Replay Analysis Part 3: Battle Result

Welcome to Part 3 of my Replay Analysis, where I analyse over 66,000 replays sent to me by you. Today we will focus on the Battle Result.

In Part 1 we looked at the Matchmaking Spread (were you bottom tier by two, by one, or top tier? etc.) and the results showed that equal matchmaking is far more common than we might realise.

In Part 2 we looked at the Battle Duration, visualising perfectly the 5-7 minute games we all love so, so much.

In Part 2 I used a different method to filter the almost two million data points and have found it far more efficient in filtering out rouge data that may skew the results, such as Ranked Battles (only equal spread), Assault games (ten minute timer) and more. It has also made it easier for me to gather the data, so I will return to Part 1 soon and republish it with the updated data set. I’m not expecting the results to change drastically, but it will provide a more accurate understanding of the Matchmaking Spread.

Once the initial round of analysis is complete (currently at least one more for the Range) I will return to each area with an updated data set and have a much closer look at the statistics. I am hoping these will look in detail at specific aspects, such as taking a closer look at preferential matchmaking vehicles, comparing battle duration with and without SPG’s, or even pre and post update 1.13 (SPG and HE changes). For this, however, I would like a far larger data set – if I could add a 0 onto the end of the 66,849 replays I already have, this would be tremendous. You can still send me your replays – for details please see this post. Thank you!

Due to Assault and Encounter battles being optional, and their different battle durations and play-style, this analysis will focus on Random, Standard Battles, but Assault and Encounter will receive honourable mentions at the end.

First, some definitions... Click to Expand
  • Matchmaking Spread
    • Bottom Tier by Two: -2 (e.g. A Tier 6 vehicle in a battle with Tier 6, 7 and 8 vehicles)
    • Bottom Tier by One: -1 (e.g. A Tier 9 vehicle in a battle with Tier 9 and 10 vehicles)
    • Middle Tier: -1 / +1 (e.g. A Tier 7 vehicle in a battle with Tier 6, 7 and 8 vehicles)
    • Equal: = (e.g. A Tier 8 vehicle in a battle with only Tier 8 vehicles)
    • Top Tier by One: +1 (e.g. A Tier 4 vehicle in a battle with Tier 3 and 4 vehicles)
    • Top Tier by Two: +2 (e.g. A Tier 10 vehicle in a battle with Tier 8, 9 and 10 vehicles)
  • Battle Duration
    • Whilst it is recorded in seconds, I round it to the nearest minute and display the minute (e.g: 470/60 ≈ 8)
    • All figures relating to duration in minutes are decimals, and not minutes:seconds (e.g. 7.21 minutes is 7 point 21, not 7 minutes and 21 seconds, unless I specifically list both values)
  • Battle Result
    • Defeat, Draw, or Victory
  • Range
    • The number of surviving team members minus the number of surviving enemies (e.g. A 15-5 win would have a range of 10, whereas a 7-10 defeat would have a range of -3)
    • It is possible to have a negative range in a victory (e.g. if your team captures the base to win the battle, but has less remaining vehicles than the enemy)
    • The Range has been tricky for me to define. For the purpose of this post, we are only looking at one aspect (Battle Result), which is why I have included it. The issue I have here is it only approaches it at a surface level: a 15-14 win has a range of 1, but so does a 3-2 win. This is why I need to approach the Range in a different way in the future, I just need to work out an efficient method.

Of the 66,849 replays submitted, 35,030 were Standard, Random Battles and contained the victory condition data. 18,159 were played on the EU servers (51.84% of all battles) and 16,871 (48.16% of all battles) on the NA servers. The victory rates were 48.96% and 49.89% respectively. North America had the highest draw frequency of 306 (1.81%), even though it had less overall battles.

Most EU battles (31.46% of all battles) were fought on the EU1 server, and for NA (44.13% of all battles), the US Central server. Both servers had a respective 49.92% and 49.84% victory rate. US Central has more than twice as many draw results as EU1 (288 to 116).

Tier 10 was the most popular (42.96% of all battles), 64.15% more Tier 8 (22.10% of all battles) which in turn was 41.61% more than Tier 9 (14.48% of all battles). Ignoring Tier 1-5, Tier 9 had the best victory rate (50.34%), closely followed by Tier 8 (50.27%). No other Tier had more than 50% victories. Tier 10 had more draws than all other Tiers combined (272 to 244).

Equal matchmaking Spread was the most common (37.57% of all battles), with -1 / +1 being the least common (7.43% of all battles). Equal matchmaking had the highest victory rate (50.76%), but +2 matchmaking had the lowest of only 48.97%. It also had the highest draw rate of 1.86%.

A Range of 8, 9, -9 and -8 were the most common (5.76%, 5.70%, 5.56% and 5.54% of all battles respectively). The closer the range to 0, the more likely the battle will result in a draw. The lowest range to achieve a victory was -6, and the highest range to achieve a defeat was 8. The table visualises the 15-5 (or rather, 15-7) simulator perfectly.

The most common Battle Durations were six, seven and five minutes (23.19%, 18.72% and 17.88% of all battles respectively). Only 3,351/35,030 (7.49%) of all battles lasted 11 minutes or longer, but 42/516 (8.14%) of all draw results were not due to the battle time expiring - a higher value than expected. The charts perfectly visualise the five (or rather six) minute simulator.

Excluding Safe Haven, as it is a new map, and Province and Widepark, as they are exclusive to lower Tiers...

Mannerheim Line had the most battles played (3.46% of all battles), and Berlin had the least (1.31% of all battles). The highest victory rate was Redshire (55.04%) and lowest was Prokhorovka (45.88%). Prokhorovka had the highest draw rate (5.21%) and only 0.15% of battles ended in a draw on Ensk.

In Conclusion...

There isn't a huge amount to look at on the surface of win rates, however there are some surprising results, such as +2 matchmaking having the lowest victory rate of only 48.97%, and 8.14% of all battles ending in a draw did so before the timer ran out. The statistics further visualise the average Range and Duration of a battle: 15-5 (or rather, 15-7) simulator and five (or rather six) minute simulator are prevalent in these results.

When I return to this data set to scrutinize it even more than I have done today, I would like to take a closer look at the battles from the NA region. I am interested to investigate why it has significantly more draw results than EU did, with fewer overall battles. I suspect it is because a large number of battles on the NA region were in an SPG, and SPGs appear to have longer games, or at least, survive for longer on average.

Thank you for reading. It'll likely be a little more time until Part 4, as I still need to figure out how to represent the Range data more efficiently, and I'll have university deadlines to meet sooner too. In the meantime, I have thrown together some tables for Vehicle data, and Assault and Encounter. You can find them below.

Battle Results by Player Vehicle - Click to expand

Assault and Encounter - Click to expand

Leave a Reply

Generic selectors
Exact matches only
Search in title
Search in content
Post Type Selectors