By quincy0191

I’ve been watching competitive HoN for a long time, and after years of intensive study, I think I’ve cracked the game. Time after time, no matter the opponent, there is a consistent theme: teams are trying to win. Now that I’ve identified the objective, it becomes much easier to determine how victories come about.

In order to do that, I examined the correlation between wins and each stat I track (K/D/A, GPM, XPM, CK, CD) to find out which ones seem to go hand-in-hand with victory. For the uninitiated, correlation is expressed by an r value, and varies between -1 and 1. An r value of 1 indicates perfect positive correlation (i.e. when one value increases the other always increases), an r value of -1 indicates perfect negative correlation (i.e. when one value increases the other always decreases), and an r value of 0 indicates no correlation. So the closer to we get to 1 or -1, the more sure we can be that there is a relationship between the two factors.

Note that it’s not always easy to identify the existence of a causal relationship – that increasing one factor increases the other – nor is it easy to determine which factor is influencing the other (if it is). Consider the relationship between kills and GPM – is it easier to get kills because a player has a high GPM, or does a player have a high GPM because they got a lot of kills, or is it both?

With that hopefully giving you enough of a foundation, let’s examine some correlation values to see what affects wins the most. These are taken from team-wide statistics, so when I mention kills I mean team kills. This is done in order to obscure the effect of role – a support might die a lot and still win, while a carry probably can’t.

1. GPM (r value 0.85)

Gold per minute has the highest correlation to team wins, and if you’ll remember a couple paragraphs ago, at 0.85 suggests that it is in fact a strong indicator of whether or not a team will win a game. This probably should not come as a surprise, but the point here is more about the order. Everyone knows getting more gold will help you win a game. What is slightly unexpected is that it seems to be the biggest factor in determining whether a team will win a game.
In order to give you a visual representation of how well these two factors are correlated, check out the plot below:

Since I just use a 1 for wins and a 0 for losses, it’s not quite as clear as a straight line, but the implication is fairly obvious: once a team passes about 350 GPM, they don’t lose. We’re looking at a sample of 171 games here, and the highest team GPM that resulted in a loss was 362 in an hour-long game. The lowest team GPM that resulted in a win was 272 in a fifteen-minute game. GPM is highly correlated with time, so correcting for that means the effective gap is even smaller.

The spectrum here runs from 69.2 (KNX) to 560 (SynC), but outside of a 90 GPM range one could determine with 100% accuracy whether a team won or lost entirely based on their farm. That range included just 55 games, by the way, less than a third of the total. If I told you a team’s GPM, and you said it was a loss below 313 GPM or a win above 313 GPM, you’d have a 94% chance of being correct.

2. XPM (r value 0.72)

XPM and GPM play off each other very heavily, with the correlation between them a whopping 0.89, third-highest of any two values. So the interaction here isn’t just with XPM and wins, but with XPM and GPM (as well as K/D/A). The same could be said for the section above. Nonetheless, high experience gain is on its own extremely closely related to winning games.

3. K/D/A (r values 0.52/-0.51/0.49)

I’m grouping these three together not only because they are commonly seen as “one” statistic, but also because their correlation values are shockingly close. For kills and assists this is not extremely odd, as they have the highest correlation at 0.96, but kills and deaths and deaths and assists are very weakly correlated at 0.28 and 0.3 respectively (yes, there is a positive relationship between kills and deaths, likely due to some players being generally active and others generally passive).

Given that, one might not expect the individual r values to be so close from an absolute value perspective. Each of the triple slash stats presents a moderate relationship with winning, teetering on the edge between predictable and useless. That in itself is surprising, especially considering the relationship between K/D/A and XPM/GPM. But even generally one would imagine that racking up a high kill score would lead to victory. The numbers suggest that it does make some difference, but it’s apparently not critical.

4. CK/CD (r values 0.09/0.09)

Creep kills don’t matter. There is almost no relationship whatsoever between whether a team wins a game and their creep score. Again, this is not an expected result – teams that farm creeps well should get gold and win games, especially given what we know about GPM, XPM, and wins. And yet this graph is clearly just a jumble:

There are teams with high creep kill scores that lost. There are teams with low creep kill scores that won. Time plays a major factor in the ability of a team to rack up creeps – it completes the top three in second place at an r value of 0.91 – but that just usually means both teams are farming effectively. So it’s not so much that creep kills don’t matter, as they certainly do, but teams generally don’t distinguish themselves from their opponents by their CK/CD scores. In fact, denies have a slightly higher correlation than kills.

Conclusion

Winning in competitive HoN might look a little different now, but the overall picture should not have changed much. Farming gold and experience is perhaps a little more important than previous thought, and getting creep kills is probably more about keeping up with the other team than allowing yourself to really pull ahead. Kills, deaths, and assists will help you a bit, but they’re not really game-breaking.

Now that we’ve got a slightly better picture of what’s important, it’s easier to examine the critical inputs. Next week I’ll be taking a look at where teams get their gold from, and how that impacts their chances of winning.