Math Wolf has an astonishing twenty-one(out of 27) top 50 finishes!

Meh, old news (and it's out of 28 actually, going to be 22/29 after this one with 2 53rd places as well). Which season did you exclude in the analysis?
I'm more proud to have finished in the top 100 of every single seasonal ladder in existence. Wonder how long I'll be able to keep that one up.

No surprises on the top ranks. Until recently, I would have put Ruthless Bastard and TeddyFSB as 1a and 1b of the Seasonals, but lately you are rigging everything in your favour of course (including the current seasonal?) ;-)

Although I have one improvement to suggest - adding "times participated on ladder" or "times ranked" after top 3|10|25|50 would give good comparison of ones performance.

This information is much harder to extract at the moment which is why I didn't do it yesterday. I'll look into it soon.

Avg rank is hard, did you only do a larger sample size (meaning a person who plays 1 ladder and finishes 10th is not counted)?

Yeah. You need at least five top 50 finishes to feature on any list.

Also, if anyone (like me) played the FFA season, I lost like 3 games on the last day to auto finishes because games took forever and you still had no idea who would win. I would throw that ladder out of any statistics really, as it wasn't a great measure of skill.

You're right. I'll compile the numbers again today and exclude that FFA season.

Next can you analyze clans? That could be a little harder since people like Dodo Commander have wins while playing for WG, but he retired and is not currently in WG, while other players like Timon have been in a half dozen clans probably.

WL doesn't preserve the information of a player's clan membership during that season. If I were to compile clan statistics, it would consider the player's current clan(I'm not sure if this is what people are looking for).

it's out of 28 actually. Which season did you exclude in the analysis?

That was a typo. This analysis contains data over 28 seasons. I will exclude the FFA season and recompile numbers soon though.

I would have put Ruthless Bastard and TeddyFSB as 1a and 1b of the Seasonals

TeddyFSB never crossed my mind even though I knew he had a few top finishes. Off the top of my head, I would have said MoD, Timi or timon as 1b.

Edited 7/26/2017 17:46:38

"WL doesn't preserve the information of a player's clan membership during that season. If I were to compile clan statistics, it would consider the player's current clan(I'm not sure if this is what people are looking for)."

Speaking of which, does the MDL track that? Meaning did Blitz lose all their wins from Linberson moving to GG (assuming he played on MDL, just an example)?

Nice work overall.

To appear on a list, a player must have been ranked top 50 in at least 5 seasons.

This filters out (needlessly?) performances that are already statistically significant. A completed season consists of ~20 games (with a 40 point penalty for each game short of 20).

It seems absurd to exclude Rubik87, winner of seasons I and III, from any list of seasonal winners.

Edited 7/26/2017 22:40:49

I know for a fact that if you change clans all of your stats will transfer with you to your new clan so for Linberson his stats they moved over to GG if it goes for clans and my stats moved over from Blitz to Hydra. This is i think also because they don't remember in which clan you have been in on that specific account in the past.

Leaving out Rubik might have something to do with the fact he didn't finish top 50 on 5 different seasonals or the fact he did it on 2 different accounts but neither of them met MotD his prereqs.

Although I have one improvement to suggest - adding "times participated on ladder" or "times ranked" after top 3|10|25|50 would give good comparison of ones performance.

Another complication is the question of what it means to have competed in a season. If a player played 4 games did they compete? 7? 15? Or do they need to have completed all 20?

Another complication is the question of what it means to have competed in a season.

Good point. Officially, joining for even 1 game implies competition in the season. There's a measure in place to curb potential abuse.

"players who play fewer than 20 games (due to leaving the ladder early or joining late) get a penalty of 40 rating points per game."
https://www.warlight.net/wiki/Seasonal_Ladder

Edited 7/26/2017 22:39:01

This filters out (needlessly?) performances that are already statistically significant. A completed season consists of ~20 games (with a 40 point penalty for each game short of 20).

It seems absurd to exclude Rubik87, winner of seasons I and III, from any list of seasonal winners.

It does filter out players as specified. Such a filter is necessary unless people want to see hundreds of names, at which point the data is not consumable. I can always run the analysis for a smaller set of outliers if necessary.

