Random Ability Draft: Hero Win Rates

June 11, 2014

As a bit of a curiosity and side project, I put together the hero win rates in Random Ability Draft from my last 6.80 sample back in January.  It may be a bit outdated, but the base hero values didn’t change drastically in 6.81, and the stats come from the period before Divided We Stand’s reign of terror.  While I don’t have much interest in the mode myself, it’s an interesting look into how the basic stat values contribute to a hero’s viability independent of their abilities.

The special caveats here:

  • My RAD match collection is a byproduct of my actual match samples.  As a result it’s fairly small at just under 20,000 games total.  Mitigating this, because the hero selection is randomized, all hero samples are around 2,200 games.
  • The matches are separated into Normal, High, and Very High.  I do not know if they actually correspond to anything.  My best guess is that RAD unranked MMR but then becomes separate.  In any case, I averaged the Win % across all three brackets in an attempt to mitigate sample size issues.
  • I tried to revert the 6.81 changes in my stat charts, but I might have missed some.
  • There might be other errors.  This was new, hectic, and I’m not extremely invested in it.

Oh, and if you’re prone to getting angry about being assigned “bad” heroes in the mode, you might want to look away.


And that’s well and good, but the more interesting question is what makes a hero a good or bad platform in RAD?  Silencer far and away at the top is a relatively easy outlier due to his intelligence steal, but beyond that it gets more complicated.  One intriguing fact you might pick out by the hero shading is that Intelligence heroes tend to do the best overall in RAD.  In terms of average win rates, Intelligence heroes come in at 51.33% with Strength at 49.49% and Agility at 49.01%.

But that’s not good enough because we want to break things down by individual stats.  To approach this, I imported the hero attributes table from dota2wiki, and found the average win rate for the top, middle, and bottom third percentile of each stat group.  Or some Excel facsimile of that.  It’s at least a relatively consistent division for above and below average.  I wouldn’t rely on it for precision, but it gives us a general idea of how influential each stat appears to be on the overall win rate.  And here’s the results:


So for stats, Intelligence actually appears to be the most valuable stat in both base and scaling, but Strength Scaling is close.  This likely explains a good portion of why Intelligence heroes do best overall in the mode.  Having good overall base stats and stat growth are arguably the two strongest predictors for a strong RAD hero.  Movespeed and Base Armor are the next two strongest predictors, and ranged heroes as a whole tend to outperform melee heroes.  What’s somewhat surprising is that short cast points provided no noticeable advantage whatsoever.

But don’t take my word for it.  I made a (rather huge) graphic that includes a percentile ranking for each stat for every hero.  For example, the movespeed for all heroes fall between a minimum of 280 and max of 330.  Heroes with 280 are treated as 0%, 330 is treated as 100%, and 305 is treated as 50%.  Extreme stat outliers are kept separate.  For example, I treat the range in average level 1 avergae autoattack damage is 44 to 69.  But Treant Protector’s damage is 85, which gets treated as 174% so he doesn’t warp the listing for everyone else.  It’s also available in spreadsheet format so you can download your own copy for sorting.



The Virtue of Not Knowing

May 20, 2014

Came back to my computer at 5 AM Sunday morning to an absolute deluge of messages, comments, and e-mails.  Redditor ThirstyCows had discovered a bug in 6.81 that was causing Pudge’s Rot to do a incorrect amount of damage.  As of yesterday, the bug was fixed and the results were immediate:

So with that out of the way, I’d like to address the idea that fact that some people appear to have missed the point of the post where I examined this Pudge shift days before the bug was discoveredThe suggestion is that I spent a lot of words trying to figure out the ‘why’ behind Pudge’s sudden win surge and then had to admit that I didn’t know, when the real goal of that article was to prove that we didn’t know.

The fact that I didn’t know the answer was never in question.  I was upfront about this a mere 24 hours after the patch dropped in my 6.81 First Look.  Pudge had one of the largest win rate increases.  I had no clue why it was happening, but I was sure that it was a big deal.

What I then ran into were comment sections full of explanations that hinged on this-or-that meta shift.  I decided to go through a couple of these and look at them in detail, and I found them overall to be wanting.  At a first glance they were neat and plausible, but, as we now know, they were wrong.

