By Using Stats, I Have Found The Trench

Warning: This is an article I’ve been sitting on for 12 months because I had some issues with it. But I thought that it’s interesting enough to  put it out there.

I lost a game as Legion Commander yesterday with 344 bonus damage from Duel. I played pretty well and still lost. It made me wonder if you could find out if you are in the Trench simply by using the data of your team, and working out how well you did. I define ‘The Trench’ as simply a place where the performance of your team-mates is consistently below your performance. Pro Players are likely to be in The Trench when they play Solo Ranked because their Solo MMR is so ridiculously high that the system places sub par team-mates on their team in order to balance the game. This puts them in a very frustrating Matchmaking Hell. For me, I tend to find that the Australian server size is so small that the pool of players to select team-mates from is too small to find good games for experienced players. This might sound a bit egotistical and ‘douchy’ – any form of positive self-skill analysis generally does – but first off you have to understand two important things.

Two Things

The first thing is that I have over 5500 games of Dota 2, including playing every hero and experimenting around. This is along with watching other people play and constantly trying to find what mistakes I’ve made, such as a few games ago where I regretted buying a Euls on Abaddon for that game. I primarily play Unranked, and I consider myself fairly highly skilled at the game, and very knowledgeable about mechanics and scenarios.

The second thing is that every day the Australian Servers only have 50 – 1,450 searching for a game at any given time. For comparison, EU West has about 1,200 – 11,500 players searching and US East 900 – 3,000. As I described above, this means that the pool of players to MM against is far smaller for me than in other places. The LC game above occurred at about 6:30 am when there was only 150ish people searching for a game, that’s an incredibly small player pool. Servers seem to peak at about 8-10pm server time every night, so if you’re really low or high skilled, you are likely to get the best pool of players to MM you against during this time. Which side-note: this is a very good time to played Ranked if you desire more balanced games.

How To Find If You Are Trenched

To statistically work out if I was in ‘The Trench’ or ‘ELO/MM Hell’, I looked at 100 recent games. I put my last 50 wins in one pile, and my last 50 losses in another pile. I wanted to look at wins and losses entirely separately. I used 3 metrics to determine how well I did in the game: Kill Participation, Hero Damage, and Tower Damage. For each game you rank yourself on these metrics. If you did the most hero damage on your team, you placed first on your team for this metric, so you get a 1 for Hero Damage. If you did the least Tower Damage, then you get a 5. The average should be a 3 if you are in solid matchmaking. An average of 3 means that for every game you do well in (1s and 2s), there are an equal amount of games you do bad in (4s and 5s). An abundance of 4s and 5s means you are getting MM’d with people better than you, whereas an abundance of 1s and 2s means the opposite. AKA lots of 1s and 2s means your have been in MM Hell/a Trench.


My Last 10 Losses Tell An Interesting Story

The scoreboard doesn’t tell the entire story of a game as complex as Dota, but it often includes very important plot points. For example, a Zeus who does well is likely to have High Damage and Kill Participation, but much lower Tower Damage. A Clinkz or Drow on the other hand, being basically entirely single target, are likely to have very High Damage and Tower Damage but lower Participation. Every hero – even Supports – should rate on these three metrics. Even my one Omniknight game I had, I placed last in none of the three metrics, even though this is the only hero in the entire game that my system doesn’t treat too kindly. It should be noted however, that this system is best for players who play a large variety of heroes.

The Mean-ing

When I lose, I actually still perform pretty well on average. In my 10 losses pictured above I had strong metrics for 7 games, including 5 games where I had the best metrics on my team. This is a similar story for my other 40 losses as well. I was playing Carries, Semi-Carries, and Supports in my games, yet still placed very highly.


Amount Of Places For My Last 50 Losses

An average player should have 10 on each metric, as that would mean they have as many 1s as 5s. I had the most Hero Damage on my team for 20 of my 50 losses, which is more than double what it should be. While across the board my performance was heavily skewed towards the upper end of the spectrum.

