Breaking Niko's rating record of 2266, Oliebol has a rating of over 2300. I'm sure he will eventually dip back down to a more realistic rating. But for right now, he is the king of the 1v1 Ladder. Congrats Oliebol!

Oliebol is a great Medium Earth specialist. But 12-14 or 12-15 vs #1s and "beat" dunga after dunga had the game won but was booted? Congrats to another one of us who was able to exploit or benefit disproportionately from the ladder's rules. More of this and the inaccurate provisional ratings from games 15 to 20 will be even less meaningful.

How many of the 2200+ ratings have been accurate so far (at least half weren't)? How many of the #1s got to 1st by actually winning every game they "won" (2+ didn't) and without exploiting the ladder's rules (1+ didn't)? At least three people have a big shiny #1 on their profile that should have an asterisk.

Ladder reform anyone? Or just rename the ladder "The Race to the Highest Rating in 15-25 Games" to reflect what is happening while others play the ladder?

I understand what you mean with the ladder being easy for new teams (and slow players) to do well. The 2v2 Ladder is just as bad, probably worse. But to back up your point, Oliebol won 2 games against people rated above 2000, 3 against people rated 1900-2000, and one against people rated 1800-1900. That makes 5-6 good wins. But he might have 3-4 losses waiting in games he is currently in, and is using the ladder to his advantage. But be that as it may, he still has a rating of above 2300 and in my opinion deserves to be in the top 5 for sure, and possibly #1. And I also don't know how Fizzer could change the system for it to work better. If you come up with a better rating system, I'm sure Fizzer would at least look it over.

Sounds to me like Gui's just jealous that people he deems as unworthy have surpassed him on the ladder.

Best bit is that most of his complaints are purely subjective and were his changes implemented there would be other equally subjective complaints made by other senior users who felt that the new system was unfair to a certain style or subset of players.

Tone down the rhetoric a bit, it's not as if Oliebol exploited the system. Oliebol won 15 games, and the system gave him a rating that did not fairly rate/rank him. You can blast the system all you want, but it seems quite out of character to blast the guy. If it were the fact he was like Billy Walsh and was parading around the forums talking about how he broke the record and is the best player ever, then I would see reason to attack him.

At least this sheds light on the systematic failure that is his ratings.

You definitely do have a chip on your shoulder, Gui. what's your deal?

A lost game vs Dunga counted as a loss and his numerous wins against 0-1500 players create more downward pressure. Oliebol wouldn't have cracked 2200 then. Ultimately, he played 15 games in which only 6 or 7 mattered. He was beaten in his game with Dunga but got a boot win. Thus, only 5 to 6 games actually mattered. And once he got that Dunga win, do you think he would surrender in a game before going 15-0?

Fizzer dislikes me and seems to avoid any changes I suggest. Plus, he has binders full of charts that look at partial problems and provide complete analysis of partial problems. So
let's just leave it be. The ladder is fine if you don't click on games to see who really won or if you don't think too much about incentives or reality.

Why is it bothering you so much? I am sure you are a very good player too.

Who is blasting Oliebol? "Oliebol is a great Medium Earth specialist."

Sir, I like optimal systems. This system is not optimal.

Dick, everything I said is based on a fact stated either in this thread or elsewhere. When Chinese gymnasts exploit the rules using fake passports, do you cheer for them? (1) Being awarded wins for games that were lost isn't the most accurate way to rate someone (though there is no good way to solve this problem). (2) Playing only 5-7 meaningful games to get a rating can be solved. Creating a system and incentives to favor the most accurately rated players over the least accurately rated would make the ratings more precise.

The more fluctuation in one's rating, the less accurate the system is in understanding a player's actual skill level. This the basis of everything. So why have a system that doesn't use this basic truth behind the ratings system to ensure the comparative ranking of players is done in a more balanced and fairer manner? Competition without optimal fairness isn't as interesting or enjoyable.

Gui, what would you propose in return for the ladder? Of course global ELO could be implemented, but ladder is the biggest membership incentive for players, so I don't see Fizzer implementing it soon. If that's the case what'd you propose as a substitute algorithm for the ladder? At least 20-30 games before you get ranked? That'd be a problem for players who play slowly/on occasion, that'd mean they'd have to play 30 games in 3 months to get ranked meaning they'd have to finish 1 game/3days on average. If something else, I'm all ears.

Compare the seasonal ladder system to the other two. How are they different? Would the seasonal ladder approach (minus the seasonal ladder inflation) work for the first 20 games on the other ladders?

Well, the seasonal ladder has the advantage of everyone playing the same amount of games in approximately the same period of time... Not sure how is it doable on a continuous ladder. I don't think it is possible to implement sth similar since on seasonal ladder all players start from 0. The more games you play the rating rises. In normal ladder it's not the case since when I join with 0 I can be paired up with people that are rated >2000. That's why i get a ridiculous provisional rating over 2000 after 3 games... In seasonal ladder as far as I see you could try to use it to your advantage and join late getting paired up with better people early on to boost your rating, but more often than not you'll just run out of time... Unless I missed something it's not doable on normal ladder.