And far be it from me to criticize someone for attempting to come up with an explanation about Dota, but the problem is that too often we use “it’s the meta” like it’s this omniscient being that directs the course of Dota in a way far beyond the comprehension of us mere mortals.  Being able to explain any change with “the meta” helps us feel like we know everything we need to know about this game that we know far less about than we pretend to.  This might actually be a necessary coping strategy sometimes for life in general, it’s pretty terrible in an environment where you’re trying to gain a competitive edge.

Dota does change constantly but in very much knowable ways, and to be able to keep up you need to be willing to throw away things that you thought you knew.  Like Wraith King.  A year ago people would constantly tell you that he’s a terrible competitive hero that will never see the light of day.  Now he’s the 11th most picked hero in 6.81.  With enough power boosts any kit can be competitive, and the wonderful thing about pub play is that it’s so big and random that any large win rate change almost has to correspond to a shift in raw power.  There are, of course, exceptions, but they’re almost never sustainable.  When a hero, any hero, sees big changes, it’s time to reevaluate their viability, and pub win rates give us a useful (if incomplete) idea of which changes are big.


And to close on a completely different note, with the TI qualifiers going on I have a couple of short articles out about the biggest hero stories so far:

TI4 AM Quals: A Wild Mirana Appears (Everywhere)

TI4 SEA: Into Forest Shade

Encyclopedia Pudge and the Case of the Strangely Surging Win Rate

May 13, 2014

As you might recall, last week I published a thing on the hero win rate shifts of 6.81.  At the time, I said that the biggest mystery of the patch is Pudge’s win rate increase.  Well, that mystery has not abated over the past week, so I want to take some time to examine the potential suspects and see how their alibis stack up.

The Scene of the Crime, as it were

The Scene of the Crime, as it were

First the facts.  At the time of this draft, Pudge has exhibited the largest win rate boost on Dotabuff in the 6.81 patch period at 3.94%.  This is odd because the usual culprit for shifts this large are official balance changes, and Pudge received none.  He also, as far as I can tell, received no direct or indirect buffs bugfixes that would have driven an improvement of this magnitude, and I can find no reported bugs in 6.81 that Pudge would benefit from.  Of course, just because I cannot find them doesn’t mean that they’re not there, but at this point it’s safe to say that on this front I have no leads.

For now, let’s assume that Pudge’s recent success has nothing to do with a mechanical power boost and examine some alternative explanations.

Theory 1: Pudge is being played by better players in 6.81

The typical explanation goes like this.  There are a lot of bad Pudge players.  These Pudge players get distracted by new shiny things in the patch, so they play heroes other than Pudge.  Pudge’s win rate goes up accordingly.

There’s some support for the idea that popularity trends can influence win rates.  The most clear and consistent example is how every new hero introduced to the game has had huge day 1 usage rates that quickly fall off, and as they fall off the hero’s win rate always creeps upwards.  Some of this win rate improvement, especially for newer and more complicated heroes like Earth and Ember Spirit, is players just learning how the hero works.  Still, even with more basic heroes there tends to be a tourist effect.  It’s not even much of a surprise; people who pick a hero because they like them tend to outperform people picking the same hero out of a sense of novelty.

The problem then is that there’s no real evidence that this effect makes much of a splash in the opposite direction, and even if there were, there’s no reason to believe it would effect Pudge dramatically more than every other hero in the game.  And on top of that, Pudge’s use rate didn’t even decline post-patch.  It actually hit a month high 39.24% last Thursday up from his usual 35-36%.

Verdict: Implausible

Theory 2: Pudge is taking advantage of trendy strats like Aghs-rush Lich and KotL

This theory we can actually test.  Using the Dotabuff Pudge matchup page we can find out how often he plays against different heroes during different patch periods and his win percentage in each head-to-head matchup.  Has Pudge fed off of a sudden surge of sloppy play?


In terms of usage rates, 6.81 did not shake things up dramatically.  More important to the theory is that the heroes Pudge is facing more often in 6.81 aren’t exactly favorable matchups.  Only Invoker, Keeper, and Phantom Assassin are better than average matchups both pre- and post-patch.  Not by much in the case of Invoker and Keeper who are 42 and 45 respectively out of 106 in 6.81.  Phantom Assassin was Pudge’s 11th best matchup in 6.80, but her6.81 buffs have improved her to 33rd.  On the other end of things Wraith King was tied for Pudge’s worst matchup in 6.80, though things have improved for Pudge such that it’s now 104 out of 106.  Ursa and Lich are also amongst Pudge’s top 10 worst matchups.

