Since the beginning of time, competitive players have always looked at faction win rates to give them an idea of which armies are performing well (or badly).

A faction with 55% is doing well, and a faction with 45% is struggling. Anything between those two figures will be considered balanced.

Win rates are always going to be the first figure we look at when judging a faction. But we need to look further than just win rates (which I’ve harped on about in the past on a few occasions).

How Much of the Win Rate Comes From the Players?

Right now Hedonites have a win rate of around 57%

But the players that have taken Slaanesh are coming to the table with a higher WoElo than their opponents. So even before taking their faction choices into account we would expect Hedonites players to win 54.7% of the time anyway.

This hopefully shows why I’m a little uncomfortable by just relying on the win rate as an indication of faction strength.

WoElo v2

The origin of WoElo was based on player matchups, and this was the case up until May last year when Aida (our stats bae) made the decision to withdraw from the Age of Sigmar community. If you know her, wish her well, it was her birthday recently!

But with Aida leaving it meant we didn’t have our dedicated stats scraper anymore and I was back to doing things by hand (I.e. going into every event and copying them down on excel).

Recently I got myself a new laptop and instead of doing things on my phone like I usually do, I was able to write my own scrapers. So I’m back in a position where I have player v player data.

So now, before a game takes place, WoElo gives us an estimate of how we would expect those two players to perform based on their previous results. This also gives us a player strength baseline for each faction.

Now we can ask:

How often would we expect these players to win?

Which then also lets us ask:

Did the faction perform better or worse than that expectation?

And this is where it becomes more interesting.

Does Player Strength Affect Faction Win Rates?

This is what I really wanted to get to.

It’s easy to say that strong players gravitate towards particular armies, and it’s something we always hear when talking about the armies in AoS.

But can we see it?

There are some interesting things in there. Cities of Sigmar are the most popular choice amongst this group, accounting for 9.1% of the players, followed by Disciples of Tzeentch and Sylvaneth at 6.3%.

Hedonites, despite the strong win rate, are only eighth at 4.7%. When I say Hedonites players were expected to win 54.7% of their games, I’m not saying that Slaanesh is full of elite players. WoElo looks at the players involved in each match. Based on their ratings before those games, the Slaanesh players were expected to win 54.7% against the opponents they were paired into.

We can answer who are the stronger players choosing? We can see they don’t really care that much at over 500+ Elo. Yes Cities is perhaps high, but otherwise there doesn’t seem to be much of a pattern in where those players are heading. Our meta leaders Fyreslayers are ranked 19th on this list.

I tested it by limiting the match data where the player and their opponent were close in WoElo, within 75 points of each other.

Does Player Strength Matter?

This is the bit of the article I’ve been looking forward to getting to. I tested the WoElo system by looking at the difference between two players before each game and whether the higher-rated player won.

The bigger the difference, the more reliable WoElo becomes.

WoElo Difference	Matches	Correctly Predicted	Accuracy
0-25	986	538	54.6%
26-50	823	488	59.3%
51-75	741	479	64.6%
76-100	616	442	71.8%
101-125	572	410	71.7%
126-150	370	293	79.2%
151-175	335	265	79.1%
176-200	252	217	86.1%
201-225	173	153	88.4%
226-250	126	105	83.3%
Over 250	342	319	93.3%

Does WoElo Work?

If I’m going to use WoElo to estimate how the players should perform in their games, then WoElo should tell us who is iikely to win a game.

So I took the difference in WoElo between the two players before each game and checked how often the higher rated playeractually won.

When two players have a difference in their WoElo of 25 or less, the higher rated player wins 54.6% of the time, not much better than tossing a coin, which is what I was hoping to see when the WoElo thinks the two players areevenly matched.

As the difference increases the results get more predictable. At 51-75 points the higher rated player wins 64.6% of games. At 126-150 they win 79.2% and when we get to over 250 points difference, the higher rated player wins 93.3% of their games.

It’s not saying that the higher rated player will always win, and I wouldn’t expect them to, as armies, matchups, battleplans and the decisions made during a game all matter.

This means I can use the ratings difference to work our what we expect to happen before the game is played.

Knowing the higher rated player wins more often is useful but not enough for what I want to do with WoElo. I want to use WoElo to work out an expected win rate before the game. So if WoElo says a player had a 60% chance of winning, I want to know whether the players in that position actually do win 60% of the time. Otherwise saying that Hedonires expected to win 54.7% of their games wouldn’t mean much.

So I grouped players by their probabilty to win the games and compare what WoElo expected to happened with what actually happened.

I was hoping to see:

Players given a 55% chance win roughly 55% of the time.
Players given a 65% chance win roughly 65% of the time.
Players given a 75% chance win roughly 75% of the time.

It isn’t perfect but if the observed win rate follows the expected then I can be more confident that WoElo will give us a useful measure of how likely that player was to win.

WoElo Predicted	Mean Prediction	Actual Result
0-10%	8.4%	0.0%
10-20%	16.0%	6.9%
20-30%	25.7%	17.2%
30-40%	35.4%	29.0%
40-50%	45.4%	40.3%
50-60%	54.4%	59.1%
60-70%	64.6%	71.0%
70-80%	74.3%	82.8%
80-90%	84.0%	93.1%
90-100%	91.6%	100.0%

It isn’t perfect but the pattern is there. As the WoElo predicted chance of winning grows, the actual results increases with it. The results are slightly more extreme than WoElo predicted and the higher rated player tends to win more often than expected.

WoElo is conservative when estimating the difference between the two players, which is useful but it still shows that the result follows what the expected predicted.

Using it with the Factions

Looking at the September results we have so far Idoneth is a good example with the players having an expected result of 52.2% but an actual win rate of 61.4%. While Hedonites had an expected of 52.9% but an actual of 43.3%.

But this doesn’t mean the difference is entirely caused by the faction. WoElo controls for player strength, not the other aspects of the game such as matchups, battlepland or army builds.

What this means

We shouldn’t stop looking at win rates. They’re the easiest way to get a quick picture of the meta. But I think we need to be careful about what we take away from them. A faction winning 55% of its games doesn’t always mean the faction itself is responsible for all of that 55%. Who’s playing them matters.

WoElo should give us a way to separate this. It’s not perfect and I don’t think a perfect version exists. But it it can tell us based on the players they faced, how well should the factions have performed? Then, did it do better or worse than that?