Two versus one, three versus zero, double jungling, quick tower pushes and crying top-laners. Lane swaps are a huge part of the professional League of Legends meta. They have been around for a while and have evolved over time. Hard data on them is scarce and as a result, analyzing them usually involves hours of VOD reviews. We are about to change that! In this post, we are introducing our own lane swap section on League of Analytics. If you want to head straight over to the stats page, here you go. To understand how we generated the lane swap variable and for some deeper insights, you should keep reading.
The lane swap variable
Lane swaps are tricky to determine with the data from the match histories, but I think we have found a way that I feel confident accurately recognizes at least 95 percent of swaps. As long as we do not have a perfect definition of a lane swap, there will always be games that walk the line between lane swap and standard lanes. Teams sometimes switch back and forth, resulting in some form of semi lane swap. In those cases, we had to make a decision on where to split the two.
How do we do it? While towers falling early are a clear sign of a lane swap, I think the determining factor, and the best way to recognize them, is the position of the ADC. As long as the red- and blue-side ADC are in the same lane, it is not a lane swap (whether they are top or bottom). With the help of available x- and y-coordinates for every full minute timestamp, we created “boxes” around ADCs by calculating the differences between the two coordinates of both ADCs and checked whether the opposing ADC is within that box at the 2, 3, 4, 5 and 6 minute mark (box defining x- and y-differences are roughly the length of half of Summoners Rift). If out of those five minute-marks, one ADC is within the box of the other ADC 3 out of 5 times or more, it is not a lane swap. If 3 out of 5 times or more the ADCs are not within each others boxes, it is a lane swap.
Disadvantages of our definition are semi lane swaps and a weird combination of consecutive events. The first one might occur if a team swaps away from a 2v2 lane after a couple of minutes or if teams swap back into a 2v2 lane after the tower destruction and match up in lane before the four-minute mark (I could only find four instances of this happening and even then, declaring it a standard lane when this happens before four minutes does not seem that bad). Overall, I would say this is a minor problem, if any at all (depending on how you view those semi lane swaps).
A weird combination of consecutive events could theoretically identify a game that is definitely not a lane swap as a swap. If Twitch ganks mid-lane at the two-minute mark, then dies right before the three-minute mark and again before the five-minute mark, he will be out of the other ADC’s box three out of five times and it will thus be declared a lane swap. While these types of things are a possibility and have probably occurred at one point, I could not find any such instances when checking the data and VODs.
Advantages of our definition is the flexibility it allows due to the focus on ADCs and the overall identification rate. While the first towers in lane swaps almost always fall before the five-minute mark, 2v1s or botched tower kills do happen, and our definition still fully captures those scenarios as lane swaps. Given possible meta changes in the future, this flexibility might be even more important. Overall, we feel like the ADC’s map movement is the best indicator of lane swaps. Top-laners double jungle, leach experience and are all over the map, making their position on the map unsuited for a lane swap definition.
The most important advantage is the overall identification rate. While I have spent a long time tweaking our identification strategy, we have also watched many VODs and have yet to find any obvious flaws or misidentifications. While, as I mentioned above, there are probably very few games in the data where the strategy fails, it seems highly unlikely that they reach a number that is troublesome for the analysis of the metric.
Stats summary and analysis
In our lane swap statistics section, we will provide basic info for each LCK, LMS, EU LCS & NA LCS team such as lane swap rate (LSR) and win rate in swaps vs. no swaps. Furthermore, we give data on experience and gold difference @ 6 & @ 10 minutes for teams overall, and experience and creep score difference @ 6 & @ 10 for the top-laners. While this is a lot of specific data and might be a bit confusing at first, we thought we would provide something that allows people to analyze the data themselves.
Below, I will dive a bit deeper into the lane swap data pool to show you some interesting insights into your favorite teams. Overall, the LSRs as off July 4th in the four regions we track are: 54.0% EU, 54.2% NA, 36.9% LCK, 30.6% LMS and 46.8% combined. Here is a graph showing the average experience and gold difference for EU, LCK and NA teams (LMS has small sample sizes with too many outliers, so I left it out).
As expected, the two (experience and gold difference) usually go hand in hand. Notable exceptions are Team Envy and Jin Air Green Wings, who are even or ahead in gold, but behind in experience, and SKT and the Afreeca Freecs, who are behind in gold, yet ahead in experience. We can also see that some NA teams (IMT, TSM and C9) seem pretty dominant in lane swaps, whereas others (FOX, P1) suffer. While this is an interesting insight, it obviously does not tell the whole story. We want to know how teams do in lane swaps compared to standard lanes. In graph 2, we will see that Team SoloMid is similarly dominant in either scenario (in terms of gold difference @ 10), while Immortals, Cloud9 and JAG gain a bigger advantage in lane swaps compared to standard lanes.
Then there are Teams like Fnatic (65% lane swap rate) and CLG (71% lane swap rate), who struggle in normal early-game set-ups, but manage to get lane swaps in most of their games to (somewhat) make up for it.
Looking beyond the early game, there are those teams that do well in lane swaps and those who do not. In graph 3 below, you see the lane swap rate plotted versus the difference in win rate between lane swap and standard games for the bottom ( 35 % win rate) and top ( > 65% win rate) teams. If that difference is negative, a team is doing worse in lane swaps and vice versa.
There are some obvious data points here. Phoenix1 for example is one of the teams with the highest lane swap rate, yet they have performed a lot better in the few standard set-up games they played. They are either an extreme outlier and got “lucky” in those games, or P1 should work harder at trying not to end up in lane swaps. Other teams like Echo Fox or Longzhu — two of the bottom teams in their respective regions —seem to benefit from lane swaps and also manage to get them in more than half of their games (FOX has an 11% win rate in swaps vs. 29% in no swaps). We can also see that IMT, a team that did better in the early game during lane swaps (in terms of gold difference), has a higher win rate in no swap scenarios. They only have five losses overall, so we have to be very careful with the interpretation. This is a general take away. While this data can offer great insight and point us in the right direction, sample sizes are limited, especially for those with very high / low lane swap rates.
Overall, it would be useful to check for strength of opponents when comparing between lane swaps and standard lanes, but I will do that some other time. Furthermore, including information on dragons and towers would be useful and we will try to add those to the stats page as soon as possible. My goal was to give a first insight into the data pool and highlight the usefulness of our lane swap statistics. As mentioned above, I will write an article about top-laners and lane swaps soon, for now you can check those stats (and the team stats) out yourself. Just head over to our new lane swap statistics page. If you have any feedback, find errors in the data or have suggestions on how to improve our identification strategy, please let us know.
If you like the content or are interested in stats updates, new articles or analysis, be sure to follow us on twitter.