It’s also worth pointing out that this group is, if anything, outperforming Pudge’s overall win rate shift of 3.94%, thanks in a large part to buffs to Huskar, Faceless Void, Phantom Assassin, and Silencer.

Verdict: Implausible

Theory 3: Pudge is benefiting from a favorable meta-game shift, and his win rate comes from a handful of advantageous matchups

This explanation gets trotted out to explain any otherwise inexplicable win rate shift, and while it sometimes is valid for competitive play, things are rarely that responsive in pub Dota.  Competitive Dota is a highly self-aware system where teams work to respond to specific matchup threats and any seemingly successful strategy will get integrated at-large as a counterpick option.  By contrast, picks (and lane assignments) in pub Dota are essentially random dice rolls weighted by the innate popularity of the heroes in question.  As a result, the influence of the metagame on pub win rates tends to be rather weak.

To demonstrate this, let’s use a thought experiment.  Invoker is Pudge’s most commonly seen foe in 6.81 Dota, so let’s simplify things and just say he shows up in 1/5 of all Pudge games pre and post patch.  In 6.80 Pudge’s win rate vs. Invoker is the same as his overall win rate at 48%.  In 6.81 his win rate jumps four percentage points to 52%, but this increase is solely because of a favorable development in the Pudge vs Invoker matchup.  As a result, his win rate in the 4/5 of games without Invoker stays at 48%.  What would Pudge’s 6.81 win rate against Invoker have to be to make this possible?

The answer.

So yeah, you’ll basically never see a win rate shift this large determined by a single match-up, at least in pub play.  At a tournament like TI3 you can see a single hero like Outworld Devourer sunk entirely by counterpicks like Razor, but that was an incredibly small environment where teams had a very specific gameplan involving these two heroes.  Pub play is a bit under a million games a day being played players that, on average, either have no concept of a Dota metagame or have an incredibly wrong one.

But perhaps there’s a broader explanation at work here.  A cluster of heroes that are working in concert to lose to Pudge.  Well, if there is, I’m sure not seeing them.

Using the Dotabuff matchup page again, I compiled the win rate shift in every individual matchup against Pudge between the two patch periods.  60 out of 106 of the heroes fell within a +/- of .75 from Pudge’s overall win rate shift of 3.94.  Of the 46 that did not, the vast majority were heroes that had their own win rate shifts thanks to buffs and nerfs they received in the patch, with their Pudge shift being a rough summation of the individual win rate shifts.  In short, Pudge’s newfound win rate appears to be distributed relatively evenly among the entire cast.  I can’t completely rule out that there isn’t something more complex going on, but it’s definitely not a simple question of 1v1 matchups.

Verdict: Very Unlikely

Theory 4: With recommended items changed in the patch, Pudge players are now spending their money more effectively.

This…actually might have some merit.  Maybe.  I can’t actually determine if Pudge’s recommended items did change with the patch, as I can’t find a reliable source on what his recommended items were pre-patch.  What I can say is that there has been some interesting item trends going on with the hero.  Two in particularly actually.


I’m skeptical that the Force Staff -> Blink Dagger is the big driver here.  From what I can tell using (Last Month – 6.81) to get an estimate of late 6.80, this trend towards Blink Dagger has been a gradual once ever since Pudge received the ability to use Blink Dagger in 6.79.  If Blink Dagger was the source of Pudge’s win rate, we would expect his win rate increase to have also been gradual and not suddenly surge on patch day.

The Boots shift is much more likely to be due to recommended items.  My estimate of (Last Month – 6.81) puts Phase Boots at a ~29% usage rate, which fits the profile for a precipitous dropoff in 6.81.  Tranquil Boots is also definitely a part of Pudge’s current core items.  It’s also worth mentioning that Pudge has a lower win rate with Phase Boots despite Tranquils being much cheaper, so there’s a case to be made that reduced Phase Boots usage should actually increase his overall win rate.  The question though is whether this boot choice alone could drive a 4% win rate shift.

Verdict: Plausible, but possibly only a partial explanation

And that’s the case as it stands.  It could be that Pudge stands out as the somewhat surprising biggest winner of the 6.81 recommended item changes, but I’m skeptical that this is the entire story.  As I’ve pointed out, Pudge’s win rate increase in the patch was the largest of any hero, and a boot switch doesn’t seem like it would be enough regardless of the fact that Phase Boots are Pudge’s least successful Boot upgrade.

