Latest / Start the game already! / [ENG] How to beat Hera

Transcript

[ENG] How to beat Hera · Start the game already!

0:17
Welcome back to another episode of the Start the Game Already
0:20
podcast. I'm your host, Chris, and today
0:23
I'm bringing you a rather different kind of episode
0:26
because I'm going to tell you about my little adventure in
0:29
search of the answer to the one burning question.
0:32
How do you beat Hera? Hera's win streak in STA
0:36
tournaments is nothing short of legendary, and there's no sign
0:40
of it ending anytime soon. The next opportunity to knock
0:44
him off his throne will be the Ling Yuan Cup in China, which
0:48
kicks off in just a few days. Hopefully, I'm publishing this
0:52
episode right while Hera is on the plane, so he won't hear it
0:56
in time and his rivals might just get a tiny advantage.
1:00
Hera has won 12 incredible STI titles in a row, establishing
1:05
himself as the most successful and versatile player of our
1:08
time. The best players in the world
1:11
often seem at loss when asked how to bring Hera to his knees,
1:15
and you keep hearing them say that he's unbeatable within the
1:19
current meta, that it would take some unconventional strategy to
1:24
stand a chance. Only one player seems
1:27
unimpressed by all of this, Leary.
1:29
Probably the greatest talent Age of Empires 2 has ever seen.
1:33
No one gives Hera as much trouble as Leary does,
1:37
especially in it in competitive settings, and even Hera himself
1:40
admits that Leary is his equal in nearly every way except for
1:46
one thing, and that's familiarity with all
1:48
civilizations. You see, Leary hardly ever plays
1:52
ranked matches and instead prepares individually for
1:55
tournaments, whereas Hera seems to spend practically all day
1:59
grinding game after game. So my first thought was to take
2:04
a closer look at Leary to see how he exploits Hera's
2:08
weaknesses. Since there are hardly any
2:11
ranked games to analyse, I turned to their past tournament
2:15
encounters, which as we all know, are always incredibly
2:19
close. But usually go in Harrah's
2:21
favor. One such match was the final
2:24
game of NAC 4, where Hera defeated Leary with the Flemish
2:27
Revolution. Again, I actually dedicated
2:30
whole podcast episode 2. Interestingly enough, that match
2:34
also marked the beginning of Harrah's ongoing STO win streak.
2:37
So if you want to recall that game, just get across that
2:41
episode. I extensively gathered data from
2:45
all their encounters, and I think it's fair to call these
2:47
two the best players in the world right now.
2:50
And well, I was planning to tell you all about Leary's complex
2:54
recipe for beating Hera, but it turns out there isn't one.
2:59
At least not one that I can derive from the data that I had
3:03
gathered. The short and honestly quite
3:06
impressive result of my analysis is that Leary and Hera are
3:09
virtually identical in most respects.
3:12
If anything, Hera plays slightly more defensively than Leary does
3:17
otherwise. Both have more or less perfected
3:19
the game, and in many cases half of the results seems to be
3:22
determined already during the map and sift draft that ties
3:26
into what ultimately decides their matches.
3:29
It's not micro or macro that determines whether Hara O'Leary
3:32
wins the game. It's often something like a
3:35
random hill being closer to 1's player's base, or who realizes
3:40
this first, or who happens to have the better civilization to
3:44
take advantage of that random aspect of the map.
3:46
And when it comes to who wins an entire set, the decisive factor
3:51
tends to be a strategic choice. In the best of nine.
3:55
Those random map aspects usually even out across the games, so
3:59
it's the strategic edge that counts.
4:01
And in that regard, Hara just seems to be one step ahead of
4:05
Leary, simply because he has a few more strategies up his
4:08
sleeve. So I think Hera is probably
4:12
right when he says that Gary just lacks familiarity with the
4:15
now 50 civilizations we have in the game, at least compared to
4:19
him, because when it comes to pure skill, they are evenly
4:23
matched, especially in today's tournaments which combine such a
4:27
wide variety of map types. After this analysis, I was left
4:32
a bit puzzled and had almost given up on the episode because
4:35
if you want to sum up the findings briefly, it would be
4:39
this Hera has no weakness, and to beat him you simply have to
4:44
be better than he is. Leary, at least, doesn't
4:47
systematically exploit any particular weakness of Hera's.
4:51
Instead, he focuses entirely on his own strengths.
4:54
That hasn't made him ultimately successful, but it has made him
4:58
more successful than anyone else in trying to wrest an STA
5:02
victory from Hera. But don't worry, the podcast
5:05
doesn't end here. In fact, this is where it really
5:08
begins, because I've gathered an entirely new set of data thanks
5:13
to a situation that couldn't have been more convenient for
5:15
me. Hera went on an incredible
5:18
ranked grind last month, aiming to set a new world record in
5:23
ELO. This grind felt a little bit
5:25
like the no hit torch only soul level 1 Dark Souls runs I've
5:29
been watching recently because he needed to reach a 3050 ELO
5:34
and every loss would cost him about as much ELO as eight wins
5:38
would gain. Fortunately, this resulted in a
5:42
huge number of recently played games which I decided to take a
5:46
closer look at. Specifically, I focused on the
5:49
period from September 8th to October 14th, just over a month,
5:56
during which Hera played a little over 200 ranked games.
6:00
I didn't analyze all of them though, only his losses.
6:04
In total, there's 39 defeats, which I examined and analysed in
6:08
detail. However, these are only losses
6:12
on Arabia. That means, for example, the
6:15
well known loss against Matza where Hara got douched is not
6:19
included, since that took place on a different map.
6:22
I also excluded another game, his recent loss against Doomsa,
6:27
because even though it was played on Arabia, it wasn't a
6:30
serious match. Hara was doing a challenge where
6:33
he deleted 1 villager for every 5 gifted Subs, so including that
6:37
would have obviously skewed the data.
