Mid Game: Objective Control
In the previous article, we looked into the aggression patterns of junglers and saw how they shaped the early game, as well as where and when junglers were most likely to succeed. We also saw that there were clear differences in terms of starts between the blue side and red side. All of this was already quite insightful, but most of the analysis was limited to the first 3–10 minutes of the game.
In this article, we’ll try to look into what happens past the 10-minute mark. We’ll see how objectives shape the flow of the mid-game by trying to answer the following questions:
- What is the first objective taken?
- When do players take the first objective?
- When do players take the first turret?
- When do players take the first Drake?
- When do players take the Rift Herald?
- Is the Infernal Drake really prioritized if it is the first dragon to spawn?
- When do players take outer turrets?
- What is the most common sequence of objectives taken?
Strap on, there’s a lot to go through!
Caveats
Most of what will be talked about in this article is stuff most people already intuitively know. This article just reinforces this intuition by putting numbers on it.
The data we are using is from high ELO players (dia2+), who *may* be more efficient and more knowledgeable about the game than the average player.
This is aggregated data. The findings presented in this article are not to be taken as gospel, but rather as a general rule of thumb.
There was a bug during the acquisition of the data, and a sizable chunk of events have no associated matches, which made it impossible to account for them when creating objective sequences, as this naturally relies on events belonging to certain matches.
TL;DR
- The first objective taken is a Turret in 65% of the matches, a dragon in 30% of the matches, and the Herald in the 5% remaining.
- If the first objective taken is a Turret, it is the Bot Turret in 44.8% of the games, the Top Turret in 34.4% of the games, and the Mid Turret in only 20.9%.
- As a first objective, the Bot Turret falls around 12:10, the Top Turret does so around 12:40 and the Mid Turret around 14:00, with a pretty large spread.
- On average, the Bot Outer Turret falls around 14:45, the Top Outer Turret around 15:30 and the Mid Outer Turret around 16:45.
- The differences in “first objective taken” by the blue side and the red side can be almost entirely explained by the difference in ganking patterns between the two sides.
- Although Infernal Drakes are as likely as other Drakes to spawn with 25% chance, they represent 38.3% of the first objective taken, when this objective is a Dragon. They completely warp the flow of the game, with them representing 38.8% of the possible first objectives in games where they are present, whereas other Drakes account for around 26–27%, and when the spawning dragon is a Cloud Drake, dragon focus falls down to 22.8%, with Tower focus rising to 71.3%.
- As a first objective, Infernal Drakes are taken around the 10' mark, whereas the other dragons are usually taken around 11:20. This difference of over a minute can cause a drastic loss in dragon control if not accounted for in warding habits.
- To have vision over 80% of standard dragons taken, players must start warding it from the 8th minute. If the dragon is an Infernal Drake, then the warding must start in the 7th minute to keep an 80% coverage. If warding is only started in the 8th minute, the control will only cover 65.6% of the fire drakes.
- There are clear differences in behavior when it comes to taking the Rift Herald depending on whether no objective was taken. If there has yet to be an objective taken, the average time around which Rift Herald is taken is 13:20, and failing to cover from 10:00 to 12:00 results in 30% chance of missing Rift Herald. Otherwise, the average time which Rift Herald is taken around 15:45, and players should start warding around 13:00 if they tolerate an 80% Rift Herald control.
- Blue Side holds an advantage when it comes to Outer Turrets, with Blue Side taking the Mid Turret about 10s earlier than Red Side, and the Top Turret about 35s earlier. Blue Side also takes Outer Turrets slightly more often than Red, by about 2%. This does not mean that Red Turrets are easier to take, but merely that the blue side advantage can be seen in this data.
- As a general rule of thumb, teams take at most 6–8 objectives during the 10 to 20 minute period, and usually take around 3 objectives.
- Side Turrets only matter more than Mid Turrets during as first objectives.
- Infernal Drakes are prioritized over other objectives for the first few objectives (5–6).
- Elder Drakes appear much later in the objective sequences than Barons, which suggests that if Barons are late-game objectives, Elder Drakes are very-late-game objectives, and are actually so late to spawn that over 86% of games are done before Elder Drake even has a chance to spawn.