So we still have an open mystery on our hands.  In the meantime, strongly consider skipping the Phase Boots when trying to get that Pudge compendium win.

One Weird Trick to Raise MMR That Raijin Thunderkeg Doesn’t Want You to Know About

May 7, 2014

In recent news, we have this tale from the PlayDota forums of one man’s harrowing tale of taking a 2900 rated Dota2 account to 5400 in the course of 144 games with a 122-22 (.847) record.  Some might say that this disproves the existence of ‘ELO[sic] Hell.’  But of course for some people it didn’t, and today I want to focus on one of the objections.

Juice, the player behind the experiment, prefers to play mid and carry and did so throughout the experiment.  A common reaction then has been that all this experiment proves is that if you want to raise your MMR, you need to pick ‘high impact’ heroes, preferably mids.  Juice seemed to prefer Ember Spirit, Storm Spirit, and Templar Assassin, though his picks overall were actually quite varied.  In any case, I want to examine the idea that maining these snowballing mid heroes are some universal secret to raising your MMR.

1. Let’s start by looking at the overall win rates for some mid heroes:


Outside of our three exceptions, things don’t look so good for traditional mid heroes, and when people recommend picking a high impact mid, Zeus and Necro aren’t typically the heroes that get brought up.

Of course it’s true that some of these heroes do perform better in higher rated games.  Ember, Storm, Tinker, Invoker, and Templar Assassin are all among the heroes with the greatest win rate improvement in Very High games (roughly >4000 MMR).  But at the same time, none of these shifts are strong enough to get any of these heroes to +50% in that category.  Furthermore, these heroes are only nearing that range of success in the hands of people with the mechanics to make it to >4000 play.  If you can’t make it to 4k using the heroes you’re comfortable with, it’s extremely unlikely that you have the mechanics necessary to make any of these heroes your ticket out of the so-called trench.  At least right now.  If you aspire to learn everything there is to know about one of these heroes, then good, go for it.  Just don’t pretend it’s going to be the fast fix that will carry you to the MMR you really deserve.

But I feel that this simple win rate examination fails to capture the entirety of what’s going on here, so…

2. Let’s step back and ask ourselves what exactly it is that hero win rates measure in the first place.

If you’d like to know, I can tell you that in your universe you move freely in three dimensions that you call space. You move in a straight line in a fourth, which you call time, and stay rooted to one place in a fifth, which is the first fundamental of probability. After that it gets a bit complicated, and there’s all sort of stuff going on in dimensions thirteen to twenty-two that you really wouldn’t want to know about. All you really need to know for the moment is that the universe is a lot more complicated than you might think, even if you start from a position of thinking it’s pretty damn complicated in the first place. I can easily not say words like “damn” if it offends you.

-Mostly Harmless

Think of a hero win rate as a giant array of outcomes condensed into a single number.  Two key dimensions within this array represent the probability that the hero will influence their team into a positive gold/XP state and the probability that a hero can take a positive gold/XP state and convert it into an even more dominant position.  The heroes that we think of as supports tend to produce good outcomes of the first type while the heroes that we think of as carries tend to produce good outcomes of the second type.

Of course it’s more complicated than that.  Lich likely has one of the highest public win rates in a large part because you can stick him into any dumb pub composition and he’ll accomplish something in the laning phase.  Meanwhile, Shadow Demon has one of the worst public win rates because his early impact is extremely dependent on having someone take advantage of Disruption initiation.  On the other end of the spectrum, Anti-Mage has possibly the strongest capability to convert GPM into total dominance, but his mere existence on a team makes it less likely to get to those stats.  Viper has a much weaker gold conversion but has a dramatically higher win rate because he lanes well in general.  Spectre provides a somewhat more puzzling counter-example, as she, like Anti-Mage, is a weak laner, and yet she still has had a consistent >52% win rate for several patch periods.

Anyway, We’ve got this idea of snowball or momentum heroes, so where do they fit in?  We could say that they’re just a lot like carries since carries snowball too, but that feels off.  Items do matter for heroes like Templar Assassin, Storm Spirit, and Outworld Devourer, but not really in a way that’s comparable to Anti-Mage or Spectre.  Anti-Mage and Spectre aren’t dependent on momentum; they’re more like time bombs, with their rate of farm determining how much time you have to disarm them.