6:40
In the end, I analysed 39 of Hara's losses on Arabia, hoping
6:44
to uncover systematic weaknesses that might offer clues on how to
6:49
actually beat the guy, even if you're not necessarily the
6:54
better player, because let's be honest, it's hard to see how
6:58
anyone could realistically be significantly better than him.
7:03
It didn't make much sense to go further back and include older
7:06
losses of his, since I specifically wanted to analyze
7:10
Hara's losses during this his world record attempt under the
7:14
assumption that he played them as if they were tournament
7:18
matches with absolute focus and determination to win full sweat
7:22
and try hard mode. I also restricted my analysis to
7:26
Arabia games to ensure better data comparability since
7:30
different maps play quite differently even on the ranked
7:33
ladder. For completeness, it's worth
7:36
mentioning that almost every match here I played during the
7:38
period was on Arabia anyway, with just a few exceptions like
7:43
an occasional arena game or a mega random match against Mr.
7:46
Yo. All right, I think that covers
7:49
all of the context, so let's take a look at the data.
7:52
But first, you should know how I actually collected it.
7:57
My data collection process had two major steps.
8:02
The second step was to confirm the data from the first step,
8:06
meaning I reviewed and verified the data from the first step and
8:09
supplemented it when necessary. In the first step, I analysed
8:14
all of Hera's losses on AOV 2 Insights and then compiled the
8:18
following information into an Excel spreadsheet.
8:22
Hera's opponent both play as civilizations.
8:25
The game length where the Hera researched loom in the Dark Age.
8:29
Hera's and his opponents openings, their mid game army
8:33
compositions, their late game army compositions, how many T CS
8:37
each had at the start of the the Castle Age, their Feudal, Castle
8:42
and Imperial Age up times, and finally their average APM
8:47
actions per minute, both of Hera and his opponent.
8:51
And I can already tell you in advance that Hera's APM is
8:54
basically always higher than his opponent's, ranging from about
8:58
20 APM more to almost double, such as in comparison to Mr. Yo.
9:04
However, APM turned out to be so insignificant for my analysis
9:08
that I ended up removing it from my sheet.
9:11
Not because it's unimportant, clearly Hera benefits greatly
9:14
from consistently influencing the game more actively than his
9:19
opponents, but simply because it's irrelevant to the question
9:22
of how to beat Hara. If you look at Hara's ranked win
9:27
rates, you'll notice that his lowest win rates are against two
9:31
very different players, Yo and Viper.
9:34
Yo is slower than average, while Viper is much faster than
9:38
average, yet Hara's win rate is still positive against both.
9:42
According to AOV 2 insights, Viper is is fourth most frequent
9:47
opponent, against whom Hera wins 56.5% of his games, while Yo,
9:53
his second most frequent opponent, wins 59% of the time
9:57
against him. Hera's most frequent opponent,
10:00
by the way, is MBL. What a surprise.
10:03
I'm mentioning all of this mainly to explain why APM will
10:08
not be brought up again in this episode, because that simply
10:11
doesn't appear to play any role in explaining why Hara loses
10:14
games, even though it is certainly one of several
10:18
relevant indicators for why he wins them all.
10:22
Right, with that out of the way, let's move on to the second step
10:25
of my data collection, which was much more time consuming, but
10:29
absolutely necessary because the data I got from AV2 Insights was
10:34
in some respects not specific or reliable enough.
10:38
For example, it wasn't always clear what the exact strategies
10:42
were in the opening, mid game, or late game.
10:45
The openings listed on AE2 Insights only matched the actual
10:48
openings in about half of the games, and even from the
10:52
military timeline it was sometimes impossible to tell
10:56
whether a player, for example, executed a successful drush and
11:00
then researched the minute arms upgrade, or whether it was
11:03
straight minute arms rush. Tower rushes, castle drops and
11:06
siege pushers were also barely visible, if at all on AOE. 2
11:11
insights. In short, I had to double check
11:14
everything myself. To do so, I went through all the
11:18
VOD's from Harrah's Twitch streams and reanalyze the games
11:23
from Harrah's perspective. This allowed me to add a new tab
11:26
to my excel sheet, because in almost every case Harrah gave at
11:30
least a brief and sometimes very detailed explanation of his
11:34
losses. After each game I took notes on
11:37
these and they turned out to be quite insightful for my
11:40
analysis. There are however 3 losses with
11:43
no VO DS as those were played off stream, and two additional
11:47
matches for which no AOE 2 insight statistics are
11:51
available. 1 is just not listed for whatever reason and the
11:56
other one is not able to analyse the game even though it is
12:01
listed. For these two games.
12:03
The exact up times are missing, of course.
12:06
I've documented all of this in my data set, and I'm mentioning
12:09
it in such detail because I'm also making this Excel sheet
12:13
publicly available both in the Discord server and in the show
12:17
notes of this podcast, so you can download it and take a look
12:21
at it yourself, or even use it for further analysis, additions,
12:25
or corrections. Now, I've given you all
12:29
necessary information of, well, how the data came to be, and I
12:34
think it's time for me to describe the sample a bit before
12:37
presenting my results. In the 39 losses that I have
12:42
analysed, Hera played against 16 different opponents.
12:47
The most frequents were MBL with eight losses and Yo with 7,
12:52
followed by Lucho with five. He suffered 4 losses to Sora,
12:56
Kuma to each against Kingston, Nikov, Sebastian and the Viper,
13:01
and one loss each to Anchor Bive or Bee Ive.
13:05
I don't know how to pronounce his name.
13:07
Antagonist Degao, Kazwa, Dr. Cosmonaut Lewis and finally
13:12
Vivi. There's nothing unusual about
13:14
this. It roughly represents the pool
13:17
of players he faces most often on the ranked ladder.