The First Objective
If the early game is ruled by the presence or absence of the jungler on lanes, the mid game is started when the first objective falls, allowing players to move more freely around the map. Depending on the nature of the first objective taken, different resources will be freed. For instance, if the top turret is the first to fall, it frees the movement of a player, whereas if the first turret to fall is the bottom turret, it frees up two players; and if a dragon is taken, it allows the jungler to go topside without risking it being taken.
Type of the First Objective
So, what type of objective is usually being taken first? Is it usually a dragon, the top turret, or even perhaps the Rift Herald?
Immediately, we can see that the two most focused first objectives are the bottom turret and dragon. This suggests a strong focus around the bottom side of the lane from players as if the key to opening the mid game was the bottom lane. However, the relatively strong presence of toplane tower seems to represent a sort of balance to that focus.
Impact of Blue/Red Side on First Objective Taken
We’ve seen in the previous article that there were notable differences between jungler pathing on the blue and red side, with blue-side junglers having a stronger presence top than their counterpart early in the game, and red-side jungler having a more diffuse presence due to their less clear cut starting pattern. This difference in gank focus could potentially lead to differences in objective control. So let’s split this distribution by side:
As we can see, there are clear differences in which objectives are taken by each side, and the differences in objective seem to echo the differences in ganking patterns for the 4th minute that we saw in the previous article. The two datasets are basically begging for a correlation analysis. However, if our ganking data set worked with 3 lanes, this dataset also includes neutral objectives — Dragons and Rift Herald. In order to compare the two, we need to distribute these neutral objectives across the lanes. How do we do that? Well, this is entirely arbitrary, but I chose to distribute them between lanes based on the number of players on that lane, so the Rift Herald would be split evenly between top and mid, while 2/3 of the Dragons would go to the botlane (where there are 2 players), and 1/3 would go to the midlane (where there’s only one player). This seemed like a relatively reasonable distribution of neutral objectives, but I have no strong arguments to affirm its validity. However, the results are absolutely mind boggling. Here’s the table with the differences in both ganking patterns and objective focus:
These two data sets have a correlation of 0.999984. This is probably the highest correlation value I have ever seen in the wild. However, remember that correlation ≠ causation. Is the difference in objectives really caused by the difference in ganks? I won’t go into a causality analysis, but my gut feeling is that it is extremely likely that the variations in the gank patterns cause the variations in objective taken, as they happen before, have a clear impact on the lane dynamics, and the causal factor of the difference in ganks (i.e. the side) doesn’t really look like it should have an impact on objectives, as the number of resources (i.e. players) available on each objective is identical due to the axial symmetry between the two side. Note that this does not suggest that where the jungler first ganks determines where the first objective will fall, but rather that the differences between the ganking patterns of each side entirely explains the differences in objective taken. (although a quick check seems to put a 0.95 correlation between jungler presence before 13' and first objective taken, not detailed in this article)
Evolution of First Objective Taken over Time
We’ve seen the general distribution of first objectives, we’ll now see how that distribution evolves over time, to try and see if anything can be learned.
It seems that the first objective prioritization respects the following pattern: Focus on Drake, if that doesn’t work, focus on botlane, if that doesn’t work focus on top, if that doesn’t work focus on mid, and if all else fail, focus on Rift Herald. This, however, is just a general observation. We’ll now focus on Tower as first objectives, then Dragon as first Objectives, and finally Rift Herald as a first objective.
Tower as First Objectives
Let’s take our previous graph, and remove neutral objectives:
We’ve already seen that the Bottom Turret was the most focused turret, followed by Top, and then Mid. But, aside from a difference in amplitude of the signal, is there also a difference in where they are centered in time? Do people take their first objective earlier if it’s a bottom, mid or top turret? And if so, by how much? This is actually rather hard to see just by looking at the previous graph. We can kind of guess that the botlane turret maybe falls slightly earlier than the top turret, and that taking the mid turret first means spending more time on that first objective, but the exact values are unclear. So here’s a table with those value, as well as the standard deviation for each turret type:
The differences are actually pretty large. A 25s difference between bot and top is enough to allow for uncompensated roams. And trying to take the midlane first usually means giving up turret gold for 2 additional minutes compared to taking botlane turret.