What I suggest is that the success of these snowball semi-carries isn’t about reaching some kind of item-based end state like it is with a traditional carry.  Instead, the goal of a snowball semi-carry is to create a gap between their damage progression and the HP levels of the opposing team.  The most straightforward example of this is Outworld Devourer’s ultimate.  Sanity’s Eclipse does damage based on the difference between OD’s intelligence and the target’s.  If OD gets an item and level lead, Eclipse will do huge amounts of damage in a large area of effect.  Conversely, if OD falls far enough behind it won’t do any damage at all.

The central unifying feature to this class of heroes appears to be a specialization in burst damage that tends to be at least as level dependent as it is item dependent.

  • Templar Assassin can use the free damage and armor shred from Refraction and Meld to melt low HP heroes.  Damage items like Desolator and Daedalus can be useful but aren’t strictly necessary and often skipped/delayed in favor of pickups like a Blink Dagger.
  • Storm Spirit thrives on hitting that point where any lone traveler is dead within the duration of Electric Vortex and Orchid.
  • Ember Spirit gets right click damage scaling off of a 5-second AoE nuke with a very long range, but struggles to keep up with the sustained damage output of a true late game carry.

If any of these heroes get ahead, they can consistently blow up out of position targets and put their opponents on the defensive with a 5v4.  If they fall behind, or even just fail to stay ahead, they can find themselves put into more prolonged fights that their kits are not as well suited for.

And that’s fine.  Part of being a hero that depends on momentum is the downside of potentially becoming dead weight if you lose that momentum.  The first choice you make in any Dota game is trying to pick a hero with a strong array of outcomes, but those arrays aren’t fixed.  Wisp’s array in pub play is pretty terrible, but his array in competitive play has been stellar.  Just like Shadow Demon, if you draft a positive environment for the hero, they can drastically outperform their performance in essentially random pub environments.  Snowball mids are just another class of hero where a good environment is essential to success.  Subsequently, their overall pub win rates kinda suck because pub teams in general are tremendously more self-centered than even the most marginal of competitive teams.

 3. So if we accept that environment influences a heroes success, why can’t MMR (and matchmaking in general) itself play a huge role in shaping the expected environment?

Imagine that we have Juice, a ~5500 rated player, queuing into his first match on a 2900 account and playing Templar Assassin.  What do we know about the match conditions?

GPMYou probably don’t recognize this graph from over a year ago, but it shows the distribution of game ending GPM for players in Dota’s three skill brackets back in 6.74.  I’m going to run with the assumption that the shape of these distributions hasn’t changed dramatically since then, partially because it’s likely true and partially because I don’t want to go through the trouble of making an updated version right now.

The blue distribution is Normal games, which roughly correlates to < 3000 MMR.  The red distribution is Very High games, roughly > 4000 MMR.  Green is everything in between.

Thanks to the Dota2 blog we also have an idea of the shape of the overall distribution of MMRs.

5% 1100

10% 1500

25% 2000

50% 2250

75% 2731

90% 3200

95% 3900

99% 4100

With a ~5500 rating, the games Juice plays on his main account would be deep in the tail end of the red distribution.  Based on the chart, we’d expect his games to fall around the 400 GPM per player mark per average.  And what do we find when we take the average GPM of all the players in his last 10 games on his main account starting with 646057416?  An average GPM of 406.41.  We don’t have a fair measure the average GPM of the games at the start of this experiment because Juice’s presence in those games inevitably warps the results, but if we trust that my 6.74 data holds, it’s likely somewhere between 300 and 325.  XPM likely has a similar, if somewhat less pronounced gap.

For a hero based on momentum this is a dream environment.  If your potential as a hero is based on the difference between your stat progression and theirs, then with 100 GPM worth of a handicap that difference will practically never go away.  On top of this, you get a guaranteed 1v1 lane against a player you almost certainly massively outskill, and there’s virtually no risk of ganks since you’re in a bracket where supports, if they even exist, largely have no idea what they are doing and believe that only mids gank.  You give a top .1% player all those advantages and it’s not surprising that the immediate result is this:


So yeah, if you’re a player playing 2500 MMR below your true MMR, snowball mids might indeed be the fastest, most reliable way to maximize your winrate and MMR growth, but for the 99% of players playing at a relatively accurate MMR, there’s absolutely no reason to believe that these heroes are a fast fix to a positive win rate.  Sure, with enough practice a hero like Storm Spirit can get you up there, but the operative word here is ‘practice.’  You’ll have to actually put in the effort to improve as a player, and you’ll be better for it than a player just trying to Flavor-of-the-Month it up the ladder.