13:20
Keep in mind that on average, there are about eight wins
13:24
between every loss, usually against the same players except
13:27
for Viper, who perform significantly better than the
13:30
rest. A look at the Civilization shows
13:33
a very diverse picture. Hera's 39 losses were spread
13:37
across 29 different civilizations.
13:40
Three of those losses were with Bulgarians, and those were also
13:45
the only three games in which Hera himself identified
13:47
civilization choice as at least a partial reason for the defeat.
13:53
Bulgarians are by far Hera's weakest civilization, and he
13:57
sometimes seems unsure how to play them effectively on Arabia.
14:02
Apart from that, there was nothing significant about what
14:05
civilizations he had to play. Just remember he doesn't pick
14:09
his sieves, he got them assigned randomly.
14:12
They include strong infantry civilizations like the Aztecs,
14:15
Celts, Goths and Japanese, strong archicifs like Britons,
14:19
Ethiopians and Malay, powerful cavalry civilizations like the
14:23
Hindustanis, Slavs, and Berbers, as well as more defensive sieves
14:27
like Byzantines and Koreans, or gold dependent civilizations
14:30
like the Mayans, Burmese, and Turks.
14:33
He lost games with economically strong civilizations such as the
14:37
Vikings, Persians, Chinese and Romans, just as he did with very
14:41
aggressive ones like Lithuanians and Mongols and Woo.
14:45
The same pattern appears amongst his opponents civilizations.
14:50
I specifically checked whether he ever mentioned a sieve win at
14:53
the end of the game, but apart from the Bulgarian matches, he
14:56
only rarely said that a matchup was inherently difficult, and
15:01
even then he emphasized that the loss came down to other factors.
15:05
As for me, the pleb that I am, I couldn't identify any clear
15:09
Sieve wins either, which is honestly an amazing testament to
15:13
the game's balance. Incidentally, Hera doesn't seem
15:17
to have any first game of the day issues.
15:19
He doesn't lose more often at the beginning of a session, nor
15:23
at the end, which is incredibly impressive and also explains why
15:27
other pros say that the longer set goes, the fewer chances you
15:31
have against Hera. The guy's stamina is simply
15:35
unmatched. Which brings us to the length of
15:39
the games. For simplicity, I rounded all
15:43
game times up or down to the nearest minute.
15:46
On average, his lost games lasted about 45 minutes, which
15:50
by itself doesn't reveal anything special except the
15:53
Terror clearly doesn't tap out early, even in defeats.
15:57
However, averages alone don't mean much without context, so I
16:02
calculated the standard deviation of the rounded game
16:05
lengths, which came up to 14 point O3.
16:09
The standard deviation tells us how far on average, the values
16:13
deviate from the mean. So if we take the average game
16:18
length of 45 minutes plus -14 minutes, we get the typical
16:22
duration of Hera's lost games, which are between 31 and 59 in
16:27
game minutes. In her future references, I'll
16:30
just say 30 to 60 minutes. There are no real outliers on
16:35
the short end. There's a game against GAO which
16:37
Hera lost after 18 minutes, and on the long end match against
16:41
Lucho that lasted nearly 85 minutes, just to give you the
16:45
full range of all games. To draw meaningful insights from
16:50
this, we first need to consider what typically happens between
16:54
minutes thirty and sixty of a game.
16:57
To interpret this properly, we need contextual data about the
17:01
most common game progressions. The best way to get that is by
17:05
looking at the up times in the matches that I have analyzed.
17:10
In these games, Harrah's average Castle Age time was just over 20
17:14
minutes, with a standard deviation of 2.38, meaning he
17:18
usually reached castle age between the minute of 18 and 22.
17:22
His average Imperial Age time was minute 37, which a little
17:27
bit of a larger standard deviation of 5.6.
17:30
So in most of these games have a reach Imperial age between 32
17:34
and 42 minutes. The higher variation here makes
17:38
a perfect sense. The further a game progresses,
17:41
the more divergent it becomes. Accordingly, the standard
17:45
deviation for the Feudal Age of times was much smaller than for
17:48
Castle Age. This gives us an important
17:51
insight Hera tends to resign during the period that roughly
17:55
spans from the second-half of Castle Age up to full Imperial
18:00
Age. However, it's also worth noting
18:03
that in a surprisingly large number of these matches, Hera
18:06
never reached Imperial Age, even in games that lasted longer than
18:10
his average Imperial Age timing. In 18 out of the 39 games, he
18:17
never reached Imperial Age, and as far as I can tell, he didn't
18:20
even click up in any of those 18 games.
18:24
That's about 46% of his losses that never went for to imp for
18:28
him. There's another important
18:31
distinction we need to make, though, one that complicates the
18:34
interpretation of these timings. The moment a pro player like
18:39
Hera resigns is usually not the same moment the game was
18:42
actually lost. In Hera's games, there were
18:46
generally two types of loss scenarios.
18:49
First one is a singular moment that snowballed, for example, in
18:52
a match against MBL where he misclicked his entire army and
18:56
immediately lost the game as a result.
18:59
He didn't give up right away, but that was definitely the
19:02
point from which the loss derived.
19:05
The second scenario is more of a complex strategy.
19:09
Strategy based reason involving a combination of factors that
19:13
don't necessarily occur at the same time.
19:16
For instance, a match where the opponent first secured key map
19:19
areas like Central Hills, and Harry kept trying to reclaim
19:22
that position or claim that position for a long time, but
19:27
couldn't, as in one of his games against Mr. Yo.
19:31
In these more complex cases, the actual cause of the defeat lies
19:34
earlier in time than the moment of resignation.
19:37
Unlike the one mistake scenarios where it's often more closely
19:41
together, Hera tends to keep playing after until he has
19:45
virtually zero chance of winning, or sometimes even
19:48
beyond that, in the hopes that the opponent might slip up.
