So, let me start by saying I've been having a really tough time this Act.

I'm sure a lot of you feel that way too, right? Feel like you constantly get toxic/throwing teammates, disconnects, or just that there's a complete team diff and like... how did you even end up in a lobby so stacked against you? Well, what I'm about to show you is absolutely anecdotal and may not actually mean a thing, but I found it curious to say the least and it felt oh-so-satisfying to have some kind of visualized data to support my unfortunate queues as of late. I finally felt like "wow, it's not just in my head!"

Over the past few days, I've been adding data from tracker.gg to a spreadsheet I made in an attempt to make sense out of the madness that is my queues (screenshot of the spreadsheet a ways down in the post). I was pretty surprised with what I found (or maybe not so much.)

Essentially, what I was setting out to prove to myself is this simple theory: When I'm losing games, the matchmaking system prefers to match me with teammates that are also losing their games, and conversely prefers to match the enemy teammates with people who are winning their games. I realize that if this is happening, it could be and likely is an entirely unintentional side effect of a matchmaking system as complex as Valorant's.

So before I show you my findings, let me explain my logic and process (I'm not a data scientist, I'm a software engineer... so forgive me):

​

1. First, I drilled into each of my teammates profiles from each match to grab their "Act win rate %" and "Overall win rate %" for competitive games (some profiles are private, so I got what data I could and this applied to both my teams and enemy teams. Enough data had to be enough in this case, there weren't too many holes.)
2. Then, I grabbed the enemy team's win rates from their individual profiles
3. I averaged my team's act win rates and overall win rates into two separate fields
4. I did the same for the enemy team
5. I then averaged those two percentages together for each team to derive an overall win-rate-vibe for them
6. I then took the absolute value of the enemy team's win rate % minus my team's win rate %, which I called the "Win Rate Delta" for the match
7. Using that delta, I graded the "fairness" of my matches based on the delta from A to F like so:
   1. A = less than 1% overall win rate delta between teams
   2. B = less than 1.5% overall win rate delta between teams but more than 1%
   3. C = less than 2% overall win rate delta between teams but more than 1.5%
   4. D = less than 3% overall win rate delta between teams but more than 2%
   5. F = anything higher than a 3% overall win rate delta between the teams
8. Based on which team had the win rate delta in their favor, I made a prediction on who would win

What did I find?

• My "predictions" of who would win based on data were right 70% of the time for my last 10 matches
• 2 of the 10 matches that I was wrong at predicting were matches that I graded "A" (basically, fair and truly competitive matches!) -- I was actually happy to see the prediction be wrong on those.
• Only one of the matches that I predicted I would lose and graded it poorly (a D) we actually won... but let me tell you it took some serious heroics and a bit of luck to win that match

Here's a snapshot of the spreadsheet. I used conditional formatting to call out disparity in win rate % for each team clearly with green and red... clearly you see a lot of red on my teams ;)

​

Spreadsheet of my last 10 competitive matches

​

I'm going to continue adding my match data to the spreadsheet for a while to see how my theories hold up. I like to capture my data in real time instead of looking at matches days or even hours before they happened, because I think going back and looking historically at individual player win rates will skew data. After all, people's win rates change over time.

I personally want my matchmaking to be as fair as possible, and from what I'm seeing, a matchmaking algorithm adjustment that funnels down like this might result in more community satisfaction:

​

• Rank > Avg. Win Rate % Delta between teams results in grade A (or B in situations where queue times would be exhaustively long)

​

All of this said. I realize this is a small sample size at the time of this posting and I'll be happy to update at 100 matches, 1,000 matches and so on to see if any of this theory holds up over larger data sets. I'd also be happy to share this spreadsheet template to anyone who wants to add their match data.

Hopefully some of you find this interesting and hopefully as I continue to add data, I can derive more trends or confirm that this entire thing was well, just confirmation bias from a salty guy tired of losing matches. Thanks for reading!