Watching millenium has been quite an experience, though, because the CD/DR/NSP match is neck-and-neck 5 hours after reset. Should be an interesting match to watch.

Good luck to us all.

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

Thanks for continuing to do this snow <3

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

SNOW! YOU ROCK!!!

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

My wish list has changed in the meantime

No more matches please.

Shutdown and fix it please, ANet.

sorry about missing last week’s numbers. it was a busy day and I forgot.

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

Thanks for doing this all, Snow, but Anet has changed its matching system since the end of Season 3. Many improbable matchups that you would expect using glicko and predicted ratings based on score have been happening — we’ve fought outside our tier almost every match in the last 2 months (Tier 5, what’s that? Hi Tier 7, Tier 4, etc), which boils down to approximately 0.006% chance as of my math last week. Less now I’m sure. Someone on our server has also told us that a dev confirmed that the matching algorithms have changed. Perhaps one could step into this thread and enlighten us.

HoDs match ups have been within the glicko draw rate.

Though it does feel like they’ve changed a bit. Then again.. We’ve been stuck with a T2 server for the past 2 months basically outside of one or two weeks.

Ark Bladesteele.2943:

Thanks for doing this all, Snow, but Anet has changed its matching system since the end of Season 3. Many improbable matchups that you would expect using glicko and predicted ratings based on score have been happening — we’ve fought outside our tier almost every match in the last 2 months (Tier 5, what’s that? Hi Tier 7, Tier 4, etc), which boils down to approximately 0.006% chance as of my math last week.

Not a single matchup has occured in NA since season 3 that suggests that the matching algorithm has changed. That being said, a small change would be hard to detect.

Imagine that you roll a 6-sided dice 8 times and record the results. The probability of the numbers showing up in that exact sequence would have been 1 in 1,679,616 (~0.00006%). However, there’s nothing remarkable about that particular sequence. The probability that any one of the 1,679,616 possible sequences would occur was 100%.

Baron Oakley.1539:

Not a single matchup has occured in NA since season 3 that suggests that the matching algorithm has changed. That being said, a small change would be hard to detect.

Why would the match up vary much when no server is doing anything to strongly suggest they should move up or down a tier?

My problem with your statement, is that you state it as if some servers pulled off some miracle play and deserve to move tiers for new match ups.

Baron Oakley.1539:

Imagine that you roll a 6-sided dice 8 times and record the results. The probability of the numbers showing up in that exact sequence would have been 1 in 1,679,616 (~0.00006%). However, there’s nothing remarkable about that particular sequence. The probability that any one of the 1,679,616 possible sequences would occur was 100%.

What does any of that have to do with anything? The system is has nothing to do with luck or dice, it is based on comparative performance over a duration of time.

Players make complaints with things of this nature, are in fact their fault. If you want a new match up, you need to recruit, practice, coordinate, to move up in tiers and perform well. People need to stop blaming the system, and see it for what it is.

In reply to above:

1) All I said is that no matchup has occurred since season 3 that would be impossible under the matchup generating system described to us by ArenaNet . This is a fact. I do not believe that “miracle play” has occurred.

2) “Luck”, as you put it, is in fact part of the matchup generating system. (Presuming that by “luck” you mean some random component, which is what I was referring to.)

If you refer to the page linked above, you can understand how to generate tables with server probabilities like those provided at the bottom of that webpage. The matchups are not generated purely based on server rankings or ratings alone, there is a random probabilistic component.

Edited: Unnecessary snark removed. I am sorry, didn’t mean to offend. Also, missing a very important word.

Baron Oakley.1539:

1) All I said is that no matchup has occurred since season 3 that would impossible under the matchup generating system described to us by ArenaNet . This is a fact. I don’t understand why you took it to imply that I believed that “miracle play” had occurred.

What does “would impossible” mean? Would be impossible?

Baron Oakley.1539:

2) You clearly don’t understand how the matchups are generated. “Luck”, as you put it, is in fact part of the matchup generating system.

After you have read and understood that, you should then be able to understand how to generate tables with server probabilities like those provided at the bottom of that webpage. Please note that if “luck” had “nothing” to do with how matchups were generated then each week would have matches that were simply the server rankings 1-3, 4-6 etc.

Please specify the section of anything you linked that says “luck” please?

And no, the statement above is not all you said. You went into some story about rolling die and all that. I don’t think anyone would have disagreed with much about what you said until you went off on your blind luck, roll of the dice tangent.

What is actual fact is that when you place in last, and do so by a certain level of points, you have a score variant that when reaches a thresh hold, you can be bumped down a tier, as long as the winner in the lower tier wins enough, to reach the point that they reach a ranking score thresh hold that will allow them to move up to the next tier. There is absolutely no luck evolved what soever.

As well, you seem to confuse subjective opinion with objective fact.

For someone being so snarky in your post, you would think you could grasp the difference between luck and “probability”, which is the only term anywhere in there that could possibly be misconstrued as “luck”.

dancingmonkey.4902:

What does “would impossible” mean? Would be impossible?

Corrected. My mistake.

dancingmonkey.4902:

For someone being so snarky in your post, you would think you could grasp the difference between luck and “probability”, which is the only term anywhere in there that could possibly be misconstrued as “luck”.

