About a month ago, I did an analysis of the 40,000 game anonymized OCTGN dataset. Particularly, I was looking at what game balance was like for the most experienced players.
I took the 1% of players with the most amount of games played (called “the regulars”), and looked at their performance a) against the whole field, and b) against each other. There were a few take-home messages from this analysis.
1. The regulars have a high win rate against the field (duh). However, it is higher when they play as runner (68%) than when they play as corp (57%).
2. When regulars play against each other, there is a tendency for the runner to win (58% of the time).
3. Gabriel Santiago does seem to be influencing these results. When one of the regulars plays Gabriel, the runner wins 65% of the time. When looking at the other runners, the runner only wins 55% of the time. This is still reliably different from a 50/50 split between corp and runner wins, but it certainly is not in the 60-70% often-reported runner win rate for experienced players.
I was a little surprised with these analyses at the time. The runner was favored when regulars play against each other. The runner win rate (discounting Gabe) was lower than I expected, but the data seemed pretty clear on what they were saying.
One of my personal gripes with the analysis was that I was not entirely comfortable with the definition of “experienced player” I was working with. On average, people with more games played are going to be better than people with fewer games played. No disputing that. However, there is more to being good than simply playing lots of games; even within the group of regulars, there was a notable amount of variation in individual performance.
I’ve recently been playing around with the OCTGN league data, and while doing so it occurred to me that I do not have to rely on number of games played as a measure of player skill. The 40,000 game OCTGN dataset is ordered by time and has date stamp attached to each game. If I wanted to, I could calculate a more direct measure of player performance (e.g., Elo, Glicko).
So, I calculated everyone’s Elo and Glicko-2 ratings and redid my analyses from a month ago. A month ago, we looked at performance of “the regulars” -- that is, the 1% of players with the most number of games played.
This time, we are going to look at performance of “the professionals” -- the 1% of players with the highest Glicko-2 ratings. I’ve done the analyses with both Elo and Glicko-2 ratings, and the main results come out the same. However, I prefer to use Glicko-2 in this case, because it more conservative about who is performing extraordinarily well.
So, first, let’s look at the distribution of Elo and Glicko-2 ratings for all people who have played on OCTGN. The 99th percentile Elo rating is 1698, and the 99th percentile Glicko-2 rating is 1578. Nothing too odd seems to be going on with either distribution.
So how do the professionals perform? Well, they absolutely crush the field, winning 78% of their runner games and 72% of their corporation games. These are good players, as evidenced by their high Glicko-2 and Elo ratings. These are notably higher than the regulars’ win rates against the field (68% for runner, 58% for corp).
When the professionals play against each other, there is a decisive advantage for the runner (60.6% win rate over 407 games). This is a little higher than the 58% reported for the regulars, but the difference is nothing to make a fuss about.
However, it gets a little interesting when we look at the performance of individual identities. I’m only going to look at the seven Core Set identities, since there’s not really enough data points for the other identities represented in the 40k dataset (Chaos Theory, Whizzard, HB:Stronger Together, Weyland:Because We Built It).
Ok, a couple points I want to draw your attention to:
1. When the professionals play, all corporations are equally capable of winning. Jinteki is not at a notable loss, nor is HB notably better. HB shows what might end up being a slight edge with more data, but this effect is not statistically reliable within the current dataset.
2. Gabe’s win rate is 68% over 280 games. Noise is pretty good too, sitting at 62% over 276 games. Kate really seems to be lagging at 54% over 142 games.
I wanted to make this post for a couple reasons. First, the corp:runner imbalance might actually be a little bit more pronounced among the professionals than among the regulars. Gabe is certainly a culprit. Noise might be too.
Second, the fact that all corp identities seem to perform equally well when the professionals play is interesting. I would not be surprised if NBN and Jinteki had higher barriers to entry for playing well, but that does not mean they are bad (relative to the other corp identities). If you want to step up your game, it might be worth thinking about how you could actually make these difficult factions work well.