Dragon as First Objectives
We’ve seen that dragons were often taken as a first objective, tying the botlane for the most commonly taken first objective (combined, they represent a whopping 59% of all the first objectives). However, there is more than one type of dragon. The four types of dragons have equal spawning chances, but do players prioritize some of them more than others? To answer that, we’ll plot the evolution of the number of first drakes taken by type and see if there are any differences.
There’s a massive difference in prioritization of the dragon as a first objective depending on the dragon type. The prioritization of the Infernal Drake being so different from the other three, which seem to be relatively identical, warrant further analysis of the impact of a first Infernal Drake on the objective focus.
The Infernal Drake factor
In order to better see the impact of the infernal dragon on the game, we’ll re-evaluate the overall distributions between turrets, dragon, and herald depending on the dragon type.
This pretty much confirms the massive impact of an Infernal Drake on the early objective focus. However, we can do further analysis. From the previous graph plotting the evolution of dragons taken as first objectives, we can see that there are differences in the times at which an Infernal Drake or the other dragons are taken as a first objective.
To see this a bit more clearly, we’ll re-plot this graph by grouping the Ocean, Mountain, and Cloud Drake together, and looking only at the distribution instead of both the distribution and the amplitude.
From this graph and table, we can clearly see that the Infernal Drake is usually taken earlier than other drakes. However, knowing when a Dragon is usually taken is not really a metric on which players can act.
Cumulative Distribution Functions
For neutral objectives, the main way to ensure control of the objective is by warding it, so that you are aware of any attempt by the enemy team, and are able to respond if possible. However, just knowing when an objective is usually taken does not provide much information into when you should ward the objective. A better metric is to see how many of the total number of first Drakes have been taken before the n-th minute. Indeed this metric gives you a clear understanding of the risk you take by starting to ward only at the n-th minute. This can be done by looking at the cumulative distribution function of each Drake type.
This is the juicy chart of this article. It gives you a clear idea of when you should start warding depending on how much importance you give to the dragon. If you are fine with missing out on 40% of the Cloud Drake due to the fog of war, you can start warding the dragon pit around the 9:30. If you prefer to have an overall control of 80% (or 20% fog-of-war loss) on the Mountain Drake, you should start warding around the 8:00 mark.
Since the Infernal Drake has a clear impact on the prioritization of the dragon as a first objective, it also has an impact on how many dragons have been taken over time. This means that if you usually ward around the 8:00 mark for all types of dragons, you’d have an 80% control on “normal” dragons, but only 63% control on infernal drakes (or 37% loss). To maintain that 80% control, you’d need to ward the Infernal Drake starting the 7th minute. Please remember that when we are talking about control in this section, we are only referring to having vision of attempts, not actual ability to contest these objectives. This involves pressure control, who’s winning where and other factors that are outside the scope of this article.
Herald as a First Objective
Since the Rift Herald is also a neutral objective, its CDF is also a good metric to tell when one should ward it.
In the event where no objective was taken yet, if you want to have an 80% control on Rift Herald, you should start warding it around the 11' mark, but that’s still ~20% Rift Heralds lost during the first minute. This means that a lot of Rift Heralds are done right as they spawn, which hints at a stronger jungler presence topside around that time, if no other objective was already taken. We’ll explore how players’ teams value Rift Heralds in general in a section further below.
Outer Turrets
Until now, we’ve focused only on the very first objective a team takes. We will now remove this “first objective” constraint and see when Turrets are usually taken. Looking for differences in the way turrets are taken can give us more information about how teams value the different objectives and when they value them the most. This, in turn, should give us hints about player movements across the map. So, is there any differences?
Well, again the botlane seems to fall earlier than other lanes, with the midlane being the last to fall.
Interestingly, we find the same time differences between the lanes as we did when we looked only at turrets taken as first objectives, with the Bot Outer Turret being taken about 30–45s earlier than the Top Outer Turret, and 2 whole minutes earlier than the Mid Outer Turret. We can also plot both graphs together to see how much of the Outer Turrets the Turret Taken as First Objective represent.
Another reason I plotted this graph is because the data from the analysis of the first objective was not obtained with the same filters as the outer turret data, and since there were issues with missing matchIds, I wanted to check if I was spewing nonsense since the beginning of the article, which would have been true if I had more of any turret type taken as a first objective than there were of any turret of that type taken at that point in time, which could have been seen by having a dotted line above the corresponding solid line. However, it seems that the data is consistent with itself (whew.). This type of double check is very important if you want to do serious stats, because interpreting invalid/meaningless data is worse than failing to interpret correct data.
