Merlini’s Mailbag and Solo Queue Clarifications

Woke up this morning, randomly clicked on a video link, and found my solo queue article getting mentioned.  A bit surprising.  Merlini gives it a fair treatment, and you can find it here if you like.  I’d also generally agree with his opinion on Vanguard right now.

Anyway, there are a few things that people have misinterpreted, really more in the reactions to the video than the video itself, so I thought I’d issue a short follow-up.

First, the letter Merlini reads misinterprets the proposal.  It’s not making every form of matchmaking other than AP into solo queue; it’s just making one particular mode (RD was suggested, but it could be any or a new mode) into the solo queue option.  Duplicating every mode would needlessly fracture the playerbase.  The goal instead is to keep the number of matchmaking options constant, but try to elevate one into a distinct secondary option to the current AP juggernaut.

One common response is that the stack win rates are no big deal because the loses are spread out over a large population and therefore barely noticeable.  I’m not going to say that this is impossible, but I think there are a couple factors that make it unlikely.  I’d expect one of two outcomes depending on how the matchmaking system handles it.

The first possibility is that the matchmaker recognizes that these stack matches are relatively likely losses.  In this case, the non-stacking team only takes a minor MMR hit upon losing.  But the non-stacking team was thrown into the match because they were the closest people to the top, and if they’re not losing much rating from losing, they’ll still be the closest to the top.  As a result, these particularly frustrating matches aren’t actually diffuse and instead concentrated in the very slightly less narrow band of players below the top 1%.

The second possibility is that losing to stacks does tank your MMR.  In this case, we have a scenario where the rating of the players in, say, the 2nd to 5th percentile range fluctuate depending on how many stack matches they happen to have triggered recently.  In this case the stomps are spread out, but as a result the ratings of the players in a larger range are now much more volatile than they would be otherwise.  An unlucky stretch of triggering stomp games will be fixed by those players now being underrated and stomping on players a few percentiles down the line.  What this ends up doing is lowering the matchmaking quality in the tier below the stomp tier because they’ll feature more underrated players who are only correcting back to their “true” rating.  So the losses are spread out, but not in a good way if you care about game quality.

My belief is that the only way this isn’t an issue is if the number of highly successful stacks is exaggerated.

Finally, the one thing that stood out to me was that Merlini mentioned that he and other people play ixdl when they’re frustrated with general matchmaking.  I don’t have an intimate knowledge of how ixdl works, but the description of it sounds an awful lot like a privately run solo queue, with the only difference being that since it’s invite only the players have more trust in the system to give them a fair game.  I think a lot of the frustration I hear is basically from people who want the ixdl experience without having to go through the vouch process.  It’s not as though the ixdl could just expand to contain them all because part of the experience is the social element that is maintained through the vouch barrier, so there needs to be additional outlets somewhere.

There’s not really a specific point here, just more of an observation.  Sometimes I wonder if we should just accept that top end matchmaking needs to work differently.  For instance, if top groups are going to deal with 15-30 minute queue times anyway, should we bother using continuous matchmaking?  Perhaps it would be better for Team Matchmaking to just say “Enter the queue whenever you want, matches will be formed for everyone in the queue every 30 minutes.”  In this case you’re waiting longer, but you know precisely how long you’ll be waiting, and you have more confidence that when you do get a match it will be against a more closely matched opponent than if the system just settles for the first “decent” matchup.

In any case, just having a solo queue shouldn’t be considered the long-term fix to any of Dota’s matchmaking ills, perceived or otherwise.  But if people really are burning out from stack fatigue, a short-term fix might be warranted if we want to maintain player population at the narrowest end of the distribution.

For those of you still waiting for actual statistical stuff, I had something ready but decided to delay it in light of Bristleback.  We’ll get back to your regularly scheduled programming either tomorrow or sometime over the weekend.

Calling in Ganks and Cost-Benefit Analysis

Time for another match anecdote and a (hopefully) helpful moral to the story.

All Random Radiant.  I swap to take Lone Druid and also accept mid.  Our bot safe lane is a Skeleton King/Centaur Warrunner combo, which is reasonably decent.  Our suicide solo is Sand King.  We have Nature’s Prophet as a jungler.  I think our lane choices were perfectly fine, with Sand King and Centaur’s positioning being more or less interchangeable.  The opponent’s send Clinkz mid, Luna and Ogre Magi to top safe lane, and Tinker/Dark Seer to bot hard lane.  Clinkz mid instead of Tinker is a little weird, but everything else is pretty standard.

Relatively early in the game (I can’t tell you precisely when because demo’s do not record team chat but definitely before 10 minutes) our Sand King requests a gank on his lane.  I’m a bit curt with him about how we just need him to get a quick 6 and gank.  He complains that we’re just letting Luna get freefarm.  At the time I wanted to give him a more elaborate response, but playing Lone Druid still demands way more attention from me than it ought to so I let it slide.  Here’s what the response would have been.

