Winning is Everything

How would you argue one Kill Team faction is better than another?

You could appeal to specific rules or interactions that suggest faction X has some kind of advantage over faction Y. Sure, that would be reasonable; But why would those points matter? Where is your argument ultimately heading?

If faction X is stronger than faction Y, then faction X is more likely to win a game against faction Y.

That’s it. That’s the bottom line.

Notice the language I’m using sounds a lot like probability. Comparable to saying something like,

Ignoring player skill and noise, faction X has a 60% chance to win against faction Y.

The simplest way to approach faction strength is Win Rate. Win rate is how often one faction beats another faction. Statistically speaking, it is a sample of games used to estimate the probability that one faction will beat another faction.

Calculation

With a sample of games, we can calculate the mean for a faction's win rate as follows:

Ties

There are three reasons why we treat ties as half a success.

1. Ties are not a failure, but an expression of performance parity.
2. It ensures the Expected Value of win rate will be 50%.
3. GW considers ties a valid outcome in Kill Team; two ties ought to count as one win and one loss at events.

Let's break these three points down.

In a single game between equal skilled players, a tied outcome implies both factions are evenly matched. After all, what expresses parity better than a perfect tie?

Since we are chiefly using the stats to analyze faction balance, why would we record both factions as failures when the outcome suggests they were evenly matched? It makes no sense to treat a tie as a failure.

The second point is really one of convenience. When we treat ties as half a success, the total successes distributed between players in any match will always sum to 1. Either one player receives a success, or both players will receive half a success.

Since there are two players in a game, the expected value is guaranteed to be 50%. This allows us to maintain 50% as a consistent benchmark for an average, balanced faction; 50% is a memorable, clean number that makes simple analysis approachable.

If we were to treat ties as failures, the expected value will drop below 50%, to some random percentage in the high 40s. This throws off the intuitive assumption that a balanced faction should have a 50% win rate.

Finally, if none of that was convincing to you, we can simply appeal to authority and point out that Games Workshop prefers to treat two tied games as one win and one loss according to their Tournament Companion.

This isn't a trivial point. Statistical models should always strive to reflect the reality they aim to represent. If the game is built to treat ties differently than losses, then the data nerds must as well.

The Meta

When we compare faction X and faction Y, we might care more about how these factions fare against all other factions, rather than just each other. Kill Team has many factions; to manage balance, we would want to know how well a faction plays in every matchup, not just one.

We can do this by pooling all faction X’s wins, ties, and games against all opponents, rather than just faction Y. This flattens the nuance, but it does give us a single, simple rate to gauge how often faction X is winning against the field at large.

Consequently, this will transform win rate into a Weighted Average; the number of games played against all opposing factions are not evenly distributed. Some opposing factions are more prevalent than others due to their popularity. The distribution of factions being played across events is called The Meta.

This move changes what our win rate actually measures. We’re no longer estimating faction strength with a simple, faction verse faction comparison. Instead, win rate becomes an estimate of the probability that a faction can win within a specific meta.

Should this be controlled? Well, not necessarily.

The meta may not represent each faction fairly, but it does represent reality. Player choice is responsible for the uneven distribution; The collective choices of all players shape the environment in which each individual player must compete.

In other words, the meta influences the pooled win rate, but in a way that is meaningful. A faction's popularity matters; it controls how likely other players will be matched against it. The meta is the context in which all factions must perform.

Finally, it can still be helpful to break down a faction’s pooled win rate into specific matchups. This can give you an idea for which opposing factions are having the most influence. You can do this in the Matchups tool.

Mirror Matches

Mirror matches are games where both players are playing the same faction. If you’ve been following the reasoning through this section, it should be obvious why we wouldn’t include observations that are mirror matches.

The goal is to compare a faction's performance against all other faction performances within the meta. It would be meaningless to include games where a faction competes against itself.

I’ve noticed a few people in the broader wargaming community refer to the removal of mirror matches as "true win rate". I don't recommend this; the term “true win rate”, or any True Mean, usually refers to the population mean.

Ultimately, it’s best to just throw away mirrors in faction win rates and not bother giving them any kind of name. Of course, we should keep faction mirror matches when calculating player win rates, but player win rates serve a different purpose than faction balance.

Up next, we'll cover a secondary metric useful for measuring faction performance: Placing Rate.