19:52
So what does this tell us? At the very least, that we
19:56
always need to look a bit earlier in the timeline to find
19:59
the true moment of defeat. And distinguish that from the
20:03
moment the game has ended. With this understanding, I turn
20:06
back to the data and try to further characterize the time
20:10
frame between minutes 30 and 60. And remember, we're talking
20:14
exclusively about Arabia games and independent of
20:18
civilizations. To me, this period, regardless
20:23
of age transitions, represents the phase of the game where the
20:27
economy is mostly or even fully established and the focus shifts
20:32
to gaining or maintaining map control.
20:35
In this phase of the game, depending on whether a player
20:37
expects to have a late game advantage or a disadvantage in
20:41
the civilization match up, they will either expand to secure
20:45
late game resources or launch attacks to claim key strategic
20:50
areas, especially central hills which serve as strong positions
20:54
for launching offensives. Another characteristic of this
20:58
phase is that it marks the end of the mid game or the beginning
21:02
of the late game when gold units dominate, making gold control
21:06
crucial. Around minute 60, gold is often
21:10
mostly mined out and the game transitions into the post
21:13
Imperial trash war phase that I despise so much.
21:18
At the very least, by the hour mark, this phase should already
21:22
be prepared through tech switches.
21:25
This aligns with Harrah's reputation as the king of the
21:28
late game and the idea that he is much harder to beat once the
21:32
game reaches that stage. After all, two things
21:35
immediately come to mind when you think of Harrah, his
21:38
unmatched defensive play and his signature move Hustlerage.
21:42
With 80 or more farms. We can already infer something
21:46
from this. When playing against Terror,
21:48
it's advisable to aim to decide the game in the mid game rather
21:53
than the late game. And I'm explicitly saying
21:56
deciding the game, not necessarily winning the game.
22:00
The only question is how to do that.
22:03
So to explore on that, let's take a closer look at the
22:07
individual faces of the game in the Dark Age.
22:10
There isn't really much to analyze.
22:13
It's very much standardized. In 79% of the games, or 31 out
22:19
of 39 analyzed games, Hera researched Loom before clicking
22:22
up. Even though Hera is known as an
22:25
advocate for clicking up without Loom, he actually does it more
22:28
often than not. I couldn't find any evidence
22:31
that skipping Loom game gave him a measurable advantage in the
22:35
cases where he did so. By comparing up times with his
22:39
opponents, I was able to rule out that differences there were
22:44
the decisive factor or even the cause of his losses.
22:48
Whoever reached the Feudal Age or Castle Age first was split
22:52
roughly 5050, and in most cases the time differences were
22:55
negligible both for Feudal Age and Castle Age.
23:00
The Imperial Age tells a different story though.
23:02
As already mentioned, in 46% of the analyzed games, Hera never
23:08
reached it at all. In seven out of 39 games, the
23:13
opponent reached Imp and Hera didn't, and in almost all of
23:16
those Hera resigned shortly after his opponent advanced.
23:20
In one game Hera reached Imp and his opponent didn't, and that,
23:25
according to Hera himself afterwards, was the mistake.
23:29
He said he should have never clicked up and that was a highly
23:32
aggressive match against MBL. Comparing up times between Hara
23:36
and his opponents is very interesting, but it's more of a
23:40
symptom of what had happened earlier rather than the cause of
23:44
the loss. So I decided to focus on
23:47
something else instead. Hara's economy and that of his
23:51
opponents. Specifically what their mid game
23:54
setups looked like. Meaning how many town centres
23:57
were added at the beginning of Castle age?
24:00
In 13 out of 35 games, that's 37%.
24:04
Hera had the same number of TCS as his opponent at the start of
24:08
Castle age, whether that was 1-2 or three.
24:11
The data set here only includes 35 games because in four games
24:16
that information couldn't be determined. 2 games were lost
24:19
already in Feudal age, so they never reached Castle age. 1 was
24:23
against Sebastian, the other one was against the GAO, and the
24:27
other two couldn't be analysed through AOE 2 insights.
24:30
So I simply don't know. And since I only have Harrah's
24:34
POV from the streams, I couldn't tell what the opponent was
24:36
doing. So once again, 37% of the games
24:40
Harrah had the same number of disease as his opponent at the
24:43
start of the Castle Age. In 49% of the games,
24:47
specifically, 17 out of 35, Hera had fewer TCS.
24:52
In 13 of these games, he had only one TC, and in the other
24:56
four he had two. That of course means that Hera
25:01
lost only 5 out of 35 games in which he had more TCS than his
25:06
opponent. That's about 14% of the total.
25:10
Four of these five games didn't reach the late game.
25:13
They were decided in Castle Age, with Hera adding one TC while
25:17
his opponent stayed on one. In the 5th case, Hera had the
25:20
three TCS and his opponent had two and one last statistic in
25:26
only four out of 35 analysed losses, Hera had three TCS at
25:30
the start of Castle age. Again, to clarify, this is about
25:34
how many TCS were added shortly after reaching Castle Age, not
25:38
about additional TCS built later in the Castle Age or even late
25:42
game. This metric serves as an
25:44
indicator of play styles. Whether the focus was more on
25:48
economy early on or more more on military.
25:52
I think these findings are quite revealing.
25:54
After all, they show that in 86% of the losses in my sample, Herr
26:00
was unable to establish an early economic lead, and in 49% of
26:05
cases he actually fell behind economically.
26:08
Combined with the analysis of game lengths and up times, this
26:13
suggests that the likelihood of beating Hara increases if you
26:17
can prevent him early on from gaining in the in economic
26:22
advantage. And you don't necessarily need
26:24
to have a huge economic lead yourself to do so.