I do think that there should be a min bar though, when we're talking of the best performers on seasonal ladders. In my opinion, performing really well for a season or two isn't as impressive as consistent results every season(Rubik being an outlier since he actually won 2 of them).

Another complication is the question of what it means to have competed in a season. If a player played 4 games did they compete? 7? 15? Or do they need to have completed all 20?

I think playing 5 games should be considered as having competed in a season. Sometimes, players forget to leave the ladder and get their initial 4 games. If they haven't left before the 5th, it shows an intent to play(you could even say 6, since some players get 5 initial games)

Edited 7/27/2017 04:53:25

I think playing 5 games should be considered as having competed in a season. Sometimes, players forget to leave the ladder and get their initial 4 games. If they haven't left before the 5th, it shows an intent to play(you could even say 6, since some players get 5 initial games)

This is a clever criteria, but isn't a perfect measure of "intent to play". It's possible someone intentionally joined the ladder, lost their first game, and then left the ladder munching sour grapes.

Because a person leaving with only 4 games gets a 16 game penalty, it's unlikely they'll make any of your lists. So there's a already "natural" filter.
Fun question: are there any seasonal winners with less than 20 games?

Well, I have played a few ladders only because the damn leaving ladder seemed to not work, and I would be auto joined next season. So I get 4 games to my surprise. I decide to play the 4, and to see if I like/hate the template, then leave. Not because I won or lost games (which this is intended to measure), but because I hated the template and never intended to play it in the first place.

So my vote would maybe be 6 games as a intent to play. If that means a person who just left after a loss is not counted, so be it. Since the ladders are opt out (based on prior season) and not opt IN, I think you have to assume people get ladder games without wanting to VERY often.

Such a filter is necessary unless people want to see hundreds of names, at which point the data is not consumable.

There are 28 completed seasons. The list of champions is already short.

The point is the "5 seasons in the top 50" criteria is non-sequitur for making a list of champions, top 3 finishers or top 10 finishes.

Say playerA has completed 4 total seasons with 4 top 3 finishes. playerB completed 28 seasons (all in the top 50) and but has only 1 top 3 finish. playerA is a stronger top 3 finisher and should be mentioned ahead of playerB.

I agree it's important to make data consumable given our limited space here. This can be accomplished by taking the top N from inherently long lists like top 10 finishes.

Edited 7/27/2017 16:14:48

Fun question: are there any seasonal winners with less than 20 games?

Yes, Heyheuhei I believe.
The penalty was increased from 20 or 30 points per game to 40 after that season.

EDIT: confirmed, Season VIII:
https://www.warlight.net/LadderGames?ID=4007&LadderTeamID=2078

Edited 7/27/2017 16:28:55

^Math Wolf to the rescue again:) Very interesting, thanks. Whenever the forum is in need of honest, rigorous statistical analysis, whenever important bits of history are about to be lost forever, Math Wolf is sure to be found!

I believe the penalty used to be 40 points per game and after Season VIII it was increased to 65.

timon is correct.

For future seasons, this penalty is being increased from 40 to 65.

source: https://www.warlight.net/blog/index.php/2013/05/season-ix/

Would it be possible to use an either/or statement as a means to include significant results (such as two winning seasons) that are otherwise filtered out by the 5 Top 50 finishes?

Additionally, Average Rank either needs to list the total number of ranked seasons (which could then be used as a tie breaker to reward longevity), or the average ranks of everyone needs to be divided by the total number of seasons and not merely the total number of ranked seasons.

Edited 7/28/2017 17:39:55

Thanks for the link Timinator. Also interesting from that blog post about the penalty:

Instead, it’s implemented by simply giving 40 [now apparently 65] points to each player every time they receive a game, up to a max of 20 games. This is also why ratings tend to always go up as the season goes on.**

** It’s worth noting that this extra rating bonus does not impact matchmaking. Determining who you get matched against gets done before this rating bonus gets applied, so this mechanism does not penalize players who join late, as long as you get 20 games before the season ends.

Does anyone have access to modify the wiki? The page for seasonal still mentions the old 40 point penalty and doesn't mention the above implementation.

https://www.warlight.net/wiki/Seasonal_Ladder

Updated wiki

I guess MotD was too modest to update the list of Seasonal winners, so I did that as well :)

Motd, Beren, u guys are fuckin MVP of WL!