4. Epilogue: The Origin of an Idea

Suppose that instead of MMR, matchmaking just threw 10 people randomly together into teams to create games for you.  What would these games look like?  Well to start we need to add ourselves, so let’s assume that we’re a vaguely defined ‘above average’ player with an MMR somewhere between 3500-4500.  The other 9 slots would tend to be distributed pretty similarly to the distribution earlier from the official Dota 2 blog.  This means that on average, 8 of the other 9 people are going to be below 3000 rating, and 5 of 9 are going to be below 2250.  So if you’re a 4000 player in one of these games, most of your games are going to going to be you as a ~4000 rated player, and a bunch of other people 3000 or below.  In other words, the games from your perspective would look an awful lot like the game’s Juice went through on his experiment.

I’m not going to claim to be an expert on the social structure of original Dota, but it’s likely that basic Battle.net custom games were pretty similar to this as the normal distribution has this pesky tendency to show up just about everywhere.  This means that if you were a decent player in the Dota days, then just about every game you played was similar to the ones in Juice’s experiment.  Is it any wonder that ‘play a snowball mid’ is common knowledge?  It would have really been true in almost every game.

For better or worse (hint: better), matchmaking has killed this, but Dota players in their opinions are often the slaves of some defunct patch period.  You can see the same effect when it comes to win rates.  In a completely random system, a 60-80% win rate would be evidence of the equivalent of a 3000-5000 rated player.  With Dota2’s matchmaking, win rates are no longer a direct predictor of player skill.  But you still have this crowd longing for the regular assurance that they’re nominally above average that you would get in a completely random.   And so you get complaints about ‘socialist matchmaking’ forcing 50% win rates.

LoL Side Advantages and What They Suggest About the Radiant Advantage

March 19, 2014

A while back we looked at Radiant vs Dire win rates in 6.74 and 6.77 and found that Radiant had a slight overall advantage that was most pronounced in short games and actually became a disadvantage during long games.  While we don’t have absolute proof, it’s generally accepted that the late game Dire advantage is driven largely by the positioning advantage Dire has for Roshan.  But when it comes to the other side of the dynamic, Radiant’s  advantage during short games, all we have are competing theories with no clear evidence.  Recently in League of Legends, a stat site released some data on the balance between their sides in various modes, and the implications could help explain what’s going on.

For those unfamiliar, the LoL Blue team corresponds to Radiant positionally and Purple corresponds to Dire.  The default LoL map has some asymmetry, but nothing to the extent found in the Dota map.  What the stats show is that on the standard League map, the Blue side has a significant advantage in all modes (Unranked: 56.3%, Ranked: 55.2%, Ranked Teams: 53.5%).

Where things get interesting is that the same advantage exists in ARAM (55.4%) which is a perfectly symmetrical map.  This is the strongest evidence I’ve seen that the Radiant advantage is related to the viewing perspective/camera angle.  The LoL map with the smallest side advantage is Dominion (51.8%), which also happens to be the only map tested without a diagonal tilt, as the map is East/West symmetrical.

It’s still not proof, but I think there’s decent evidence here in support of the camera angle theories.  Dire’s inherent Roshan advantage could actually be rather appropriate.  Roshan’s positioning would likely have the greatest net effect in low level games as low level games have the longest average duration.  These also might be the players least capable of dealing with the camera disadvantage, and therefore most in need of a larger counterbalance.   It’s fairly speculative, but it would be a pretty unique feature if both the Radiant and Dire advantages separately scaled at a similar rate with the overall skill of the players in the game.

[Guest Post]A New and Improved Test for Farm Dependency

March 17, 2014

For a few months now I’ve been in an e-mail correspondence with a reader by the name of Bishop’s Guest on the subject of putting together a better farm dependency test, and this week he provided me with a .pdf writeup on the test and a table of the results.

Given that this has been in the works for several months, the data used is from 6.77, so it’s not the 6.80 Farm Dependency test that people have been asking for.  Think of it as foreshadowing.