26:28
Of course there are different ways to achieve this, but based
26:32
on the up times we at least know that in these games Harris
26:36
opponents did not succeed by because they invested heavily
26:40
into feudal age aggression. Still, a slightly stronger
26:44
feudal investment might be worthwhile as long as it's not
26:48
an all in, since Hera almost never loses to all insurance
26:52
except if it's Matza and TTC drops him.
26:56
But to say that with more confidence I want to rely again
27:00
on my data, specifically by analysing the openings used by
27:04
Hera and his opponents. It's important to note that
27:07
openings are often determined by the match up and by both players
27:11
maps, especially how resource resources are distributed.
27:16
This is why I specifically looked at all maps very
27:19
detailed. And at the end it's clear to say
27:23
that in a few games Hera did have a bad map, but in most
27:27
cases his maps were fine and the same applied to his opponent's
27:32
maps as far as I could tell. Moreover, when Hera had a bad
27:36
map, his opponent's map was usually bad too.
27:40
And what exactly defines a bad map?
27:42
It's a map that's hard to wall and where resources are
27:45
difficult to protect, meaning berries and gold are exposed and
27:50
secondary gold piles are on hills, making it impossible to
27:54
place TCS next to them or behind wood lines, which carries its
27:58
own risks. And all wood lines are long and
28:01
thin and spawn towards the center of the map.
28:03
There's also bad set up for your base because tower rushes or
28:08
Archers can easily deny your wood access.
28:11
As mentioned, Hera did have such maps in a few cases, but they
28:15
were the exception. Now back to the openings.
28:19
In 12 of his 39 losses, almost 31% Hera opened with minute arms
28:25
or a drush, but mostly minute arms.
28:29
The follow up in these cases was skirmishes in seven out of 12
28:32
games and Archers in the remaining 5.
28:35
In another 14 games, Hera opened scouts, followed up with
28:38
Archers, skirms, or more rarely just Spearman.
28:42
In five games, about 13% he opened skirmishes, usually
28:46
paired with Spearman, and in six games he opened Archers, often
28:51
followed by Spearman or scouts. The remaining two games featured
28:55
unusual openings. 1 was a tower rush as Byzantines with a scum
28:59
follow up against Lucho, and one was a Flemish militia opening as
29:04
Burgundians against Lewis. Given this distribution, it
29:08
becomes clear that there are no significant patterns or
29:12
irregularities in Harrah's openings.
29:14
The opening itself is, I would say, not a reliable indicator of
29:19
whether Harrah would lose a game.
29:22
So what about the openings of his opponents?
29:26
In 21 out of 39 analysed games, Harrah's opponents opened with
29:31
scouts. That's about 54% of all games.
29:35
Half of those were followed up with Archers, the other half
29:39
with scouts. Scouts are among the most
29:42
aggressive openings because aside from a drush or a minute
29:45
arms opening, nothing reaches the opponent's base as quickly
29:48
as a small group of scouts. Even a single scout can cause
29:52
disruption, and three scouts can already take on a Spearman.
29:56
The Archer follow up represents the most aggressive play style
29:59
in my opinion, while the Scum follow up aims for a faster up
30:02
time since it doesn't require you to invest in gold that much.
30:08
In addition, there were seven more games where Harris
30:11
opponents opened Archers, then either stayed with Archers or
30:14
transitioned into scouts or added scums for support.
30:19
Combined with the scout openings, that makes 29 out of
30:22
the 39 games, or 74%. Roughly 3 out of four matches in
30:28
which Harrah's opponents chose highly aggressive openings.
30:32
The remaining games included 2 tower rushes, some rushes and
30:36
minute arms openings, and one trash opening by Nikhov with
30:39
Incas, which is not that surprising to be honest.
30:42
And only three games where skirms were the main opening
30:45
that actually won, which is quite astonishing and
30:49
remarkable. Overall.
30:52
Hera's opponents opened with men at arms far less often and
30:56
scouts much more frequently than Hera did.
31:00
Most notably the scout Archer combination.
31:03
And Archers in general were chosen much more often by his
31:06
opponents than by Hera himself, while scones were used much less
31:10
frequently, mostly as a follow up to scouts or as a supplement
31:15
to arches. In Feudal Age.
31:17
Scouts and Archers, especially in combination, seem to be
31:20
something Hera struggles more against than other openings.
31:24
That's somewhat logical. In my opinion.
31:27
It's the strongest opening overall, since it allows you to
31:30
take the initiative, but it's still a notable pattern.
31:34
So what can we learn from this when asking how to beat Hera?
31:38
Nothing conclusive yet, but it's at least hints that early
31:43
aggressions tend to pay off. Let's keep that in mind because
31:47
it might become relevant again later.
31:50
Now let's jump to the mid game and look at the armour
31:53
compositions played there. Looking at Hera's armies, one
31:58
thing stands out. Skirmishes appear very
32:01
frequently. Out of 36 relevant games, three
32:05
had no real mid games, so I couldn't include them in my
32:08
analysis. 25 featured skirms as part of the Castle Age army
32:14
composition of Hera. That's roughly 69% of the games.
32:19
In nine of those 36 games, about 25% or one in four, Hera played
32:25
only skirmishes. Only skirmishes in the mid game.
32:29
In other eight games he played Knights and skirts, and the rest
32:32
were outliers. Combinations like cavalry,
32:35
Archers, crossbows or long swords mixed with skirmishes.
32:38
They That may not sound surprising given the persistent
32:42
skirm meta, but looking at the mid games of his opponents
32:47
reveals a different picture. They hardly played skirms at
32:52
all. Here all 39 games are relevant
32:55
since every opponent reach the mid game.
32:58
Only 9 out of 39 games, which is 23%, featured skirmishes as part
33:05
of the army composition and skirms only was played just once
33:10
in a game where Hera went arches only.