Lone Druid as a mid will not, and should not come help you.  If letting Luna have freefarm is bad, letting Clinkz have freefarm is an order of magnitude worse.  Lone Druid’s accomplishing everything he needs to focusing on the Clinkz 1v1, and if Clinkz does start roaming, LD’s bear has the pushing power to punish the enemy team for it by taking their most valuable tower.

You are correct that after Clinkz, Luna is the biggest carry threat on their team by far.  It would be nice to shut her down, but at the same time most of our team is across the map from her.  Instead of trying to shut her down, the three of them could converge on a relatively weak Tinker/Dark Seer lane.  Tinker ought to be an easy pickoff for Skeleton King/Centaur/Nature’s Prophet.  While it unfortunately did not work out that way, conceptually it’s by far the safest way to play it.  Freefarm Luna is not a good thing, but if we’re getting a freefarm Skeleton king, a farming Nature’s Prophet, and a lot of kills out of it, while our Sand King is getting a relatively unopposed quick 6, then that trade is working out in our favor.

To safely threaten Luna, we probably need to move 2 people top.  Prophet can just teleport there, but given our lineup this basically demands sending Centaur.  Doing this greatly exposes our Skeleton King so it’s immediately a risk.  If the gank isn’t successful then we’ve spent a lot of resources not stopping her farm while greatly inhibiting our own carry’s farm.  Meanwhile, we’re ganking into two stuns, and Luna’s ultimate is an extremely threatening countergank tool.  By comparison, focusing our efforts on making Tinker and Dark seer miserable is certainly not as valuable.  But it’s much less costly because no travel time is involved, and it’s much higher probability because neither hero has much in the way of CC.

What really gets to me is that the gank call came across as rather selfish.  Sand King had a lane where he couldn’t farm and felt it was his team’s responsibility to buy him some farming space.  Look, an early Arcane Boots + Blink Dagger is great on Sand King, but when we have a Lone Druid and Skeleton King it’s just not a priority.  We’re fine with you not getting farm.  Just get the levels you need and go gank.  Give us some warning and we’ll even be able to send either Prophet or Centaur to top to clear waves occasionally and protect the tower.

The other side lesson of the game is if you’re playing a melee, please strongly consider buying a Stout Shield.  It’s a very small investment for a very large increase in early survivability.  You don’t need to start with a Circlet so you can rush a Bracer as suicide Sand King.  Centaur and Skeleton King also didn’t buy Stout Shields ever, and had they both had one I think they would have had a much easier time dominating their lane.

Finally, this probably sounds like a game that we lost, but we did end up winning it.  I’m not sure we deserved to, but the moment of truth was buying a Gem of True Sight, almost immediately after catching Clinkz at the start of a gank attempt, blowing him up, and proceeding to crush the rest of the team 5v4 and pushing for a win.  So please remember to buy stealth detection against stealth characters.

 

6.77c Win Rate Shifts for Nyx, KotL, Drow, and Treant

Sample is complete, and we should be on pace to do some stuff with it early this week.  For now, just a quick look at the early effects of the 6.77c changes that went into effect last Thursday.  I cannot create sufficiently large samples from such a short stretch of time, so we’ll be using DotaBuff’s daily trends for our data source.  Numbers are going to be rounded, so check out the source if you want more exact totals.  And keep in mind that a 4 day trend if very preliminary, so grain of salt and all that.

Nyx Assassin

The relatively minor hits he took to Spiked Carapace appear to be relatively minor indeed.  His 3/22 – 3/25 win rate hovers around 54.3%, down from 55% over the entire month.  In Nyx’s case, it would be very interesting to see the effects of this change at different skill levels.  What I’ve seen suggests he performs significantly better in high level games from his already high baseline.  Regardless, I wouldn’t be surprised to see him receive further changes in 6.78.

Keeper of the Light

KotL’s Illuminate nerfs on the other hand appear to have been much more significant.  His 4 day win rate is fairly volatile, but the average is about 52.5%, down from 55.5%.  He’s still definitely a viable support, especially since Illuminate spam is likely a much less integral part of his kit at higher levels of play.

Drow Ranger

Our long national nightmare is over?  Ok, so it’s just 56% -> 53%, but at a pub level this is looking pretty good so far.  She’s still quite strong, but given her weaknesses she needs to be to have any kind of a presence in higher level play.  She might demand further tweaking in the future to try to normalize her with respect to varying skill levels a la Ursa, but for now this might actually be a solid balance point.  It’ll be interesting to see if she gets a Captain’s Mode addition soon (along with Centaur).

Treant Protector

The most interesting story of the patch, Living Armor might have gotten the final tweak it needs to deliver precisely what it promised.  Treant’s 4 day win rate is up to 51.5% from a previous monthly average of 45%.  It’s a pretty huge jump, which suggests to me that 6.77c was actually a huge boost to our dendritic friend.  If the change proves to be as significant as it appears, we could see Treant being picked up as complement to a trilane vs trilane strat, so he can be the 4th participant in the trilane from across the map.  It definitely requires some coordination, but if the Living Armor buff alone is responsible for a 6% jump at the basic public matchmaking level we might be looking at an incredibly potent spell to add to the arsenal of a competitive lineup.