A basic rule of thumb: Any 1v1 system that has Zibik or HHH with over 40 games played and doesn't rank them in the top 5 should be questioned.

Actually if Fizzer is reading this, could you provide us with some statistics about the ladder? For example a graph showing relation between top 5 and how many games they have played (a cumulative with options <20 20-30 30+ for the past 6 months counting daily). Also there is one more basic rule of thumb: play more games ---> worse play.

If that cumulative graph shows strong distribution towards left then yeah, Gui is right and getting into top 5 has a lot to do with accepting losses only at the right time postponing them after you get the wins.

Well how does the seasonal ladder determine ratings based on total number of games played (the issue I'm focusing on here)? Could that system work on the other ladders?

Descent with modification might lead to the "fittest" system...Fizzer has learned to make the seasonal ladders' rating systems better. Why not use what he has learned to make similar yet ladder-specifically appropriate adjustments to the dinosaur ladders?

Haha. The math boys and their reliance on graphs and charts based on a subjective determination of what should be graphed and charted. I love it.

Ok, it's like that... Ladder does 2 things, makes games and builds ratings. In order to make games ladder algorithm needs to see a rating of a player, thus we have provisional ratings. In seasonal ladder everyone starts with 0, so there is no way to get up several thousand points in 2-3 games. In normal ladder you already encounter ranked players, so algorithm can pair you up with someone very strong after your first 1-2 games. What you could do is win 1 game with a weak player, stall 4 other, get a game with players ranked ~2000, win it and get a new provisional ranking. Since the algorith still needs to put you on a provisional ranking it tries to approximate it based on 2 games, so basically your first 2 recorded wins might be against players with rating 1000 and then 2000, so your provisional ranking would have to be over 2000. That is not the case in seasonal ladder since there are no already ranked players and if you want to join late you won't be able to complete enough games. That's why average rankings go up in seasonal ladder when time passes. Again I'm asking, what would be the specific changes you'd want to see? I don't know what you exactly mean by descent with modification... About the graph part, you should be able to recognise why exactly that would be the necessary data to evaluate that statement, it it shows no deviation it will show how little impact postponing games early to get higher ranking has on the ladder. If it shows deviation it'll prove you were right, nothing complicated there.

5 Gruils of Thumb:

(1) Each game completed enables the system to more accurately rate a player. Sample size is important for the system to rate and rank most accurately.

(2) Great fluctuation from completed game to completed game means the system previously had a less accurate rating of you (due to a low sample size of completed games). Thus, the following inference: The system rates more accurately the more one plays (less fluctuation) and rates less accurately the less one plays (more fluctuation).

(3) Zibik and HHH are 2 of the best 5 1v1 ME active members of all time. Most of the time their rank is based on their having completed over 40 games. However, they do not consistently[i/] rank in the top 5 when they have completed over 40 games. My conclusion: If 2 of the best 5 1v1 ME active members play over 40 games and are unable to crack the top 5, I say the system does not measure ratings accurately enough.

(4) Best ever rankings for players who reached #1 are almost all based on that players rating after having completed less than 20 games. Counter examples: Chris, Zibik, HHH consistently have or had over 40 games played. Chris and Zibik didn't reach #1 or get their highest rating until their completed games were reduced and the system had to guess more about what their actual ability might be. HHH's highest rating ever is low compared to players who finish only 15-20 games. Is the system accurate if it can't properly rate the best players who play more games?

(5) The map and settings are loaded with luck. (See my comparison of 1v1 ME and 3v3 Europe No Cards at the bottom of the following thread: http://warlight.net/Forum/Thread?ThreadID=2832.) Thus, the more games you play, the more likely lady luck will catch up with you. Combine this rating/rank factor with (1) and (2) above, and you have a recipe for inflating the rating and ranking of players who play 15-20 games (system is less sure of your skill level AND you can better avoid bad luck) and creates a disincentive for players to go beyond 20-30 games completed (if they want to be a consistent and serious competitor for the top 5).

(6) Play 2-3 games at a time to avoid ladder fatigue and boredom so you can focus more. Though Piggy and Zibik are doing just fine with 5 at a time. So this rule of thumb is relative to your actual ability (some people need more time to think about picks/moves than others) and lifestyle (some have less time to play). I don't think this argument is the end all counter-argument to (3) above, since I think (1), (2) and (5) above better explain the underlying reasons for why Zibik, Chris and HHH don't rate/rank as well when they have 40+ games played compared to when they have 15-20 games played.

(I switched computers. Signed in as real-time account.)