33:14
In the remaining seven games skirmishes were used as a
33:17
support unit alongside Knights, Archers and Spearman.
33:21
And in one wild match MBL as Koreans used skirmishes and fire
33:26
lenses to support a mid game trash.
33:29
But in this game skirmishes were were more or less insignificant.
33:33
Fire lenses were the most used unit and if you're cross
33:38
checking this with my excel sheet, don't be confused.
33:41
I also included the game where you played with janitors, since
33:46
janitors are essentially mounted skirmishes.
33:49
So in summary, Hera played skirmishes in 69% of his mid
33:53
games while his opponents did so in only 21%.
33:59
So what did they play instead you might ask?
34:02
Well, I have the answer for you. Pretty much everything that
34:07
costs gold, most notably crossbows, Knights, cavalry,
34:11
Archers. Sometimes in combination with
34:13
other units, though rarely with trash units.
34:16
Instead we see far more unique units or regional units from his
34:21
opponents. Archers with eagles, cave
34:24
Archers with srevamshahs, Archers with step Lancers,
34:27
karambes, fire Lancers, camels, war elephants, kamayuks,
34:31
slingers, hussite wagons. All of this in castle age.
34:36
And all of these units are gold units.
34:39
They played a lot of double gold compositions.
34:43
In other words, in these analysed games, Harrah's
34:45
opponents fielded many more gold units and also a greater variety
34:50
of units overall than he did, especially less trash and in
34:55
particular fewer skirmishes. How can we interpret that?
35:00
Well, as a rough approximation, we can propose 2 hypotheses.
35:05
First is that these diverse army compositions are simply the
35:09
result of the preceding game state, essentially A
35:12
continuation from the Fuel Age where things already looked
35:16
similar. That's quite plausible, since
35:19
players naturally tend to carry their momentum from Feudal Age
35:22
into Castle Age. This also fits with Harrah's
35:25
general defensive play style. In that case, these games would
35:29
be the games where the opponent managed to break Hera's
35:32
defenses. Alternatively, this is the
35:34
second hypothesis. The double gold compositions
35:38
commonly played in Castle H by his opponents in these games
35:42
might have forced Hera into more defensive army compositions.
35:47
In that case, he was essentially overwhelmed and tried to
35:51
withstand the pressure by going into more trash units.
35:56
Both hypotheses are supported by the earlier analysis of game
36:00
lengths and up times, because, as you might recall, I
36:03
characterized this face of the game as the one where gold plays
36:07
a crucial role. To explore this further, I
36:11
reviewed his dreams, specifically the moments where
36:14
he decided on his opening and his Castle H composition.
36:18
It became clear that in Fuel Age, Harrah's decisions were
36:21
usually based on the map and the match up, though in a few
36:24
exceptions he first scouted what the opponent was doing and then
36:28
adapters adapted his opening accordingly, typically opting
36:32
for skirms in those cases. In Castle Age, however, things
36:36
look different. Hera focuses intensively on
36:40
anticipating the opponent's army composition and then adjusting
36:44
his own accordingly. If his scouting gives him that
36:47
information, he'll try to play the counter.
36:50
If the opponent is heavily walled and he can't scout these
36:54
informations, he'll infer the army composition based on
36:58
something like unit upgrades. And he's rarely wrong with that.
37:04
Hera just proved in many, many cases, even when he was losing
37:09
games, that he had an incredible game sense.
37:13
Yet despite all this careful consideration, in the games
37:17
analysed here, Hera almost always stuck to the composition
37:21
he had already been playing in the feudal age, with the only
37:24
major change being scouts turning into Knights.
37:28
If he did deviate, it was usually towards skirmishes.
37:33
His opponents, on the other hand, performed far more tech
37:36
switches and fielded a much greater number of gold heavy
37:39
units. This connects well to another
37:43
factor that can only be indirectly inferred from AV2
37:46
Insights data through the number of town centres placed early in
37:51
Carthage, but which became much clearer through the Vods that
37:56
I've looked at. And that's the use of siege
38:00
weapons and map control. The striking commonality across
38:04
nearly all analysed games is that Hera failed to secure real,
38:09
lasting control of the map centre.
38:12
In six games he tried to apply pressure through a siege push,
38:15
but those pushes were either completely stopped or just
38:19
didn't pay off as they failed to deal enough damage despite
38:23
having investment. And by siege push I specifically
38:27
mean the mid game strategy where he builds a siege workshop in
38:31
the centre of the map or near the opponent's base.
38:34
Of course, siege units played some role in almost every game,
38:38
but less as a strategic focus and more as a just, well,
38:42
support unit. And to be honest, Harrah didn't
38:45
look particularly good in those siege pushes.
38:48
Losing a siege push to MBL is nothing surprising, but his
38:52
pushes also failed against Lucho, Kazwa and Mr. Yo.
38:55
The consequence in each case was that the opponent gained Mac
38:59
control and from there the game snowballed out of Hera's RE.
39:02
And there is something that struck to me when I was
39:07
analysing these games and that is that Leary, when playing in
39:12
tournament games against Hera, significantly plays a lot a lot
39:18
of siege pushes and especially with Manganiels.
39:22
There were more than a couple of games where Leary was actually
39:27
pushing Hera from one angle with up to 5 manganiels at the same
39:33
time, trying to get Hera into giving up his defensive position
39:38
and forcing him to play with his own siege to defend.
39:43
In two other matches, one against YO, one against
39:45
Kingston, Hera never even had the opportunity to take map
39:49
control as he was under pressure from enemy siege pushes from the
39:54
start. I think it's not it's not so
39:57
much about the siege push itself to be honest, but rather about
40:01
what the siege push does to him. When Hara goes for a siege push,
40:04
he often ends up with overly complex army composition and
40:08
tries to micro three groups at once while his opponents try
40:12
trade much more efficiently. Hara spikes to 200 APM in these
40:16
moments, while his opponent successfully defends with less
40:20
than half of that. Simply because they only have
40:23
manganals and maybe one other unit type on the field.