Is It Time for a Dota 2 Solo Queue?

Back in my Dota Matchmaking FAQ series, I said that in my opinion a lot of the perceived failings of Dota’s matchmaking are social issues rather than technical failings.  Today I want to talk about one of those social issues that keeps popping up: the problem of highly rated stacks, stacks being a group of players who queue up together and play in a highly coordinated way.

Dota 2’s matchmaking allows players to play in a group.  It needs to allow this.  Some people cannot enjoy the solo queue game, and Valve is obligated to offer them a form of casual matchmaking so they can play with their friends.  In order to expedite the process, matchmaking is willing to allow these pre-made groups to face solo players because it selects solo players with a higher overall skill level to compensate for the pre-made group’s greater level of coordination.  It’s not a perfect system, matchmaking never is, but for the most part it gets the job done, and I expect it’s still undergoing constant refinement.

The problem this system runs into is that when a group of sufficiently highly rated players queue together, the matchmaking system doesn’t have a sufficient population of even higher skilled solo players to pick from.  The matchmaker does what it can to make decently competitive games while not allowing queue times to skyrocket, but the results are not pretty.  The picture here is a little clouded.  We can only see who these players have played with in the past month, inactive players are not filtered out, and Dotabuff cannot show us players who have decided the remain anonymous.  Nevertheless, if you click on most of the players on that list you will find many who regularly play in groups that achieve +70% win rates.  It’s true that Dotabuff only shows us 50 people who have a greater than 70% win rate, but it is very likely that win rates are distributed in such a way that the 60-70% group has a much higher population than the +70% group.

What this means is that there must exist a group of players who are absorbing those 30-40% win rates on the other end.  Solo queue players who are constantly matched into games that we know are not statistically competitive but that are created for expediency.  What this does is create a burnout point, where the primary way these players interact with the game is so frustrating that they lose interest in the game as a whole.  Some of these players will quietly quit, or at least scale back their playtime, and we will then just throw the next tier of players into the grinder so that the cycle can continue.

Some say that these players should be grateful that they have the opportunity to play against the pros and learn from the experience, but I find this argument rather unconvincing.  If you were to create a game with 8 talented amateurs and a profession player on either team, that could easily be an educational experience for everyone involved.  You’d have a source of direction and insight that you might not normally have while also being tested by a similarly directed team.  Being stomped by a pro+friends stack in under 15 minutes?  Yes, there are things that you can learn for it, but if it is the 5th time that night you’ve been stomped in such a manner then you’re probably at the point where you’re no longer learning much.  And let’s remember that if you can recognize a pro player’s stack by the names, it’s pretty likely the other 4 people on your team can as well.  Some of those players are going to give up internally before the creeps even spawn, and being essentially down 4v5 against a far more coordinated team is not an environment well-suited for learning anything.

A alternative solo queue would address this issue, but the danger is that the solo queue could become so popular that finding a casual team match would become difficult or impossible.  As I mentioned earlier, casual team matching is a necessary feature of the system and is as important as addressing this burnout point.  So, how can we create a solo queue that won’t completely cannibalize the more general queue?  My answer starts with this:

It’s not a perfect sample by any means, but a 75% representation for All Pick jives with everything else that I’ve seen.  The most straightforward solution is to make a solo queue that does not include All Pick.  The most appropriate default mode for solo queue would be some variant of Random Draft, and if it’s proves popular enough it could eventually add something like All Random or Single Draft.  Meanwhile, solo queue is treated as the opt-in mode and doesn’t include the most popular mode in Dota 2.  These two facts should allow the general matchmaking queue to maintain at least a 30-40% All Pick representation, which should be more than enough for efficient group matchmaking.  General Random Draft might suffer, but in exchange we would have a solo queue Random Draft with a significantly higher population than we have now, which should improve the match quality for dedicated RD players.

Most importantly, this would give highly rated solo players the means to avoid stack burnout.  The question then remains, what does matchmaking do with the stacks who now have even less competition than before.  My suggestion is to just let the matchmaker search for however long it needs to to find an actually even matchup, warn the stack that the queue might take a significant amount of time to process, point out that they would have a much shorter queue time in Team Matchmaking, and then automate team creation and switch them over to that queue.  This would serve as a much needed boost the TMM population, and frankly, if your stack is at a +60% win rate in general, TMM is probably where you belong.

This is not to say that TMM doesn’t have issues right now, but the best way to address those issues is to get more players playing that mode and giving feedback on the problems they have with the mode and how they’d like to see it develop in the future.

So How Does Any of This Make You a Better Dota Player?

No big release this week most likely.  I need a few more days to complete a 40k x3 sample extension for the end of this patch period.  This should allow me to do general sample tests with a much higher accuracy for every hero.

In lieu of an actual update, I thought I’d take some time this week to talk about why any of this matters, especially to a person who’s just trying to get better at basic, public matchmaking Dota.  To do this I’ve dusted off a particularly neglected account last weekend, played some games, and we can use them as examples for how the things I do here influenced my decision making.

