Amazing Video here, very relevant to Bloodbowl!

Everything dealing with the video game developed by Cyanide!
triperis
Posts: 26
Joined: 06 January 2017, 06:31

Re: Amazing Video here, very relevant to Bloodbowl!

Postby triperis » 06 June 2017, 11:08

Given you just fiddled with coefficients ever since S1 shows that..
Going to stop you right there - while that formula was used for the first two or so seasons, it was changed completely after that. It's not simply knob turning and graph shaping anymore as far as I know - it is based on very specific breakpoints.
I might have misread something then. Under my impression, it always was an idea that your winrate is nerfed based on number of games you played, with nerf disappearing at hard cap and becoming insignificant at the soft cap. Only the locations of the caps were changed, while irrespective of the form the design remained the same. While simple and clean design, it does not address common flaws I tried to present. There are a lot of mathematical tools you may add to counter them, but I totally get prioritizing it being simple and clean, especially when it's mathematical form actually is not that clean for a layman :D
If you count that at some point matching record X becomes easier from a fresh team than a developed one, it suddenly becomes advantageous to grind fresh teams until your winrate combines with luck "correctly".
That's quite an assumption... where is that point, exactly? You're basing a lot of your argument on that idea so maybe you could take a moment to explain where it lies rather than hand-waving your way past these sweeping statements?
It is actually very practical think. E.g. it's very easy to check such stuff with gobospy. For a practical example, take top current record (orc) of PC cladder, 26-3-3. Let's say I have a reason to believe I can do better. 4th best team sits on 19-3-4. They'd need 10 straight wins to better that arriving at 36 games played. Then they would also need to get more points in 6 games than the other team in 10, until both hit hard caps. Or he may win straight 19 games from fresh to better the record, staying on full-games played reward, while the other coach is on the soft cap. All at the cost of extra 9 games. Orcs are good at mid-to-high TVs, it should pretty stable there, but one might have to reroll a few early losses. This might not be the most clear example, but you kinda can see the logic of consideration when comparing your record with others. Looking back on the player with 19-3-4, he did enter 30 orc teams last season alone into cladder. While he did not benefit from it (best record only marginally better than combined), one may see that it is certainly viable way to play. And having leader who went up from S5 14-7-2 record with same race (15-7-3 if last season record combined) to 26-3-3 this season is certainly encouraging rerolling until you get major luck (in current leader's particular case S5 record he conceded once and had no concedes against; S6 he did not concede, and got 5 concedes against).

Disclaimer: Just took examples based on their performance, nothing personal meant and I do not know them.

User avatar
VoodooMike
Posts: 1614
Joined: 14 July 2009, 07:44
Location: Calgary, Alberta
Contact:

Re: Amazing Video here, very relevant to Bloodbowl!

Postby VoodooMike » 06 June 2017, 15:52

Under my impression, it always was an idea that your winrate is nerfed based on number of games you played, with nerf disappearing at hard cap and becoming insignificant at the soft cap. Only the locations of the caps were changed, while irrespective of the form the design remained the same.
No, the first few seasons (and really, there haven't been that many seasons thus far) were not based on win% at all... they were based on arbitrary curve-shaping. You can find the very friendly discussions of that elsewhere on this forum.

Current seasons are based on a % of win rate which is based on games played. There's no hard cap (that would mean that games played stops mattering) only a soft cap at which point games played doesn't contribute tons more to your ranking, but the contribution does not disappear over time.
While simple and clean design, it does not address common flaws I tried to present.
What you were suggesting was that we use the lowest calculated potential win rate for each person. That's not a clean system at all, and while I enjoy the idea in theoretical terms, it's a bad, bad idea in practical terms. It means someone can achieve a great record for a season and be ranked lower than someone with the same number of games and a worse record simply because we calculate that his good record was probably just the result of having a lucky season.

Yes we want to know who the most skilled coaches are, but there's no point in having qualifying seasons if we're not basing it on their seasonal performance... I mean, we could just have people sign up and use their past record to calculate their general skill instead, right?
It is actually very practical think. E.g. it's very easy to check such stuff with gobospy. For a practical example..
...sorry, but that's not a practical example, it's a wall of meandering text that is again using BOTH "hard cap" and "soft cap" despite the fact that there isn't a hard cap in the first place.

I'm all for a clear example if you have one. What you seem to be suggesting is that, as people did in the WC qualifiers, people benefit from rerolling teams over and over in an attempt to get a huge win streak based on luck. In WC your qualification was based on the first 20 games you played, and basically every team that qualified had a 20-0-0 record... and anyone who got a draw or loss just started over.

In present CCL you are no longer restricted to 20 games, and the point at which 100% of your win% is considered is something silly like 42 games. Additionally, because it is matched based on TVPlus rating, you can't rely on getting easy opponents and concessions to pad your wins... the more sequential wins you get the more likely you are to be paired with a tough opponent who will end your streak because unlike you, his record is not based on pure luck.

While we could refine the curves and resisting pressures further, the current system is pretty damned good... and quite superior to what you've suggested both in terms of pragmatism and even logic.
Friendly Reminder: Correlation does not equal Causation - tattoo it on the inside of your eyelids if it'll help.


Return to “General”

Who is online

Users browsing this forum: No registered users and 8 guests

cron