You are right, of course, that luck and probability are not synonymous. However, if you recognize that there is a random component in the matchup generating system, then you should be able to accept that one may describe a particular outcome as “lucky” or “unlucky”. I presumed that when coglin said that “luck” had nothing to do with how matchups were generated he was suggesting there was no random component. I made this presumption because he had responded to my original post which made reference to probabilities (I did not mention luck) with a reply that “luck” had nothing to do with it.

I apologize for the tone of my original post, it was snarky and unnecessarily so.

dancingmonkey.4902:

And no, the statement above is not all you said. You went into some story about rolling die and all that. I don’t think anyone would have disagreed with much about what you said until you went off on your blind luck, roll of the dice tangent.

It wasn’t a random tangent. Ark suggested that the exact sequence of matchups that HoD has had since season 3 has had a very low probability. My point was that you can’t use probability that way, because if you rolled a dice 8 times (by analogy with 8 weeks of matchups) that exact sequence would have been very unlikely to occur. However, you would not conclude that there was anything wrong with the dice, because one of the 1,679,616 possible sequences had to occur.

Edit: Okay, so it turns out there’s been a lot more than 8 weeks since the last tournament. But it was intended as a theoretical example.

I never said impossible. I said highly improbable. Which it has been. Not just my tier, but others as well.

It’s true that from the outset, it’d have been highly improbable for HoD to get the sequence of matchups that it did. The point I was attempting to make (perhaps clumsily) is that it’s dangerous to use probability to argue backwards.

For 11 weeks after season 3, HoD has had the most likely matchup or 2nd most likely matchup seven times, and for the other four weeks has had an improbable matchup with a likelihood of about 2%. This is actually not very surprising.

It’s true that the chance of this exact sequence of matchups was vanishingly small (about 1 in 110 billion). But even the chance of HoD getting the most likely matchup for all 11 weeks in a row would have only been ~1 in 145,000. To put this in perspective, the chance of HoD never getting the most likely matchup for all 11 weeks in a row would actually have been ~1 in 124. This particular scenario is more likely than the others, however, because it allows for any matchup each week except one.

Even the probability of HoD getting either the 1st or 2nd most likely matchup for all 11 weeks in a row would have been ~1 in 1220. To put this another way, you would expect there to be a 99.92% chance that in at least one of those 11 weeks HoD would get an “unlikely” matchup. (ie. not the 1st or 2nd most likely)

TL;DR: After 11 weeks pretty much any specific sequence of matchups is going to seem unlikely, and there’s nothing particularly surprising that in an 11 week sequence that a few unlikely matchups would show up somewhere.

Having four T2 wvwvw population servers really plays havoc with T3 and T4 level ones.

Predicting that 3% chance of seeing DB (350+ glick points ahead of us) is going to come to fruition again. That’ll be fun!

I like to use the lottery as an example (the kind of lottery where you pick, say, 6 numbers out of 36, and the jackpot can get into the millions).

each week, the numbers that come up are extremely unlikely (so unlikely that many weeks have no winners at all), but despite being unlikely they came up anyway. the unlikely results continue to occur week after week.

if a matchup comes up that was predicted to have 0% probability, that would be a fairly good sign that the matchup algorithm may have been adjusted, but it would not be a certainty because when I calculate probabilities I only generate 100,000,000 random matchups (to save time). there are some possible matchups that are so unlikely that they wouldn’t show up in 100,000,000 test runs, but no matter how unlikely they are, they can still happen.

just like lottery numbers.

there are also matchups which are mathematically impossible according to the known matchup rules. if we saw one of those matchups occur, we could be certain that something has changed.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

here are the estimated probabilities for the new matchups, assuming that predicted ratings are reasonably close to what we will see at the end of the current matches. as always, if predicted ratings turn out to be different from the actual end-of-match ratings, that will throw the numbers off.

the “na1” and “eu1” files have probabilities of getting a particular server as an opponent, while the “na2” and “eu2” files have probabilities of getting specific pairs of servers as opponents.

also, I’ve included the predicted ratings that I use for these calculations. these are the “na” and “eu” files. these values should fairly closely match predicted ratings from other sites at the time I post them.

-ken

And we just got the match that only had 1.18% chance of happening.

yeah, sometimes the unlikely matchups happen. it would in fact be very unlikely to get only likely matchups. there are statistical randomness tests out there that verify that you get the expected number of unexpected results. or something like that, I’m not an expert.

here are the estimated matchup probababilities for the upcoming matchups. same files as usual, with the same naming convention.

-ken

Hmm interesting. Apparently CD showed up and with its charm, caused a divorce between GoM and DH.

0,98% of IoJ/NSP/SBI… make it happen, Anet!

azyume.6321:

Hmm interesting. Apparently CD showed up and with its charm, caused a divorce between GoM and DH.

0,98% of IoJ/NSP/SBI… make it happen, Anet!

Oh that would be awesome. Otherwise I’m hoping for SoS.

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken

here are the estimated matchup probababilities for the upcoming matchups. “na1” and “eu1” files show the probabilities of getting any single opponent, “na2” and “eu2” files show the probabilities of getting a particular pair of opponents. the last two files are the estimated rating outcomes for the current matchups, assuming that final scores don’t meaningfully differ from today’s scores.

-ken