Now they’re not great games.  This account was inactive for a long time, and I’d have to say the general level of play was significantly lower than when I last used it.  Maybe there’s ratings decay from inactivity.  Maybe the MMR system has just filled out since then.  Maybe it’s pure placebo.  In any case, it’s fine for us because it’s a level of play where bad decision making alone can easily lose you a game.  And decision making is exclusively what I want to focus on.

So let’s start with a somewhat frustrating game that began with an interesting decision.  It was All Random.  Their team was Broodmother, Chaos Knight, Spectre, Ursa, and Warlock on Radiant.  Dire was Clinkz, Clockwerk, Enigma, Bane, and me on Earthshaker.  An stark contrast of compositions, and we’ll get to that, but for now let’s focus on lanes.

Predicting their lineup is more or less impossible.  Pub teams in general can be essentially random when it comes to who’s laning where, and in their case they had so many carries that the only safe prediction was Ursa in the jungle.

As for our team, sometimes you can lobby your team into a certain lineup.  Sometimes you can’t.  In this case Clockwerk called top safe lane, Clinkz called mid, Enigma was obviously a jungle, and Bane was heading hard/bot.  They expected me to go hard lane with Bane, but I went top.  This confused them, and it might confuse some of you, but here’s why.

First, what exactly is a Bane/Earthshaker lane going to accomplish?  You might be able to set up a kill, but in general Bane’s laning is built either around using  Brain Sap for burst damage (or maybe attrition if you can get a quick Soul Ring) or using Nightmare and/or Enfeeble to set up kills for other heroes.  Earthshaker on the other hand is entirely build around using Fissure to lock people into bad conditions and set up kills.  I’m not going to say that you can’t make the lane work, but there’s no real synergy and a big risk that Earthshaker is just going to be standing around an an ineffective bodyguard.

The most universal advice for getting better at Dota is always be accomplishing something.  Time is the most valuable resource, and you can’t afford to spend the first 5 minutes hanging around an area doing nothing.  People like to send 2 people to the hard lane as insurance, but when it’s a weak lane that can’t accomplish anything you’re paying way too much for that insurance.

In our case, Bane was relatively safe.  The only real CC threat was Chaos Knight, who wouldn’t have a long stun for several levels.  Bane could use Nightmare and Brain Sap to keep himself relatively safe.  He’s probably not going to compete in CS, but who cares?  We have strong reason to believe that Bane’s CS level isn’t really that important to the teams success.  He doesn’t win through CS.  In the vast majority of cases he can’t win through CS.  Every hero has a window they hit, and all Bane needs is a quick 6 so he can get some Fiend’s Grip kills.  Having Earthshaker down there doesn’t do much to help him hit 6 and actually saps experience.  Yeah, suicide lane Bane is a risk, but big deal.  You can’t remove risk from a team composition.  You should mitigate it with smart picks, but good lane selection should be driven by maximizing your advantages, not minimizing your risk.

So how does sending Earthshaker to our safe lane maximize our advantages?

1. Clockwerk is an all-in kinda laner.  If a fight breaks out, Fissure can do a lot to control where we allow our opponents to move in such a way that allows Clockwork to land a good Battery Assault or Power Cogs.

2. Safe-lane Earthshaker can use pulls.  This puts us in a scenario where all 5 people are actively earning experience, which is a lot better than sticking Earthshaker bot as an ineffective bodyguard.

3. We know that the enemy team has 4 carries by virtue of the random gods.  Most likely this means one in each lane and one in the jungle.  Using a combination of aggression and pulls we can strangle the farm on whichever carry gets sent to the Radiant hardlane.  This should allow us to make them underfarmed and underleveled and give us a big advantage in the midgame while they’re struggling to catch up to relevance.

(Not as much of an advantage, but putting Earthshaker in our safe lane also gives a measure of security to our jungler, while also giving our jungler stronger opportunities to gank.  It wasn’t especially relevant since Enigma didn’t do anything but free-farm, but it’s worth mentioning.  These also cover a lot of the reasons why I dislike 2hard/1easy + Jungle comps as I mentioned in The Dota 2 Metagame)

As it turns out, their suicide solo is Spectre.  We zone her out, but Clockwerk pushes the lane anyway.  I respond by stacking with a heavy emphasis on denying.  Enigma comes over and starts converting my pull creeps.  The strategy still does a pretty good job delaying Spectre’s relevance significantly, and given the exact same scenario again I would absolutely endorse sending Earthshaker to the safe lane.

What I might do differently is put Clinkz in the safe lane, Bane in the mid, and Clockwerk in the hard lane.  Clinkz is our farm priority, so giving him freefarm with little to no chance of denies is our strongest early option.  Bane is our exp priority, so putting him mid allows us to guarantee a quick 6/7 into roaming.  Clockwerk is the least exp dependent of the three, and not significantly more farm dependent than Bane.  He also has the best tools for surviving a suicide solo while still getting Rocket farm.

Now let’s move onto part two where we look at what happened past laning.