40:28
That same pattern holds true when Hera defends using Siege,
40:32
at least in the games I have analysed.
40:34
It almost seems as though Hera has a particular game flow that
40:38
he tries to keep up. Very rhythmic, very fast.
40:42
Of course, Siege by contrast, is too slow for that kind of style.
40:46
For his rhythm maybe so he tries to multitask two or three other
40:51
things simultaneously, but that's just not easy to
40:54
reconcile. In the games where siege pushes
40:58
weren't a major factor, control of the map center is still
41:01
played, still played a massive or decisive role, especially
41:05
control over key resources and hills.
41:08
Whenever Hera didn't see a way to gain control over these areas
41:12
he resigned and I think he simply recognized that sooner
41:15
than non pros like I would do. That also explains the
41:19
phenomenon where we sometimes think a player resigns too
41:23
early. We're looking only at metrics
41:25
like populations, score and resources, while players also
41:29
consider map control and they know that without it all the
41:33
resources in the world won't save them.
41:36
Since Hera frequently comments on his own games, we can gain a
41:40
good insight into his perspective.
41:43
Apart from the three games where he played as Bulgarians, he was
41:46
generally able to give very plausible explanations for why a
41:50
game was lost, and was almost never because of a single fight,
41:55
but mostly because he lost the map.
41:58
That can happen through a battle, a single battle maybe,
42:02
but also through worse timings or unpredictable moves.
42:07
I'm going to play you a short audiophile.
42:10
Now that's a little bit unfair towards Hera, because if there's
42:15
one thing to say about him, then it's that he has an incredible
42:18
sense of sportsmanship. Win or lose is always respectful
42:22
and fair. In most of the losses analysed
42:25
here, Hera not only explained why what he did wrong, but also
42:29
praised his opponent. Only one game was a little
42:32
different and that's the source of this audiophile.
42:36
A match against Sora Kuma where Hera played Mongols versus Huns.
42:41
It was the second-half of the mid game.
42:43
Hera had added TCS and was trying to secure a key strategic
42:47
position, kind of guarding his TC with a castle when Sora Kuma
42:52
suddenly showed up out of nowhere with military units.
42:56
No, the force, of course. What do we store Kuma if he
43:02
doesn't just lock his way into every aggressive position he AM?
43:11
Can't believe it. I literally can't believe it.
43:16
I just make a castle. Make a fucking castle.
43:22
Lucky bastard, honestly. It was the first and only time
43:26
that Hera got visibly upset. Personally, I can totally
43:31
understand that Sora Kumar played aggressively, didn't let
43:34
him boom freely, and caught him right as he was trying to
43:38
transition into a defensive set up.
43:40
If that had happened to me, I'd probably had closed Age of
43:44
Empires 2, complained about Smurfs and then reopened Age of
43:48
Empires 2. Very annoyed just to repeat the
43:50
same thing again. Harry unfortunately is a much
43:53
more professional than I am and he simply kept on playing.
43:57
So what exactly happened there? Harris Castle wasn't in the
44:01
center of the map, but rather to the north, close to his
44:04
expansion, which he wanted to protect with it.
44:07
When Sarakuma denied that castle, it completely threw Hera
44:11
off balance. For several minutes he was
44:13
housed, barely producing military and entirely focused on
44:17
that one spot on the map. At that point it was already
44:21
clear that the game was over, even though he kept fighting
44:24
back for a while. In all of my analysis, I haven't
44:28
seen any other instance where an opponent's attack had such a
44:31
strong psychological effect on him and such a huge impact on
44:36
the gameplay. So the question is, what can we
44:39
learn from that? Well, Hera said that Sora Kuma
44:43
just got lucky. More precisely, he said he locks
44:46
his way into every aggressive situation in the game.
44:49
That sounds suspiciously like Sora Kuma forces his own luck.
44:54
From everything I've seen, I can only explain it by saying that
44:58
Sora Kuma takes risks, which makes him unpredictable in some
45:02
places. And normally that's not a good
45:05
idea against a stronger opponent like Hera is, which is why Hera
45:09
usually wins those situations, especially against Sarah Kumar.
45:13
But it also creates games where the opponent breaks Hera's
45:17
expectation and that might lead to victory.
45:21
That leads me to another another observation, something that Hera
45:25
has often said when he has lost a game.
45:28
And that was while still while playing the game.
45:32
And he says I can't play the game.
45:34
By that, he of course doesn't mean that he's bad by any means.
45:38
He means that he can't play the game, he wants to play it, he
45:43
loses strategic access, or in other words, he can't make
45:48
anything happen anymore. And that happened, for example,
45:51
quite often against MBL or Yol, especially when they were
45:55
surrounding him or when Hera was denied access to any part of the
46:00
map outside of his base. And finally, this brings us back
46:05
to the observation that in all of Hera's losses, he first lost
46:09
control of the map. Considering the game lengths and
46:12
the argument that games are often lost before they are
46:15
surrendered, we can locate this very well in 30 to 60 minute
46:19
window with an average of 45 minutes game lengths.
46:24
All in all, the analysis regarding how to beat Hera can
46:27
be summarized as follows. You don't actually need to win
46:31
directly against Hera. You need to win the map.
46:35
And if you can defend that map against Hera, which you should
46:39
be able to because you have an advantage, you will probably win
46:43
against Hera. Once you control the map, Hera's
46:47
probably unmatched strategic versatility is constrained and
46:52
he is pulled out of his defensive comfort zone which is
46:55
his real strength. He is forced to play the mid
46:59
game with limited options and cannot focus on his late game
47:03
set up where he dominates thanks to his extremely high APM, his
47:09
very precise play, and just his probably unmatched game
47:13
knowledge. I don't think the goal should be
47:17
to defeat Terror in the early or mid game as Tarto attempted in
47:22
past tournaments using risky off meta strategies.