High Placings On Average Indicate A Trench

High Placings On Average Indicate A Trench

How Do My Wins Measure Up?

My 50 wins were even more incredible. When losing I often did very strongly on my team, however when I won I placed highly in my Metrics non stop. To the point where a disproportional amount of my wins were likely caused my strong performance, reflecting the sentiment people have of Ranked Matchmaking where a “dominate snowball high impact hero” will increase your MMR. The idea being that heavily carrying your teammates in any role is the recipe for success.

Overwhelmingly Performing Towards The Better End Of My Team

Overwhelmingly Performing Towards The Better End Of My Team

I had the highest kill participation in 50% of my winning games, and was the second highest in 25% of them. My Hero Damage was even better, resulting in me having the most or second most amount of damage in 82% of my games (41/50). I faltered on Towers for some wins, getting low results for a wide variety of heroes such as Necrophos, Zeus, Faceless Void, Phoenix, and Lion. I still however performed well on Towers, with 52% of my games being 1s or 2s, which is still above the ‘good matchmaking’ average of 40%. All three metrics for both winning and losing had me performing better than my teammates on average.

 Final Analysis

Performing Well In Multiple Roles

Performing Well In Multiple Roles While Winning

Scoreboard doesn’t mean everything, but it does make a powerful statement. Sure, it can’t tell me ward placing or smokes, but it still can give a good impression of game performance. All of my 100 games were before I devised this MMR Hell Classification System (I’m calling it PDT Rating), and thus I never tried to inflate the numbers during the game, which is very important. I also played a lot of different heroes in a variety of roles (as seen above). Another limitation currently is that it sometimes doesn’t factor in feeding. While this isn’t as much of a problem while winning, it can occur during lost games. However often when someone feeds, they are taken off the map for so long that they end up scoring lower on the Metrics. Although there are some problems, the rating will still give a good impression of the game. To further prove this, I went back to where it all began.

My First Few Losses Ever Are Filled With My Terrible Performance

I went back (just under 4 years) to the start of the Dota 2 Beta, and looked at the data for my first set of games. In the above image are my first Dota 2 losses, and as you can see they were dominated by me being bad. My metrics were consistently 4s and 5s as I was new to the game and was frequently the weakest link in these games. To me, this further cements the integrity and accuracy of using the PDT Rating. I remember being very self aware of how bad I was when I started, and my PDT system has correctly identified how trashy I once was.

You Can Work This Out Too!

It’s great that I’ve learnt so much over the years, but it is also disappointing where I’ve ended up. Essentially The ‘Dota 2 End Game Content’ just isn’t as enjoyable as the journey there, especially when you are on the smaller servers. I thoroughly enjoyed the journey over the years, but in a sense I currently have a diminished ability to have enjoyable pubs. You start to take the game more seriously, while simultaneously moving up the pack.

If you’ve managed to come this far in this rather lengthy article, you should be able to work out that you can do your own recent game history to see what story is waiting to be told. Your PDT Rating is awaiting to be found. First off, make sure you look at the overall average per metric, and not per game. Also make sure you separate your wins and losses into two different piles. Your data might also come out a bit odd if you play too much of the one hero, especially if they have a tendency to not do much metric-wise.

My Trench Theory is that there is no static Trench for the entire game, instead every player has their own Trench locations. A trench is simply a place where your a disproportionate amount of your teammates are consistently metrically worse than you. I believe that Dota has a mobile trench, which is relative to every individual. Using this system you can work out if you a currently in a Trench or not. I tried to evade confirmation bias at all turns, and have used a fairly large sample size of my most recent games (about a month of Dota). You too can use your last games, that were played unknowing of PDT Rating, to see how you measure up. It should take less than a game of Dota 2 to work out, which isn’t that long considering how valuable you might find the information. Being able to identify MM Hell is a beautiful thing. But it’s also something that if you say, people will give you stink eye for and say you’re just shifting blame.

Either way, I sat on this article for so long I thought I should finally put it out there, even with the whole “focus on your own game” mantra that’s popular these days. You aren’t the only variable in your games.