The enemy team unfortunately did the smart thing and pushed our bot lane hard.  They took the tower, but Bane did get the levels he needed.  Clinkz was mid against Broodmother and it went ok but not great.  Overall we came out pretty well, but they ended up getting our two most important towers fairly early.  We responded by forcing a series of teamfights at the last tower which we won pretty resoundingly.  Both teams did a reasonably good job trying to play up to their strengths, and it’s actually kinda interesting to note just how dissimilar those strengths were.

They had the late game and they had the push.  Specifically, we couldn’t afford the resources mid to really shut down Broodmother, so our outer towers were really difficult to protect.  On the other hand, because they had so much late game they had trouble finding the farm necessary for all their carries and so their early fight was actually pretty mediocre.

We, on the other hand, had virtually no late game.  We had some amazing teamfight ult potential, and we generally did a pretty good job leveraging it.  Let me tell you that I’ve never had an easier time landing Chaos Dunks than that game.  We massively cleaned up early teamfights, and we took towers after we won them.  But the problem was that once we took the early towers we had no impetus to do anything, and we stalled.  When you have a team like ours, you cannot afford to stall.  Every single characters’ relevance fades off in the late game vs a lineup that only gets stronger the longer the game goes on.  Given their comp and our comp, we’re in a situation where if we do not take (multiple) barracks before late game rolls around we are probably going to lose.

This brings us to Clinkz.  Due to the random gods our gameplan revolved entirely around Clinkz, and he quite frankly did not deliver.  This isn’t to dump on the guy, but Clinkz is possibly the most misunderstood character in the game, and he’s the perfect example of big general principle number 2: if you fail to accomplish something with your character’s strengths, you’re not really doing much better than a feeder.

That may come across as controversial, but it’s 100% true.  Every character has a window a relevance, and if you don’t take advantage of yours you are dead weight.  Some characters have an easier to understand window than others.  A carry like Spectre just gets better and better the longer the game goes and the more farm you feed into her.  Batrider is always going to remain relevant because of the huge disable and displacement on Lasso.  But some characters have a much narrower window in exchange for an earlier or larger power spike, and Clinkz is the crown prince of those characters.

Think about it.  A hero with a strong windwalk, long range, really good early orb, and damage bonuses all over his kit.  This guy should be total pubstomp material, and yet he has a 42% win rate.  That’s a 42% win rate on a mechanically braindead character who also happens to be competitively viable, if a bit of a niche pick.  Unsurprisingly, everything I’ve looked at suggests that Clinkz sees one of the largest win rate increases as player skill level increases, so better CS probably plays a roll, but there’s something else going on.

Clinkz has what I feel is the strictest window of relevance in the game.  He has massive early damage increases, but you need the CS to get a quick mana item in order to use them regularly.  The value of these damage increases are fairly static, which means their value diminishes as HP pools increase with game duration.  Clinkz also will never become tanky.  Evasive maybe, but never tanky.  As the rest of the heroes in the game gear up and can force teamfights, this pulls Clinkz into a phase of the game that he does not want to be in, which means to play Clinkz well you have to essentially cripple the other team before they can reach that point.

So what did our Clinkz do wrong?  He played relatively passively and stuck with the team.  Despite our urgings to the contrary he would just farm passively after we won teamfights instead of just windwalking straight to the enemy base and taking a tower or barracks.  He never split off to scout and punish solo farmers.  Our early play gave him a massive advantage that he could have used to finish the game but instead he squandered it.

This is why I almost universally disagree with the assessment that new players should pick evasive heroes.  Take Weaver.  He’s not a carry who happens to have an escape.  He’s a carry that’s very carefully balanced around the idea of having an escape.  If you’re not constantly putting yourself into suicidal positions, getting the kill and Time Lapse/Shukuchi’ing out of those suicidal positions you’re quite frankly not playing the character very well.  Yeah, someone like Luna doesn’t have an escape, but the flip side of that is that she doesn’t need balls-to-the-wall aggression to be an asset to her team.  Just get decent farm, use your Eclipse at remotely appropriate times, and then glaive away during teamfights.  Yes, there’s more to being a great Luna, but being a decent Luna is far more straightforward than being a decent Weaver or Clinkz.

At a more general level, what we’re running into here is that player’s consistently invest way too much in insurance.  For example, “I don’t want to die so I’m going to build a Vanguard.”  If you’re just buying it for insurance, Vanguard is not a cost-effective item.  This is actually a really good thing for the balance of the game; but as a player trying to get better at the game, you have to recognize that the biggest first step to winning is constantly making decisions that allow you to constantly be the aggressor.  If buying a Vanguard is part of your plan to live through otherwise suicidal dives and wrack up a big kill advantage for your teams, then yeah, go for it.  But in that case you’re buying the Vanguard so you can be aggressive.  You’re not sitting on the Vanguard just because it might help you live through a bad time fight, and then limp back to your base while the rest of your team is dead and they’re taking your barracks.  You might be buying insurance for your K/D/A, but you’re not doing anything that’s helping your team win the actual game.

6.77c Patch Notes

Available here.  Overall unsurprising.  Let’s look at them in more detail.