Blue Side Advantage
We’ve already seen that there were differences between each side in focus on first objectives taken. Does this translate after the first objective? I’ll spare you the unhelpful charts that maybe show a slight advantage and go straight to the table that compares the average time it takes for outer turrets to fall, as well as how likely turrets are to fall.
There are some small differences, but are they really significant? Could they be only artifacts due to the randomness of the sampling? Well, this question is relatively simple to answer for the average time, as these distributions have a relatively goodness-of-fit for normal distributions, which means that a t-test should give us the answer. For the frequency differences, we can approach this by seeing it as the summation of a variable that follows a Bernoulli distribution over many samples, which should behave as a normal distribution by the central limit theorem, and so we can do a simple null hypothesis test on this value.
As we can see for almost all metrics, the blue side holds a significant advantage over the red-side. The only non-conclusive difference is the timer difference between the time when blue side takes the bottom outer turret, and the time red side takes it.
So it seems that the blue side takes the mid outer turret 10s earlier, and the top turret 25s earlier and that the blue side team takes 1 to 3% more outer turrets than the red side team. However, this does not necessarily mean that the red side has weaker turrets than blue side — it doesn’t. This rather suggests that whatever hidden advantage the blue side is visible through an analysis of the objectives taken during the mid game, and therefore that at least some of that advantage predates the mid game.
Rift Heralds
We’ve already seen how players value Rift Heralds as a first objective. In this section, we’ll look into how valued Rift Heralds are in general instead of just as first objectives. In order to do so, we’ll look at its cumulative distribution function.
As we can see, the focus on the rift herald is much lower in the event where teams already got an objective. In practice, to cover 80% of Rift Herald, you’d need to ward starting the 13' mark, as opposed to 10% of all Rift Heralds falling in the first minute when no other objective has already been taken.
Objective Sequences
So we’ve looked into differences around the first objective, and then looked into timing differences for outer turrets. Another approach to understanding the mid game is not to look into timings but rather sequences, which are orderings of objectives, although this is markedly harder to make sense of. In any case, remember that the analyses that are presented afterward are done on aggregations of sequence, and are still not absolute truths — it’s still only stats.
Process
In order to order events, I first took all events that involved destroying an objective, like dragon kills, turret kills, etc. I then grouped them by matchId and side and ordered them based on their timestamp (i.e. when the event happened). This allowed me to create sequences of actions and count how many times this specific sequence of actions happened. To my surprise, over 20k matches, there were very little duplicate sequences (like 0.2% or something), which suggests that players rarely experience the exact the same sequence of objective taken.
Rift Herald Sequences
The Rift Herald is quite the unique creature in the Rift, as its presence is strongly constrained by time, appearing at 10:00 and disappearing at 19:45. By analyzing when in the sequence of objectives the Rift Herald is present and disappears, we can infer information about how many objectives are usually taken during the 10 to 20 minutes period. This, in turn, can be useful to learn the relative pace at which objectives are taken. So, when does the Rift Herald usually appear in the sequence? Is the Rift Herald usually the 2nd objective to be taken by a team, or is it closer to the 10th?
Rift Herald disappears around the 6th objective (the tail after the 6th — 6th excluded — only represents about 4.6% of all sequences). This suggests that teams usually take at most 6–8 objectives before the 20-minute mark. We can confirm this by actually plotting the graph representing the number of objective taken before the 20 minute mark:
This puts the average number of objective taken before 20 at 3.1.
Turret Sequences
We’ve seen that side turrets were strongly favored as first objectives. Using sequence analysis, we can look into how this focus evolves as objectives fall.
As we can see, as soon as we’re past the first objective, the focus moves from the side lanes to the middle lane. Since the difference is so large for so long, it suggests that teams tend to take the side outer turrets and the two mid turrets. There’s a small slump of the midlane focus around the 8th–10th position in sequences, which could suggest that this is the period where players even out the difference between the side lanes and the mid lane. However, this is just speculation.
To try and answer this, we can plot on the same chart the distribution of all inner turrets, as well as the mid inhibitor turret.
