Which RAGL Season had the best line up?

[TTTPPP]

We're now entering the 15th Season of RAGL and it seems like people are always interested in how the registrations compare with previous seasons. To some extent this is completely subjective - is a season going to be good because the top players are very skilful or because it's hard to predict who will win? Is it better to have more players signing up or more divisions, and maybe we should consider how likely players are to forfeit?

I thought it would be good to try to answer this question by comparing the seasons across different metrics and, to start with, using the RAGL ladder rating system.

The RAGL ladder uses Glicko2 with a one week period and then ranks players by three standard deviations below the best guess at their skill level (mean). For example if one player has played a hundred games and another has only played one then it's possible that the best prediction is that the two players have exactly the same skill level, but there will be a lot more error in the guess at the second player's rating and so they will end up a lot lower on the RAGL ladder.

Plotting the ratings for Masters, Minions and Recruits as separate series we get the following chart:

StartofSeasonRatings.png (118.39 KiB) Viewed 5407 times

Perhaps the first thing the stands out is that Season 1 looks very different from the other seasons. This is because at the RAGL Ladder data starts at the beginning of Season 1, so all players have roughly the same rating. In order to plot something on the chart this shows the ratings at the end of week 1, whereas the other seasons are from the start of week 1.

By Season 2 we can see that some players that won matches have moved to Masters, and some players that lost matches have moved to lower divisions. The highest ranked player was OzzyOuzo in Minions, since he dominated Recruits the previous season and so has an incredible record. Similarly the highest ranked players at the start of Masters were SoScared, kazu. and abcdefg30 who all performed well in Minions the previous season.

It's not really until Season 3 that we see mixing between divisions having effect and players in higher divisions having higher ratings. However the highest rated player at the start of Season 3 is LorryDriver from Recruits in Season 2, and he's almost 200 points above FiveAces.

Seasons 3 and 4 had parallel Recruits divsions, which are shown by slightly different shades of yellow. I don't think it's very important for answering the question of which season had the best players, but I thought it was interesting information. For example in Season 4 then Recruits D contained kazu. and Zaqzorn, the two highest rated Recruits players, who were in Masters and Minions in S02 respectively. Recruits A on the other hand didn't contain any returning players.

So which season had the best players? Going purely on ratings for the top two players then Masters in S05 had Lorrydriver on 1900 and kazu. on 1872, and this was followed by S11 which had Kav on 1884 and goat on 1874. Going instead based on the lowest ranked player in Masters then S10 had Sigil on 1294 and S11 again had Sigil, this time on 1248.

The massive gap between 1st and 2nd in Masters Season 13 is Kav on 1841 followed by despro on 1562.

If we look at the fourth placed player in Masters, on the basis that this makes for a good playoffs then Season 5 comes top with Smitty on 1721. If we want exciting playoffs then maybe we want the closest top four which happened in Season 6 with barf_openra on 1682, Orb on 1678, ZxGanon on 1669 and Unano on 1620 (in fact FiveAces on 1572 made it to playoffs instead of Unano).

[TTTPPP]

Although ratings stabilise over time then perhaps it's not fair to compare a rating in Season 14 with one from Season 3. Consequently I thought it might be worth looking at the difference from the mean rating for each player. This could in theory show how much better players in Masters were than the average 1v1 player.

DifferenceFromAverageRatings.png (132.18 KiB) Viewed 5027 times

As might be expected we can see that often players in Recruits are below average (i.e. have a negative y-value) while players in Masters are usually better than average.

This graph puts Kav in Season 13 as the player who was most "better than average", while Season 11 had Kav and then goat both well above average. Players at +1000 on this graph have somewhere in the region of 98-99.5% chance of winning against the average player, while a +500 point difference corresponds to roughly 80-90% chance of winning. Season 15 is in third place by this metric, with Fazzar, Kav, happy and Blackened within the space of roughly 100 rating points. This season has the top four rated Masters players further above the average than any other season, with second place going to Season 11 (Kav, goat, happy, ZxGanon).

One issue with this method of comparison is that we only have ratings for players who competed in recorded matches, so primarily in RAGL or the Ladder. The OpenRA Ladder was only introduced just before RAGL Season 10 and so there were fewer bad players with ratings before that point. The Ladder has made it incredibly easy for anyone to get rated. I'd assume many players have become rated by accident by playing an opponent waiting in a Ladder server.

There are also still some players who aren't represented on the graph. Morkel and i like men hadn't played in a rated competition before Season 9, and so they don't have a rating on this chart despite finishing second and third in Masters that season.