Starting with the two item changes

Tranquil Boots

• Tranquil Boots disabled speed reduced from 50 to 25
• Tranquil Boots heal increased from 170 to 250
• Tranquil Boots cooldown increased from 40 to 60
• Tranquil Boots heal duration increased from 10 to 20

I think everyone had a suspicion that Tranqs were headed for a nerf of some kind.  The actual nerf is kinda weird, so it’s hard to predict with any precision how things will fall out.  The heal increase and cooldown change mostly offset each other.  In terms of healing per minute it’s an almost unnoticeable nerf and in terms of healing per mana an almost unnoticeable buff.  The real goal to the bottom three changes was to make you commit more to each individual heal while keeping everything else about the item mostly constant.  Now there’s a larger window for your heal to be interrupted and there’s more potential healing lost if you do get interrupted.  It’s a pretty reasonable change.

I feel the disabled speed is the far bigger nerf and maybe too harsh of a nerf.  Being knocked below brown boot speed when disabled is a huge disadvantage, and it might if anything make Tranquils exclusively an option for safe farming which if anything was the aspect of the boots that needed to be nerfed more.

Like a lot of other things in this patch, it’s an indirect nerf to Phantom Lancer.  Luna might be hit a bit, but I wonder if the biggest losers will be big tanky characters who liked keeping Tranqs for a long time like Batrider, Axe, and Centaur Warrunner.

This will also be an extremely hard thing to test for.  Items in general are difficult to test right now because you need a really large sample and can’t control when they’re bought, but Tranqs being commonly disassembled makes them even worse to test than other items.

 

Heaven’s Halberd

• Heaven’s Halberd’s Disarm is no longer dispelled by Manta

Good change.  Makes Heaven’s Halberd a much better Agility hero counter, particularly ranged carries like Luna, Gyro, and Drow.  But there’s still always BKB.  Also difficult to test for.

 

Now for the hero changes.

 

Keeper of the Light

• Illuminate manacost increased from 150 to 150/160/170/180
• Illuminate AOE reduced from 400 to 350

A bit of a light hit, but targeting his pony spam is way better than the ult nerfs people were suggesting.  Maybe this is a test to see whether any further changes are necessary in 6.78.  I don’t think this will push him out of competitive viability, and I honestly expect only a minor hit to his public win rates.  I’ll likely test him a bit with a pre and post patch sample, but of the heroes changed I think he’s the lowest priority for what I intend to look at.

 

Nyx Assassin

• Spiked Carapace duration reduced from 2.75 to 2.25
• Spiked Carapace cooldown increased from 14 to 23/20/17/14

Nyx has overall been one of the most successful heroes of both pub and competitive in 6.77, and his recently reworked Spiked Carapace was often considered way too powerful.  Giving it a scaling cooldown makes a lot of sense to cut down on the effectiveness of his early game skirmishing.  Like KotL, this might be a trial nerf that could be followed with more in 6.78 depending on how it pans out.  One thing that could be interesting is to do a pre and post patch Skill Build Analysis to see how players react to Carapace being less of a one point wonder.

 

Drow Ranger

• Marksmanship bonus is removed when there are nearby enemy heroes (instead of just being halved)
• Reduced Marksmanship enemy detection aoe reduced from 400 to 375

Continuing to whittle away at 6.76’s monstrosity.  It’s not surprising, it’s not unwarranted, and it targets the skill I would expect.  The fact that it’s accentuating the counterable element of Marksmanship is a little strange.  It fits the design MO, but it risks eventually hitting her too hard to be an eventual competitive option while still allowing her to be the end-all be-all of low level pubstomping.

 

Treant Protector

• Living Armor damage reduction is now done for all instances of damage that reduce its charges

And of course there’s the complementary Treant Protector buff.  The target here is to cut down on worst case scenarios for Living Armor and make it have a more steady effectiveness for all forms of incoming damage.  It’s a good change, and it being the only change suggests two things for me.  First, Icefrog appears to be happier with what Treant is approaching than he was a couple patches ago.  Two, Living Armor is intended to be the star of his kit.  On the last point I completely agree, but I think like Wisp he’ll never see much overall success in low level play because, like Relocate, Living Armor is insanely difficult for low level players to use effectively.  I still have a feeling that we’re eventually going to get to a point where Living Armor is secret OP, and someone’s going to pull it out and surprise a lot of people.  Like Nyx, Treant might be an interesting case for a Skill Build Analysis, but I think the more interesting thing will be to look at the rate of change in his win rates at different brackets.  This will probably be something I delay to give players a bit of time to adapt to.

 

The biggest element that will complicate testing is not really knowing how long it’ll be before 6.78 proper.  It takes a certain amount of real time to generate a decent sample, and if this is another short patch period there might not be enough days to get a sufficient sample.  I’ll just have to play it by ear.

Which Heroes Belong in the New Player Pool?