It seems that indeed, the mid inner turret tends to fall earlier than other inner turrets, but that, as a whole, inner turrets tend to fall before mid inhibitors, which combine with some additional context about how many turrets are taken over time, by turret type and by lane, tends to confirm this intuition.
Dragon Sequences
We’ve seen that the presence of an Infernal Drake had a strong impact on the first objective focus. We’ll now see if that early focus carries over during the rest of the game.
It seems that, indeed, the Infernal Drake has a prolonged impact on the prioritization of teams, with the Infernal Drake appearing more frequently earlier in the objective sequences. However, it seems that this stronger focus only carries over during the first 5 to 7 objectives, with the Infernal Drake only being marginally taken more often. This may suggest that whichever team is in control past 5 to 7 objectives has overall control on the remaining dragons.
A quick look into the late game: Baron vs Elder Drake
When do teams prioritize Elder Drake over Baron? We’ll first look into the position distributions for these two objectives.
Although this was already common knowledge, Baron is usually taken earlier than Elder Drake. This makes the Baron a mid-to-late game objective, and the Elder Drake a late-to-very-late game objective. However, this graph has a severe flaw: it only graphs the distribution of each objective, not its amplitude. So, to complement this graph, we’ll also include a graph that represents the proportion of Baron vs Elder over the Position in Sequence when the sequences for that position included either a Baron or an Elder. This will give us a look into when players favor which.
Now that’s pretty interesting! It seems that as the sequence of objective progresses the Elder Drake becomes more and more of a factor. However, it seems that you’d need a massive amount of objectives already taken to make the Elder Drake more attractive than the Baron, and games with that many sequences are rare, very rare. This is the reason why the graph starts to oscillate as we increase the position and why we cut the graph at the 18th position — because the sample size became smaller and smaller as we progressed.
Finally, another way to look at this data is by looking at how many Barons or Elder Drakes are taken every minute of the game, either as raw number, or as a percentage of the number of games.
The first observation we can make is that the average baron per minute is relatively constant as a percentage of games, if we disregard a minor cycle with a 7 minutes period. However, the number of barons falls drastically as time progresses (from first graph). This suggests that baron objectives and end of game are strongly correlated, but also that number of games that reach past the 35 minute mark is actually pretty small. A quick check of our match data indeed put it at over 86% of games finish before 35 minutes. Another observation we can make is that in the few remaining games that go past the 35 minute mark, the Elder Drake is taken as seriously as Baron by players.
The “Standard” Sequence
When taking into account dragons, early game behavior, Rift Herald and Baron patterns, we can try to piece everything together into a “standard” order of objectives. Note that this is still statistics about how people play the game. This is not even about how to win the game, just about how people play. Therefore the following sequence is not the key to winning game, it’s approximately what most people tend to do:
- 1st Drake
- Bot outer turret
- Top outer turret
- Mid outer turret
- Rift Herald
- Mid inner turret
- Drake
- Bot inner turret
- Baron
- Top inner turret
- Mid inhibitor turret
Of course, the exact order of each objective is not set in stone. The bottom outer turret can fall before the 1st drake is taken, and so can the top outer turret, as we’ve seen from the first objective analysis. This is more to be regarded as a general cheat sheet about what the next objective should be, and where players should move, which can greatly improve the mid game decision making.
Conclusion
This was a lot to process (and to write). We learned how the early game jungler pressure can shape which objective is prioritized. We’ve also seen that this priority goes out the window the instant an Infernal Drake spawns. We’ve discovered which turrets usually fall, and in which order. We’ve learned when to ward to cover Drakes correctly, and also for the Rift Herald. We’ve noticed how the “hidden advantage” blue side has transpires into the control of outer turrets. We’ve also explored sequences of objectives and learned a general rule of thumb on how many objectives are usually taken during the mid game. Finally, we’ve seen that the Elder Drake spawns too far into the game to matter for most games and ventured a guess as to what the most common sequence order is.
In the next article, we’ll go back to the early game and mid game and try to find when people back and what that entails in terms of purchasing ability.
Read more about this series (and bonus picture)
- League of Legends: What we don’t know about the Meta-game
- The Treasure Chest of League: Riot APIs
- Building a League data scrapper
- Gold, XP and Role
- Jungling, Kills, and Deaths
- Mid Game: Objective Control (this article)
- Back Timers
- Warding and Vision Gaps
- Putting it all together: Jungling Paths