6.77 Farm Dependency pdf

Farm Dependency Table in Google Drive

Anyway, I’m going to let the pdf speak for itself.  It explains how the test works, and also comes with visual representations of each hero’s farm dependency, like this comparison between Anti-mage and Faceless Void.


Hero Win Rates by Match Duration: 6.80 Edition

March 10, 2014

The duration test isn’t that heavily requested compared to farm dependency (which is coming), but I needed part of this in preparation for something else.  Turns out there are some interesting trends emerging since we last looked in 6.78.

I should mention in advance that I have changed the definition of Short/Mid/Long divisions.  Previously they were determined by splitting the sample evenly into 3 parts, and the dividers turned out to be 33:20 and 41:46.  To simplify things I’ve changed to using 30:00 and 40:00.  They’re close to the 1/3 split, but it’s much easier to remember that short matches are less than 30 minutes and long matches are longer than 40 minutes.  It also has the added benefit of being directly comparable to how datDota‘s duration filter works for competitive comparisons.

Anyway, let’s start things off with how the 5 new heroes since 6.78 scored:


Preferring the long end of things we have Earth Spirit and Legion Commander, but I would caution against reading too much into these two.

It’s no secret that Earth Spirit players are largely struggling since the 6.80 nerf, and it’s my suspicion that this might be capable of creating or exaggerating a late skew.  Essentially, if a hero has a proclivity for being blown out, then a match that lasts 40 minutes has better than expected odds of not being a blow out.

As for Legion Commander, a lot of players are still very much dedicated to her (not very good) jungle.  If you have a passive, farm oriented jungler and the match ends in under 30 minutes, chances are the contest did not resolve in your favor.  I think there’s some evidence of this being a general jungler trend, with the junglers that escape from it being those that aim for early objective control (Chen, Enigma, Lycan, and Ursa).  If we had a collection of purely laning Legion Commander games, I suspect she would still have a late skew but that the skew would not be nearly as dramatic as this one.

We also have Phoenix with a moderately late skew, which is somewhat surprising to me as I’d expect his minus attack speed oriented strategy to fall off in late game.  Perhaps the scaling on Sun Ray keeps him relevant in late game.  Alternatively, it could just be that there are a lot of bad Phoenix players in this sample, as this was release week Phoenix, and they might be disproportionately likely to lose games early.

On the short side of things we have Terrorblade.  He has a pretty early skew for a carry, but he does fit the pattern of having a lot of free damage directly packed into his kit in Metamorphosis.

Finally we have Ember Spirit who has no discernible skew in either direction.  This isn’t terribly surprising.  He doesn’t have any of the features you’d expect to see in a short skew hero (support orientation, pushing power, free auto-attack damage), but he also doesn’t have a passive start or dominant item-based late game.  Ember Spirit  is likely better off aiding his teammates’ preferred match tempo than trying to set his own.

With the new heroes addressed, I want to move on to the top 15 heroes preferring short games because that side of the list has changed dramatically since 6.78.


Push heroes always skew early, but 6.80 has taken it to new extremes.  The new and improved Lycan leads the pack, but 6 of the top 7 are pushing heroes.  An interesting note is that this push mini-meta has had interesting effects on the push heroes that haven’t recently received changes.  Nature’s Prophet and Leshrac have both seen an increased early skew, while former top 15 entry Luna has dropped down to 42nd.  One explanation for this is that Nature’s Prophet and Leshrac can support any push strat, whereas Luna might be in direct competition with Lycan, Pugna, or Death Prophet as the centerpiece.

I also want to mention the former top 3 short game heroes in 6.78, Treant Protector, Spirit Breaker, and Huskar.  All 3 have seen nerfs since then, and correspondingly their early prowess has diminished.

  • Treant Protector:  16.53% -> 9.46%
  • Huskar: 14.28% -> 7.37%
  • Spirit Breaker: 12.82% -> 2.22%

This appears to be the inverse of the earlier Earth Spirit theory.  Being a dominant hero in a patch period seems to exaggerate the heroes short skew.

On that note I’m going to close things with the complete chart.  This time I’ve put the Very High and Normal charts side-by-side.  One interesting thing about the Normal bracket (est. < 3200 MMR) is that it exhibits a much stronger correlation between a hero’s overall win rate and their short skew despite having longer games on average.  So for those of you concerned with digging yourself out of the trench, don’t agonize so much over getting mid.  The simple trick is just learning how to make an impact early, and you can reliably do that from a variety of positions.