A few test server leaks are out of the upcoming New Player matchmaking mode.  This mode appears to be an AP variant designed for new players that limits the available hero choices to newbie friendly characters.  The hero composition for this mode is still tentative, but I thought it’d be fun to discuss which heroes belong, which heroes do not, and precisely what the selection criteria should even be.

Let’s start out with the current list:

Strength: Earthshaker, Sven, Dragon Knight, Axe, Tidehunter, Skeleton King

Agility: Drow Ranger, Juggernaut, Vengeful Spirit, Sniper, Clinkz, Viper

Intelligence: Zeus, Lina, Lich, Lion, Witch Doctor, Necrolyte, Warlock, Death Prophet

Actually, a surprisingly good list.  I have 3 big objections to their inclusions.

Earthshaker is a risky addition.  He doesn’t lane well.  His ult is fairly hard to use.  Fissure is somewhat difficult to use, particularly when it comes to managing your mana constraints.  There are some new players that I would recommend Earthshaker to, people new to Dota specifically but they get the general idea and want a fun way to be aggressive.  For the absolute new player he might be too challenging to use effectively.

I don’t like Witch Doctor’s inclusion.  The heart of his kit involves using Cask and Maledict effectively, and both skills actually require quite a bit of game sense to use effectively.  His ult also involves some tricky positioning demands, and altogether he’s just far less forgiving than what you’d like from a newbie support.

My biggest objection though is Clinkz.  In every sample I’ve looked at, Clinkz has had low 40% win rates that go up dramatically in the High and Very High brackets.  As far as I’m concerned he’s the most timing sensitive farm carry in the game.  What I mean by that is that he needs a farm advantage to be useful, but unlike most hard carries that farm advantage has an expiration date.  Most low level players simply cannot last hit well enough to take advantage of this window.  What’s worse is that if a low level player does manage to farm a quick Orchid, this is also the bracket least capable of knowing what to do in response.  He simply does not belong in the New Player Pool.

Now, are there any heroes that weren’t included but deserve to be?

Starting with Strength, I think you can make cases for Slardar and Centaur Warrunner.  Both pretty simple.  Not farm dependent.  Encouraging new players to learn how to initiate is a good thing and they’re pretty good candidates.  Centaur Warrunner might not be appropriate if he’s due for further balance changes in upcoming patches.  Sand King might be ok as well.

Spiritbreaker and Nightstalker might be alright to add in a few gankers.  On the other hand, if you’re selecting heroes that aren’t too overwhelming for new players to play against, these might be bad additions.  On the other other hand, if overwhelming to play against is a selection criteria should you really be including Drow and Viper?

Chaos Knight might be fine.  Lifestealer too, but Lifestealer might be worth not including if you don’t want to add any heroes capable of semi-legitimate jungling.

For Agility, Ursa should be treated similar to Lifestealer.

Riki and Bounty Hunter might be ok, but they might also have been left out specifically to exclude stealth heroes from the pool.  But then why include Clinkz, the least appropriate of them all?

Luna is a perfectly appropriate addition.  Venomancer as well.

Bloodseeker might be ok, but he might have been skipped specifically to avoid adding Rupture to the pool.

I wouldn’t outright object to adding Spectre and Slark, but I wouldn’t really push for their addition either.

Moving on to Intelligence, I feel the only possible additions are Crystal Maiden, Shadow Shaman, Jakiro, Ogre Magi, and Keeper of the Light.  Crystal Maiden and KotL in particular are much easier to contribute with than Lina and Lion in my opinion.

Intelligence is the trickiest category.  You want to include it because new players simply don’t pick enough intelligence heroes.  But on average I’d say Intelligence heroes are harder to use than Strength and Agility heroes.  That’s why I like the additions of Necrolyte, Death Prophet, and Zeus.  Those three heroes have consistently maintained a high matchmaking win rate despite seeing very little professional play, and every test I’ve done suggest that their effectiveness is highest in the Normal bracket.  If anything, I’d say they are the three easiest intelligence semi-carries for new players to win with.

People also complain that not enough heroes with escapes are included, but I honestly encourage that.  Dota is balanced in such a way that escape-based heroes typically require higher standards of play  to win with.  They often have weak early games or weaker scaling than similar heroes that lack escapes.  To really leverage an escape hero you can’t treat their escapes as a free insurance policy; you actually have to push the boundaries of what you can get away with before escaping.  New players in general are extremely incapable of this, and encouraging them to rely on escapes to survive is a really bad habit to develop if you actually want to get better at the game.

Anyway, that’s my take on which heroes are newbie friendly.  Feel free to agree or disagree in the comments or wherever else.

Odds and Ends

Updated the Table of Contents page and added a new Sample Methodology page so that new visitors to the site could easily find a description of what I do without it having to be tacked on to each article.

There will probably be no new testing update next week.  I’ve decided I want to expand my general sample for the upcoming tests, and my target is March 20th.  Maybe next week will be devoted to some theory ramblings in preparation for some of the future tests.

A lot of questions have come up over this chart

Yes, that High 15-20 point is total BS.  I took for granted that this would be obvious.  I probably shouldn’t have.