47:25
Instead the focus should be on winning the mid game through map
47:28
control so that Hera enters the late game in a weakened state.
47:33
You don't need to stop his late game, I believe you just need to
47:36
be to to have a better setup for it.
47:39
Based on the games I've analysed here, in 46% of cases it never
47:44
even reaches the the late game if you have a good setup in mid
47:48
game. As silly as it sounds, The old
47:51
Viper who played purely from map control would arguably have the
47:55
best chance against Hara if he played that way again.
48:00
He would just need to be probably a bit less greedy and
48:04
invest more into military to secure his positions.
48:07
Otherwise Hara simply gets ahead.
48:10
Especially because Hara's not stupid, he himself plays for map
48:15
control, so you probably have to beat him in contesting map
48:20
control. Much of this data supports that
48:22
hypothesis. And there are many ways to gain
48:25
map control. For example, siege pushes
48:28
overwhelming aggression like Sora Kumar does, or simply
48:32
double gold compositions to gain the upper hand as early as
48:37
possible. Is this a guaranteed victory
48:40
against Hara? Absolutely not.
48:43
Hara is probably the best mechanical player and the most
48:46
versatile, but he is predictable in the sense that he is an
48:51
absolute matter player with a very defensive baseline.
48:55
The window to win against Hera is in the mid game to early late
48:59
game. But to be clear, none of Hera's
49:02
losses came from an all in like a phosphor rush.
49:06
They were won strategically through map control and better
49:10
army compositions. And by better I mostly mean more
49:14
expensive and stronger. That's why APM doesn't matter in
49:18
my analysis, it only comes into play.
49:21
I mean APM only comes into play post Imperial where Hera can
49:25
simply control 200 units more efficiently.
49:28
But this doesn't really happen in the mid game when they are
49:33
fighting for map control. For the next king of the desert
49:37
tournament. I hope pros focus more on
49:40
playing the map and feel less pressured to rely on off meta
49:45
strategies against terror, because at least in these
49:48
analysed games it shows that the meta is the only reliable way to
49:53
actually beat him. However, for the sake of
49:56
completeness, I must also mention a few very important
50:00
limitations regarding the interpretation of my data and
50:04
its conclusions. The biggest limitation of
50:06
everything I've put together is that I only analysed his losses,
50:11
not his numerous wins. For example, if Hera in his wins
50:15
also plays significantly more skirmishes than his opponents
50:20
and usually defeats their gold heavy units, that would
50:23
completely change the interpretation of this data.
50:26
One could then argue that maybe it's the opponent's heavy gold
50:30
investment that is the problem, and these 30 nines gave I have
50:34
analysed are just random exceptions, while if they had
50:37
invested more in trash units they might have won more games.
50:42
The same applies to his siege pushes.
50:45
If he wins more games through siege pushes then he loses.
50:49
Then my analysis loses some of its significance.
50:53
Finally, the same goes for the relevance of map control.
50:56
If Hera usually relinquishes map control in the games he has won
51:01
and then regains it in the late game, then what I concluded here
51:06
would simply not hold. None of this can be answered
51:09
from the games I have analysed. We would need data from his
51:14
games that he has won, ideally from the same time period.
51:18
What can be said, however, is that the patterns observed in
51:22
Hera's losses do exist, at least in the games that I have
51:26
analysed. As I always tell my students,
51:28
statistical correlations must mean something for us to draw
51:34
conclusions from them. There's also room for
51:37
improvement when analysing the data that I have gathered.
51:41
For example, the economic situation of the players could
51:44
be measured much more precisely by tracking the village account
51:48
of both players at multiple points in the game, which is
51:52
probably a better indicator of economic investment than the TCS
51:56
at the start of the Castle age. Resources collected is also a
52:00
better metric in this context I believe, but these were all data
52:04
that I had not gathered. Furthermore, 39 games over the
52:08
course of a month is a relatively small sample size.
52:11
Ideally we would have many more games played under the same
52:15
conditions with a broader representations of match ups.
52:19
There is still a lot that could be added and improved.
52:22
Nevertheless, I hope you have enjoyed this slightly different
52:27
podcast. If you want, you can even try
52:29
your own data analysis and expand the spreadsheet that I
52:32
have provided. You can find it in the show
52:35
notes and on my Discord server for download.
52:37
You can also send me your corrections and additions via
52:40
e-mail at age of cast@gmail.com, that's ageofcast@gmail.com or
52:47
send me a message on Discord. This podcast is only possible
52:51
thanks to the supporters on Patreon.
52:53
Steady. Thank you very much for your
52:55
ongoing support and if you are not yet a supporter, I would be
52:59
very happy if you consider supporting me on Patreon
53:03
thatspatreon.com/start the Game already.
53:07
It also helps a lot if you rate the podcast on Spotify and
53:11
elsewhere so it gets more visibility.
53:14
I'm very curious to hear your theories on how you think Hera
53:18
can be beaten. Feel free to send them to me and
53:21
I will discuss them in the next magazine issue in the mail
53:25
section. If you want more podcasts, just
53:28
check out the feed on Spotify or wherever you're listening to it.
53:31
Several episodes are now available in English as well,
53:35
giving you many hours of content.
53:37
And if you enjoyed, follow the podcast on Spotify.
53:40
The more followers I have, the better it is for the podcast.
53:44
That's it for this episode. I wish you all a lot of fun with
53:48
the Lean Yuan Cup and hopefully we'll hear each other again in
53:52
one of the next episodes.