Personally, I don’t like looking at the data this way, but I know most people are more visual than I am so I thought including it would be a good way of describing the general trend in a more visual context.  The problem you’re going to run into when displaying data like this is that you’re creating thinner sample slices, which can dramatically up your error estimates.

In the Radiant vs Dire table, we can see that the sample size for High 15-30 is 1763.  This isn’t huge, but it’s workable.  More importantly, the trend that High 15-30 displayed was repeated in Normal 15-30, Very High 15-30, and 15-30 in all brackets in the 6.74 sample.  Given all this we can be reasonably confident in a ~56% Radiant win rate at that time period.

But if we’re slicing down to 15-20, well, we’re looking at 1/3 the area, so 600 is the max expectation for sample size, and 600 isn’t very good.  What’s worse is that 600 would be an extremely generous estimate.  Why?  Let’s visit the wayback machine for that.

Match duration appears to follow a normal distribution, and 15-20 minutes is on the extreme edge of that distribution.  Without even looking at the sample numbers, we can expect samples around 40 minutes to be the largest, and therefore the most reliable.  Samples under 20 minutes or over 60 minutes will be tiny and therefore prone to extreme error.

And as it turns out, I did actually look at the sample sizes for the chart yesterday, and the entire 15-20 bracket was composed of only 200 games at every skill level.  In the interest of clarity I should have probably just started the chart at 20 minutes, but oh well, mea culpa.

I do have one more thing to add about Radiant vs Dire.  A lot of people have responded with their theory on how to explain it.  That’s great.  I encourage this.  But just keep in mind that your theory, no matter how great it sounds, is still unproven.  “Explanations exist; they have existed for all time; there is always a well-known solution to every human problem — neat, plausible, and wrong.”

On a whim I watched Gabe Newell’s acceptance speech at Games Fellowship 2013.  Everyone else is ooohing and aaaahing at 3.5 terabytes per second or whatever.  Meanwhile, my reaction was an angry, “No!  It was T.S. Eliot!”  Props anyway to mentioning Super Mario 64 as one of his three favorite games.  I feel it doesn’t get enough love for being the first of Nintendo’s only 3 good 3-D conversions.

Finally, DotaMetrics hit 100k views earlier this week, which I guess is some kind of a milestone.  Thanks for all the support, and as a reward, have some sexy, wrestling Invoker.

Radiant vs Dire Win Rates

Today we’re looking at the Radiant vs Dire win rate in relation to match duration.  It’s speculated that the Dire has an advantage the longer the game goes due to their positional advantage with regards to Roshan, and the stats appear to support this conclusion.  Overall Radiant has a small advantage over Dire (I estimate the overall win rate at 50.83%), but this advantage clearly dissipates the longer the game goes on.  What’s really interesting is that the Radiant advantage seems huge in games that last less than 30 minutes, particularly those in the Very High bracket.

Here’s another way of looking at what’s going on.

If we accept Roshan as the explanation for Dire’s late game success, what drives the early Radiant advantage?  Some explanations that have been thrown around include:

1. Radiant side is easier to play due to the way the map is oriented and how we visually process information.
2. Radiant has better pulling opportunities.
3. Radiant has a safer jungle.
4. Radiant has a safer midlane.
5. Radiant’s off-lane tower is the less valuable than Dire’s off-lane tower.
6. Valve knows about the late-game Dire advantage and gives them a slight handicap during matchmaking.

There’s no clear answer, but I will say that I find 1 and 2 of dubious value.  If 1 were the case I would expect the effect to shrink in Very High where players have simply played more games and have had more time to acclimate.  Instead, Very High is the bracket where the Radiant advantage is the strongest.  As for pull camps, if this were the sole driving factor I would expect the advantage in Normal to be negligible.  I’m almost 100% positive the ability to pull in Radiant’s middle like has had virtually no effect on Normal matchmaking.

This also raises the possibility that perhaps Radiant vs Dire have asymmetrically preferred line-ups.  Radiant could possibly be more effective with a lineup designed for early game dominance and pushing, whereas Dire would be more effective with a lineup designed for counter-push and stalling, while building around some kind of a late game carry.  This is still speculative so don’t yell at your teammates for not going along when you slam down insta-lock Dire Medusa, but it’s a possibility worth considering.

One last curious piece to the puzzle is  that this same effect existed in 6.74 and was actually stronger back then.

This is curious because the talk of 6.75 was speculation that the Aegis duration change would weaken Dire’s Roshan advantage.  I do not have 6.75 or 6.76 data to look at, but at least in 6.77 Dire appears to be stronger than it was before the change.

Popularity of Game Mode by Bracket

Minor update, but I went through the sample and tallied up the mode usage.  Not terribly exciting.  Single Draft falls in popularity in the upper brackets.  Random Draft gains popularity.  All Random is rather healthy at all skill levels.  Least Played not so much.  Captain’s Mode is quite a bit lower than I expected.  Perhaps TMM’s release early in the sample collection drew away a lot of the CM playerbase?  Surely there should still be some Normal results, but unless it’s a quirk of the sample collection in some way I’m just not seeing any.